PATTERN TO
PROCESS
METHODOLOGICAL INVESTIGATIONS
INTO THE FORMATION AND INTERPRETATION OF
SPATIAL PATTERNS IN ARCHAEOLOGICAL
LANDSCAPES
Martijn van Leusen
RIJKSUNIVERSITEIT GRONINGEN
PATTERN TO PROCESS:
METHODOLOGICAL INVESTIGATIONS INTO THE FORMATION AND
INTERPRETATION OF SPATIAL PATTERNS IN ARCHAEOLOGICAL
LANDSCAPES
Proefschrift
ter verkrijging van het doctoraat in de
Letteren
aan de Rijksuniversiteit Groningen
op gezag van de
Rector Magnificus, dr. D.F.J. Bosscher,
in het openbaar te verdedigen op
donderdag 30 mei 2002
om 16.00 uur
door
Pieter Martijn van Leusen
geboren op 28 augustus 1962
te Utrecht
Promotor
:
Beoordelingscommissie :
Prof. dr. Peter A. J. Attema
Prof. dr. John L. Bintliff
Prof. dr. Marianne Kleibrink
Prof. dr. Kenneth L. Kvamme
Prof. dr. Douwe G. Yntema
Preface
Chapter 1 .............................................................................................Introduction
1................................................................................................................. Aims and Background
2................................................................................................................Structure of this Thesis
2.1 .....................................................................................................Methodological studies
2.2 ...................................................................................................Field walking campaigns
2.3 ........................................................................................................................Case studies
1
4
6
6
7
Chapter 2............................................................................Patterns and Processes
1..................................................................................................................................Introduction
2...................................................................................................Regional Settlement Dynamics
2.1 ..........................................................................Settlement dynamics of the study areas
2.2 ..................................................................................................Core concepts and terms
3...........................................................................................Towards Interregional Comparison
3.1 ...................................................................................Why, how and what to compare ?
3.2 ...............................................................Explanatory models of socio-political change
3.3 ..................................................Detecting macro-archaeological quantitative patterns
4.....................................................................................................................................Discussion
1
1
1
1
14
14
16
18
21
3................................................. Introduction to the Wroxeter Hinterland Project
....... Extending GIS Methods for Regional Archaeology: the Wroxeter Hinterland Project (297-304)
..............................................................Aspects of Romanization in the Wroxeter Hinterland (133-143)
4.........................Dealing with Recent Post-Depositional and Research Biases in
.....................................................................................Archaeological Landscapes
1..................................................................................................................................Introduction
1.1 ............................................................................................................Aims & definitions
1.2 ................... History & treatment of biases in Mediterranean landscape archaeology
2.............................................................................................................................. Bias Modelling
2.1 ........................................................................................................... Dealing with biases
2.2 ..........................................................................................Identification and assessment
2.3 .................................................................................................Recording and evaluation
2.4 .......................................................................................................................... Correction
3................................................................................................................Concluding Discussion
1
1
3
7
7
9
14
16
18
5....................................... A Review of Wide-Area Predictive Modelling using GIS
1..................................................................................................................................Introduction
1.1 .............................................................. Aims and approaches of predictive modelling
1.2 ........................................................................................................ Theory and concepts
2.................................................................................................................................Methodology
2.1 ............................................................................................................ Impact assessment
2.2 ........................................................................................................................ Data quality
2.3 ......................................................................................................................Data quantity
2.4 ..........................................................................................................................Extensions
3...................................................................................................................................Conclusions
1
2
4
10
10
11
13
14
16
6...............................................Line-of-Sight and Cost Surface Analysis Using GIS
1..................................................................................................................................Introduction
1.1 ....................................................................................................................................Aims
1.2 ........................................................................................................... Theoretical context
1.3 ...................................... CSA and LOSA: twin tools for cognitive landscape analysis
2...................................................................................................................Cost Surface Analysis
2.1 ...............................................................................................Principles and applications
2.2 ......................................................................................................Algorithmic confusion
2.3 ..........................................................................................................................Discussion
1
1
2
3
4
4
5
7
3..................................................................................................................Line-of-Sight Analysis
3.1 ...............................................................................................Principles and applications
3.2 .......................................................................................................Methodological issues
3.3 ....................................................... Visibility, perception, and the cognitive landscape
3.4 ......................................................................................................................Further work
4...................................................................................................................................Conclusions
9
9
10
12
13
14
7............................................................ Educating the Digital Fieldwork Assistant
1....................................................................................................................................Introdution
1.1 ............................... Improving the efficiency and accuracy of field work procedures
1.2 ............................................................... Development of a digital field-work assistant
1.3 ................................................................................................. The SIBA2000 campaign
2............................................................................................................................. The Field Tests
2.1 ............................................. Highland survey: the recording of transhumance routes
2.2 ...............................................................................(Re-) location and recording of sites
2.3 ................................ Recording of topographic reference points and collection units
2.4 .....................................................................................................................GPS accuracy
2.5 ................................................................. Spatial accuracy and the problem of identity
3................................................................................................................................Further Work
3.1 ......................................................................................Enhancing current functionality
3.2 ..............................................................................................Hardware and functionality
3.3 ............................................................................................ User and network interfaces
4...................................................................................................................................Conclusions
1
1
2
3
6
6
6
7
8
10
11
11
12
13
14
8.......................................................................The RPC Field Surveys, 1998 – 2000
1..................................................................................................................................Introduction
1.2 ........................................................................................................................Background
1.3 ........................................................................................................................ Approaches
2............................................................................................................................................Results
2.1 ................................................................................................................... Pontine region
2.2 ................................................................................................................. Salento Isthmus
2.3 .............................................................................................................................Sibaritide
3.....................................................................................................................................Discussion
3.1 ...................................................................................... Finds collection and processing
3.2 ................................................................................................................. Data processing
3.3 ........................... Interpretation: (re-) constructing settlement and land use histories
4.................................................................................................................................... Conclusion
1
2
2
3
4
4
4
5
5
6
7
8
9.......... Archaic Settlement and Early Roman Colonisation of the Lepine Foothills
1..................................................................................................................................Introduction
2........................................................................................................... The Lower Lepine Slopes
2.1 ....................................................................................................................Earlier surveys
2.2 ........................................................................................The Doganella di Ninfa survey
2.3 ..............................................................................................................Settlement history
3.....................................................................................................................................Discussion
3.1 ............................................................................................................Settlement patterns
3.2 ................................................................................................. Locational characteristics
1
3
3
4
6
7
7
9
10 ....................................................................................... A Marginal Landscape:
................. Field Work on the Beach Ridge Complex near Fogliano (South Lazio)
1..................................................................................................................................Introduction
1.1 .................................................................Marginal landscape units in the RPC project
1.2 .........................Outline of the physical and human landscape of the Pontine region
2..........................................................Evaluating the Agricultural Potential of the Landscape
2.1 ....................................................................................................Description of the units
2.2 ............................................................................................Ppreliminary land evaluation
1
1
2
3
4
5
3........................................................................................................The Archaeological Surveys
3.1 ......................................................................................................................Methodology
3.2 ................................................................................................................ Summary results
3.3 .............................................................................................................. Site interpretation
4.....................................................................................................................................Discussion
4.1 ............................................................................................ The Protohistoric landscape
4.2 ...................................................................................................... The Roman landscape
4.3 ............................................................Correlating the physical and human landscapes
5...................................................................................................................................Conclusions
6
6
8
11
11
11
12
13
14
11 .............................................................................................Walking the Murge:
........................... Interim report of the Ostuni field survey (Apulia, southern Italy)
1..................................................................................................................................Introduction
1.1 ..................................................................................... Comparative settlement analysis
1.2 ........................................................... Aims of Dutch research in the Salento Isthmus
1.3 .............................................................................................................The Ostuni survey
1.4 ......................................................................................................................Methodology
2..................................................................................... Landscape, Settlement and Agriculture
2.1 ..............................................................................Evaluation of the physical landscape
2.2 ....................................................................................................Settlement and land use
3.........................................................................................Results of the Archaeological Survey
3.1 ........................................................................................................ General observations
3.2 ..................................................................................................The Protohistoric period
3.3 ............................................................................... The Hellenistic and Roman periods
3.4 .................................................................................................... Post-Antique to Recent
4.....................................................................................................................................Discussion
4.1 ................. Centralisation of settlement in the Late Bronze Age and Early Iron Age
4.2 ....................................................................................Early urbanisation and rural infill
4.3 ...................................................................................................... The Roman landscape
5................................................................................................................. Concluding discussion
1
1
2
3
4
7
7
8
8
8
9
15
16
18
18
19
21
22
12 ........................................... Regional Archaeological Patterns in the Sibaritide:
..................................Preliminary results of the RPC field survey campaign 2000
1..................................................................................................................................Introduction
2................................................................................................................................The Sibaritide
3.......................................................................................................................... Research History
4.......................................................................................................................Patterns and Biases
5............................................................................................................. The Survey: Approaches
6...........................................................................................................................Finds Processing
7............................................................................................................................................Results
8..................................................................................................... Settlement and Infrastructure
9...................................................................................................................................Conclusions
1
2
4
6
9
12
14
19
20
13 .................. A Comparison of Archaeological Data Sets for the Pontine Region
1..................................................................................................................................Introduction
2...........................................................................................Towards an Interregional Database
2.1 ....................................................................................................................................Aims
2.2 .....................................................................................................................................Data
2.3 ............................................................................................................ Unusual data types
2.4 .................................................................................................................... Classifications
2.5 ............................................................................................................................ Fuzziness
2.6 .........................................................................................Conclusions and further work
3........................................................................... Comparing Data Sets of the Pontine Region
3.1 .......................................................................................................................Introduction
3.2 ..............Comparison within the same landscape unit: the northern colluvial slopes
3.3 ..............................................Comparison across landscape units: the Pontine region
1
2
2
3
7
8
12
14
15
15
16
19
3.4 ......................................................................................................................Further work
4.................................................................................................................................... Conclusion
25
26
14 ..................................... Land use / Land cover Bias in the Wroxeter Hinterland
1..................................................................................................................................Introduction
1.1 ..................................................................................................................................... Aim
1.2 ........................................................................................................................Background
1.3 ........................ Two approaches to the use of LULC history in locational modeling
2............................................................................................................A Quantitative Approach
2.1 ......................................................................................... Properties of the LULC maps
2.2 .........................................................................Properties of the Shropshire SMR data
2.3 .............................................................................................................Univariate analysis
2.4 .......................................................................................................... Multivariate analysis
3................................................................... Modelling Ancient LULC: a Historical Approach
3.1 ................................................................................................. Stability in the longue durée
3.2 ......................................................................................................Place-name etymology
3.3 .......................................................................................................Documentary sources
4...................................................................................................................................Conclusions
1
1
1
2
2
2
3
5
8
13
14
14
16
17
15 .................Settlement hierarchies, Territorial divisions, and Visual dominance
1 ..........................................................................................................................Introduction
2 .......................................................................... Settlement and Territory in Protohistory
3 ................................................................................Roman Colonies of the Lepine Scarp
4 ...........................................................................................................................Conclusions
1
2
9
13
16 .........................................................WHP Case Studies in Visibility and Fricion
1...................................................................................................................Visibility and Control
1.1 ................................................................................................................. Implementation
1.2 ..........................................................................................................................Discussion
2.................................................................................................. Structuration of the Landscape
2.1 ............................................................................................... Catchments and territories
2.2 .................................................................Modeling Iron Age/ Roman trade networks
3...................................................................................... Edge Effects and Background Indices
3.1 ....................................................................................................................... Edge effects
3.2 .................................................................................................... Viewshed radius effects
4...................................................................................................................................Conclusions
1
2
6
6
6
10
17
17
18
20
17 ........Interpreting Field Survey Results in the Light of Historic Relief Change:
......................................................the Fogliano beach ridges (south Lazio, Italy)
1...................................................................................................................................Background
2................................................................................................ Tracking Historic Relief Change
3...............................................Extraneous Sources of Differences Between the Two DEMs
4........................................................................................................... Interpreting the Evidence
5.................................................................................................................................... Conclusion
1
4
4
10
11
18 ..................................................................................Summary and Conclusions
1................................................................................................................. Aims and Approaches
2............................................................................................... Field work in Italy (1998 – 2000)
2.1 ......................................................................................................................... Field work
2.2 ................................................................................................................... Field methods
3............................................................................. The Methodology of Regional Comparison
3.1 ............................................................................... Modeling data formation processes
3.2 ......................................................................................... Modeling settlement patterns
4............................................................... Conclusion: Regional Archaeological Data Analysis
1
3
3
6
8
9
12
15
Nederlandse Samenvatting
What a useful thing a pocket-map is! I remarked.
That's another thing we've learned from your Nation, said Mein Herr,
map-making. But we've carried it much further than you. What do you
consider the largest map that would be really useful?
‘About six inches to the mile’.
Only six inches! exclaimed Mein Herr. We very soon got to six yards to
the mile. Then we tried a hundred yards to the mile. And then came the
grandest idea of all! We actually made a map of the country, on the
scale of a mile to the mile!
Have you used it much? I enquired.
It has never been spread out, yet, said Mein Herr: the farmers
objected: they said it would cover the whole country, and shut out the
sunlight! So we now use the country itself, as its own map, and I
assure you it does nearly as well.
- Lewis Caroll, Sylvie and Bruno Concluded
PREFACE
The volume now before you represents most of my research of the past seven years. It has grown
out of two successive research projects and the papers and articles I have written for them since
1994. From 1994 to 1997 I was a Leverhulme research fellow based at the University of
Birmingham Field Archaeology Unit (Birmingham, UK) working with Dr Roger White, Simon
Buteux, and Dr Vince Gaffney on the Wroxeter Hinterland Project, and from 1997 until the present
I have been part of the Regional Pathways to Complexity project at the Groningen Institute of
Archaeology, directed by Dr Gert-Jan Burgers and Prof Peter Attema. Parts of this thesis make use
of original and compiled data generated in the course of these two projects, and you will therefore
find a mixture of work carried out in Britain and Italy being discussed.
The Wroxeter Hinterland and Regional Pathways to Complexity projects are not only very similar in
the kinds of questions they confront, they also operate within a similar geographical scale (the
‘region’) and theoretical context (‘landscape archaeology’). They both intend to investigate spatial
patterns in the compiled regional archaeological data, and to explain these pattern – and deviations
thereof – in terms of underlying historical processes. The title of this thesis, Pattern to Process,
encapsulates this. The investigation does not start with a tabula rasa, however: we bring along our
baggage of pre-existing models and interpretations of the past, hoping either to confirm or refute.
In the manner in which we go about this task, the uneasy position of the field of Archaeology, split
between the Humanities and the Sciences since the New Archaeology of the 1960’s, becomes
apparent. Archaeological remains can be studied as a means to support and enrich the culturehistorical paradigm, or they can be studied as ‘archaeological landscapes’: on their own merits and
with an appropriate methodology. Hence my subtitle: methodological investigations into the
formation and interpretation of spatial patterns in archaeological landscapes.
Since much of my work has already been, or will be, published with co-authors as articles in journals
and conference proceedings, I decided to submit this thesis ‘in articles’(as the expression goes)
rather than re-use the material in a single-author monograph. I have tried to organise the material in
a logical fashion and have provided introductory and concluding chapters which I hope will help
you, reader, find your way. One health warning is in order: if you intend to read this volume from
cover to cover, you will inevitably encounter repetition and even contradiction among the chapters
that follow. My advice to you is therefore to regard this volume as a buffet rather than a formal
dinner.
ACKNOWLEDGEMENTS
This thesis, and the research on which it is based, could not have happened without the support and
encouragement of colleagues both here and in England. I would like to express my gratitude to my
colleagues at the Birmingham University Field Archaeology Unit, for their hospitality and for
helping me and my family cope with life in a strange country. In particular, Vince Gaffney for
inviting me to become part of the Wroxeter Hinterland Project team and then arranging for the
team to meet the Queen at Buckingham Palace; Roger White for sharing Wroxeter, Shrewsbury, and
central Shropshire with me; and Simon Buteux and Sally Exon for helping out under all
circumstances. At the Shropshire County Council, Penny Ward was most helpful with proving
access to the county records.
Whilst much of the Herculean work of the Wroxeter Hinterland project team has by now reached
print in technical papers, journal articles and volumes, and the occasional book, the final synthetic
volume is still in preparation. I have very been fortunate in that my employers at the Groningen
Institute of Archaeology have agreed to allow me to finish some of my own contributions to this
work. I would like to thank my colleagues in the RPC project for the help, discussions and the
shared fieldwork. In particular, I am grateful to Peter Attema who invited me to apply for one of the
research openings in his then brand new project, has been my mentor for the past four years in a
very agreeable hands-off style, and is now my promotor; to Marianne Kleibrink and Douwe Yntema
whose research at Satricum, Francavilla Marittima, and the Brindisino provided the ultimate raison
d’être for the RPC project; to Marianne again for convening a series of stimulating staff meetings
where current research issues were discussed; and to Gert-Jan Burgers for making me think again
about field methods, terminology, and the functional interpretation of ceramic surface scatters.
Peter, Gert-Jan and Marianne also commented on draft versions of this thesis, and I am grateful for
their helpful comments. Nick Ryan encouraged my interest in using ‘executive toys’to improve
fieldwork procedures, programmed and tested soft- and hardware during our fieldwork, and
together with me described this work in a paper and article. Hendrik Feiken, then an M.A. student
at the GIA, provided welcome help in preparing, executing and publishing methodological studies
related to our fieldwork on the Pontine coast.
Other archaeologists elsewhere have also helped me in various ways. Most directly, I should thank
Bert Voorrips, Susan Loving, and Hans Kamermans for allowing (and indeed encouraging) me to
use the ceramic data collected during the Agro Pontino Survey - Bert even supplied me with a digital
copy of the complete finds database. Kenneth Kvamme provided helpful leads and insightful
comments especially on the GIS studies presented in this thesis; besides him, I should also thank
the subscribers to the GISARCH discussion list for their comments and interest. A special mention
and thank you should go to Jan Hartmann, a kind teacher who gave me the opportunity to pursue
my interest in computer applications and methodology as an M.A. student at the then Institute for
Pre- and Protohistory ‘Albert Egges van Giffen’of the University of Amsterdam. He put me on the
trail which I have been following now for almost 15 years.
Finally, an apology rather than a thank you is due to my wife Monica, who has had to cope with my
absences and more than her fair share of the household, especially over the last year or so of work
on this thesis.
Groningen, 1st May 2002
SAMENVATTING
NEDERLANDSE
S A M E N VAT T I N G
Dit proefschrift, getiteld ‘Pattern to Process: methodological investigations into the formation and
interpretation of spatial patterns in archaeological landscapes’(‘Van Patroon to Proces: methodologisch
onderzoek naar de vorming en interpretatie van ruimtelijke patronen in archeologische landschappen’),
vormt de schriftelijke neerslag van onderzoek dat werd uitgevoerd in het kader van twee verschillende
meerjarige projecten in twee ver uiteen liggende delen van Europa. Van 1994 tot 1997 maakte ik deel uit
van het Wroxeter Hinterland Project (WHP) dat onder leiding van Dr Vincent Gaffney aan de
Birmingham University Field Archaeology Unit (BUFAU) werd uitgevoerd. Dit project had ten doel het
ontstaan, de bloei, en het verval van de Romeinse civitas-hoofdstad Viroconium (tegenwoordig Wroxeter,
Shropshire) te relateren aan haar inheems-Romeinse achterland, in de late IJzertijd woongebied van de
stam der Cornovii. Vervolgens voerde ik van eind 1997 tot begin 2001 aan de Universiteit van
Groningen mijn promotieonderzoek uit binnen het project Regional Pathways to Complexity (RPC).
Binnen dit project werkte ik onder leiding van Dr Peter Attema (RUG) en Dr Gert-Jan Burgers (VU)
samen met drie andere promovendi aan de vergelijking van nederzettingsdynamieken en landgebruik van
de late protohistorie tot en met de Romeinse tijd in drie Italische regio’s – het Pontijnse gebied in zuidLazio, de Salento Isthmus in Apulië, en de Sibaritide aan de Ionische Golf.
In beide projecten concentreert de problematiek zich rond de verhouding tussen interne dynamieken van
de inheemse samenlevingen en de rol van externe kolonisatoren. Waar tot voor kort algemeen werd
gezien als de drijvende kracht achter maatschappelijke ontwikkelingen als centralisatie en urbanisatie, is
sinds de jaren ’80 een kentering op gang gekomen waarbij het inheemse perspectief een gelijkwaardig, zo
niet primair, belang krijgt toegedeeld. Omdat dit perspectief niet of nauwelijks door historische bronnen
kan worden ondersteund, moeten archeologen gebruik maken van andere instrumenten zoals
etnografische vergelijkingen; daarnaast moet het gebrek aan kennis van het inheemse niet-urbane
landschap gecompenseerd worden door nieuw veldonderzoek en de studie van inheemse patronen van
landgebruik en bewoning. Naast de problematiek van centralisatie, urbanisatie en kolonisatie vormde
daarom ook de geografische inslag van het onderzoek een constante in beide projecten, waarbij vooral
(het onderzoek naar) de toepassing van geografische informatiesystemen (GIS) voorop stond.
In de vroege jaren ’80, juist toen archeologen alom teleurgesteld raakten in de hoge verwachtingen van de
New Archaeology, kwam voor het eerst betaalbare software beschikbaar waarmee geografische
informatie – kaarten – aan alfanumerieke gegevensbestanden gekoppeld kon worden: geografische
informatiesystemen (GIS). Het nut van dergelijke software werd eerst vooral in de Verenigde Staten
onder beheerders van archeologische terreinen herkend, maar begon daarna al gauw ook in Europa en bij
universitaire onderzoekers interesse te wekken vanwege haar analytische potentieel. Deze krachtige
software bood de mogelijkheid de in de archeologie gebruikelijke, op combinaties van vele geografische
factoren gebaseerde, nederzettingsmodellen te formaliseren en op betrekkelijk eenvoudige wijze te
genereren en te visualiseren.
Onder andere vanwege het bovenstaande, maar ook omdat veel delen van dit onderzoek in verschillende
vorm al eerder gepubliceerd zijn of nog zullen worden, heeft dit proefschrift de, onder Nederlandse
archeologen ongebruikelijke, vorm ‘in artikelen’gekregen. Deze zijn weliswaar gegroepeerd in vier delen
– inleiding, methodologische studies, veldwerk, en GIS-toepassingen – maar dit blijft een achteraf
opgelegde structuur, en de individuele hoofdstukken kunnen dus het beste als losstaande artikelen
1
V A N L EUSE N : P A T T E R N T O P ROCESS
gelezen worden. Om de samenhang tussen de delen en, daarbinnen, tussen de artikelen, te vergroten heb
ik aan het geheel een introducerend en een samenvattend hoofdstuk toegevoegd, terwijl aan het
veldwerk-deel nog een aparte inleiding vooraf gaat.
DEEL I: INLEIDING
Dit deel vangt (hoofdstuk 1) aan met een presentatie van het kernprobleem van dit onderzoek – dat
patronen in archeologische gegevens op elke schaal veroorzaakt kunnen zijn door externe vertekenende
factoren. Wanneer het, zoals in de landschapsarcheologie, de bedoeling is om archeologische resten en
landschap in samenhang te bestuderen moeten dus manieren gevonden worden om dit probleem
beheersbaar te maken, en het GIS-instrumentarium speelt hierin een belangrijke rol. Aangezien deze
problematiek wordt onderzocht in het kader van de WHP- en RPC projecten worden ook deze kort
ingeleid. Tenslotte wordt een overzicht gegeven van de structuur en inhoud van het proefschrift.
Hoofdstuk 2 werkt het kernprobleem nader uit aan de hand van voorbeelden uit de drie Italiaanse regio’s
van het RPC project. Er wordt begonnen met het geven van een overzicht van de bestaande
nederzettingsdynamieken zoals dat uit recente literatuur naar voren komt. Hieruit worden de
belangrijkste maatschappelijke processen van het 1e millennium v. Chr. naar voren gehaald, en de
belangrijkste archeologische concepten, theorieën en methoden besproken die in deel IV terug zullen
keren. Centralisatie, urbanisatie, en kolonisatie worden als belangrijkste, maar tevens problematische
concepten geïntroduceerd; naar aanleiding van een analyse van de theoretische basis voor het maken van
interregionale vergelijkingen worden de voor- en nadelen van verschillende benaderingswijzen besproken,
waarbij gekozen wordt voor een dichter bij de gegevens staande kwantitatieve benadering dan de tot nu
toe gebruikelijke socio-politieke verklaringsmodellen; en tenslotte volgt een eerste voorlopige verkenning
van kwalitatieve en kwantitatieve vergelijking tussen de drie gebieden.
Aan dit inleidende deel is nog een hoofdstuk (3) toegevoegd waarin de doelstellingen en problematiek
van Romanisatie in het Wroxeter Hinterland in meer detail geïntroduceerd worden. Dit hoofdstuk bestaat
uit twee artikelen die in 1996-97 in congresbundels gepubliceerd werden. Het eerste, met coauteur Vince
Gaffney, plaatst het vinden van een verklaring voor het bestaan van Wroxeter zelf – de vierde urbane
nederzetting van Romeins Brittannië in grootte maar ogenschijnlijk zonder het ontwikkelde rurale
achterland dat bij zo’n stad hoort – centraal. Doelstelling is om Wroxeter’s plaats in de vigerende
modellen voor urbanisatie en Romanisatie in provinciaal-Romeinse context te bepalen door middel van
zowel een studie van de beschikbare archeologische gegevens als een uitgebreid programma van
veldverkenning. Ook wordt het beoogde instrumentarium voor geografische analyse beschreven. Het
tweede artikel, met coauteur Roger White, verwerpt op grond van direct bewijs en theoretische
argumenten een drietal bestaande verklaringen voor het bestaan van een welvarende inheems-Romeinse
stad temidden van een nauwelijks geromaniseerd achterland – dat Wroxeter als stad te ambitieus gepland
was en feitelijk nooit een grote bevolking heeft gehad, dat de lokale bevolking vijandig zou staan
tegenover de kolonisator, en dat Wroxeter altijd economisch onderontwikkeld is gebleven. Waar de eerste
van deze verklaringen door middel van een vlakdekkende geofysische en luchtfotografische studie van de
stad direct kon worden weerlegd, worden in dit artikel ook de andere twee verklaringen verworpen op
grond van het argument dat het succes van de stad impliceert dat ook het achterland rijk moet zijn geweest
en dat deze rijkdom naar de stad moet zijn gevloeid. Als alternatieve verklaring voor de schijnbare
tegenstelling tussen stad en platteland stellen wij daarom voor dat de rijkdom van de inheemse Cornovii
archeologisch onzichtbare vormen (vee, zout) aannam, en dat bovendien onze kennis van nederzetting en
landgebruik in de late IJzertijd en Romeinse tijd binnen het gebied vertekend is door een gebrek aan
systematisch onderzoek.
DEEL II: METHODOLOGISCHE STUDIES
Zoals hierboven al werd aangeduid, heeft de interpretatie van grootschalige patronen in archeologische
landschappen voorheen altijd plaatsgevonden binnen de kaders die door de voorhanden historische
bronnen uit de klassieke oudheid werden geschapen. Wie zich aan die kaders wil onttrekken door zijn
interpretaties direct op patronen in de beschikbare archeologische gegevens te baseren, moet daarvoor
2
SAMENVATTING
eerst geschikte methoden ontwikkelen. Twee soorten methodologische studies werden ondernomen; ten
eerste studies die ten doel hadden om greep te krijgen op de kwaliteit van de archeologische gegevens die
aan de basis liggen van regionale nederzettings-geschiedenissen en daarmee ook van vergelijkingen tussen
regio’s (hoofdstukken 4 en 7). Ten tweede studies die ten doel hadden de bruikbaarheid van het GISinstrumentarium voor de analyse en interpretatie van patronen in die archeologische gegevens te
evalueren (hoofdstukken 5 en 6).
Interpretatie van regionale archeologische bestanden zoals die samengesteld worden door middel van
literatuurstudie, eventueel aangevuld met veldwerk, staat of valt met de kwaliteit van de aldus verzamelde
gegevens. Met uitzondering van enkele projecten die, onder invloed van de New Archaeology, tegen het
eind van de jaren ’70 werden ontworpen en in de jaren ’80 uitgevoerd, waren die gegevens nooit
verzameld met als doel een representatief beeld van het regionale archeologische landschap te verkrijgen.
Regionale interpretaties moeten dus expliciet met de mogelijkheid rekening houden dat de gebruikte
gegevens niet representatief zijn voor dat landschap. Eenzelfde probleem gold ook de interpretatie van de
in diezelfde periode steeds populairder techniek van de archeologische veldverkenning, waarbij steeds
intensiever en beter gecontroleerd werk de belangrijke rol van een aantal vertekenende factoren bij het
verzamelen van veldgegevens naar voren bracht. In hoofdstuk 4 worden methoden uitgewerkt om met
deze problematiek om te gaan, zowel pro-actief door procedurele verbeteringen in de planning en
uitvoering van lopend veldwerk, als retroactief door uitgebreide ‘bronnenkritiek’te plegen op in het
verleden gevormde gegevensbestanden.
Hoofdstuk 7 is gewijd aan een beschrijving van de experimenten die in samenwerking met Dr Nick Ryan
van de Universiteit van Kent te Canterbury zijn uitgevoerd tijdens veldwerk in de Sibaritide in 2000. Het
doel van deze experimenten is om het opnemen en verwerken van informatie tijdens en na archeologisch
veldwerk accurater en efficiënter te laten verlopen, door gebruik te maken van programmeerbare,
lichtgewicht, en halfautomatische digitale registratieapparatuur. Door de veldadministratie van
landschappelijke parameters en de verzamelde archeologische materialen direct in het veld digitaal uit te
voeren (dwz zonder ‘papieren’tussenstap) wordt de efficiëntie vergroot en de kans op fouten verkleind;
door tegelijk deze administratie automatisch te voorzien van nauwkeurige digitale plaatsbepalingen door
middel van GPS is de kartering van veldeenheden en archeologica niet langer afhankelijk van de minder
nauwkeurige plaatsbepaling met behulp van vaak verouderd kaartmateriaal. Met de apparatuur werden
onder andere routes, veldgrenzen, en de kern en omtrek van archeologische sites vastgelegd. Uit deze
experimenten blijkt dat dit soort apparatuur goed toegepast kan worden in zowel intensieve als extensieve
archeologische surveys, zij het dat het gebruiksgemak en de fysieke betrouwbaarheid van de systemen
nog verbeterd moet worden.
Om te komen tot een evaluatie van de twee hoofdstromingen in de GIS-literatuur van het afgelopen
decennium wordt in de hoofdstukken 5 en 6 een uitgebreide analyse gepresenteerd van de theorie,
methodologie en methoden die ten grondslag liggen aan zgn. ‘voorspellingsmodellen’(voorspellende
locatiemodellen, voornamelijk gebaseerd op eigenschappen van het fysieke landschap) en zgn.
‘cognitieve’modellen (voornamelijk modellen die betrekking hebben op de mate van zichtbaarheid en
bereikbaarheid van delen van het landschap).
Voorspellingsmodellen zijn internationaal voornamelijk ontwikkeld in de context van archeologisch
beheer en behoud, maar vormen als locatiemodellen ook sinds lang onderwerp van academisch
onderzoek. In dat laatste geval is het doel meestal om bestaande patronen van nederzetting en
landgebruik te verklaren door ze te relateren aan aspecten van de natuurlijke omgeving, en hiervoor werd
het potentieel van GIS al vroeg herkend, hetgeen in Europa geleid heeft tot een gestaag groeiende stroom
van publicaties sinds begin jaren ’90. In diezelfde periode echter krijgt ook de postmoderne theorie steeds
meer voorstanders binnen de Europese archeologie, zodat de ‘ecologisch deterministische’
voorspellingsmodellen onder hevige kritiek kwamen te staan, en er ‘cognitieve’alternatieven worden
voorgesteld. Dit heeft geleid tot een levendig maar chaotisch debat over zowel de theoretische
grondslagen als de doelstellingen en methoden van dit soort geografische voorspellingen. Hoofdstuk 5
geeft aan de hand van internationale literatuur een overzicht en evaluatie van alle in dit debat gebruikte
argumenten, die overheerst lijken te worden door een reeks dichotomieën voortkomend uit
3
V A N L EUSE N : P A T T E R N T O P ROCESS
gepolariseerde theoretische posities. Vervolgens wordt een argument opgebouwd dat niet theoretische
zuiverheid maar procedurele transparantie het belangrijkste kenmerk van voorspellingsmodellen dient te
zijn; een transparantie die alleen bereikt kan worden door alle in het hoofdstuk gepresenteerde stappen in
het modelleringsproces te formaliseren, de kwaliteit van de toegepaste gegevens en methoden te
verhogen, en de resulterende modellen ook daadwerkelijk te toetsen. Hiertoe worden specifieke
voorstellen gedaan.
Een belangrijke bijdrage van de postmoderne archeologie aan het debat over aard en doel van GIStoepassingen, hierboven reeds genoemd, is de aandacht voor het landschap zoals dat door de mens werd,
en wordt, gepercipieerd en geconcipieerd. In tegenstelling tot de externe, fysieke, kenmerken van het
landschap gaat het in deze benadering om de interne, cognitieve kenmerken. In de praktijk heeft zich
deze benadering vooral vertaald in archeologische toepassingen die gebruik maken van een tweetal GISinstrumenten in het bijzonder – de analyse van respectievelijk zichtlijnen en kostenoppervlakken. In
hoofdstuk 6 worden vrijwel alle internationaal gepubliceerde toepassingen van deze beide technieken, die
overigens nog volop in ontwikkeling zijn, beschreven en geanalyseerd.
DEEL III: VELDWERK
Onze kennis van de archeologie in alle drie de gebieden was, bij de aanvang van het RPC project,
voornamelijk opgebouwd uit Italiaans onderzoek sinds de jaren ’60 en Nederlandse projecten vanaf circa
1980. Lacunes in deze kennis vallen dan ook vrij direct te herleiden tot (in de Italiaanse topografische
surveys en een deel van het Nederlandse onderzoek) een overwegende belangstelling voor de klassieke
cultuur ten nadele van eerdere en latere perioden, gekoppeld aan de veronachtzaming van ‘lage’cultuur
en het rurale landschap ten faveure van heiligdommen en urbane grafvelden en nederzettingen. Waar het
de meer intensive en systematische Nederlandse surveys betreft was er sprake van een onevenwichtige
ruimtelijke spreiding over de onderzoeksgebieden ten nadele van "marginale" gebieden. In het RPC
project is in de jaren 1998-2000 op bescheiden wijze een bijdrage geleverd aan het opvullen van deze
lacunes in de bestaande gebiedskennis door middel van intensieve en systematische archeologische
surveys in alle drie de onderzoeksgebieden. Daarbij is echter steeds ook veel aandacht besteed aan het
ontwikkelen van een geschikte methodologie voor het uitvoeren van het veldwerk zelf en voor de analyse
van de resulterende gegevens. In dit deel zijn de vier met coauteurs uit het RPC project gepubliceerde
verslagen van dit veldwerk opgenomen, voorafgegaan door een inleiding die de overkoepelende
doelstellen en resultaten van dit veldwerk samenvat.
Van de RPC surveys nabij Ninfa aan de voet van de Monti Lepini, en rond het Lago di Fogliano aan de
Pontijnse kust (1998-1999), wordt verslag gedaan in de hoofdstukken 9 en 10 van dit proefschrift.
Hoewel het oorspronkelijke doel van het veldwerk bij Ninfa de kartering van zogeheten platform-villa’s
was, viel het onderzoeksgebied nog net binnen één van de kaartbladen van de Forma Italiae (Cora, door
Paola Vittucci Brandizzi,1968) en kon dus ook getoetst worden in hoeverre dit oudere Italiaanse
gegevensbestand representatief was voor het totale archeologische landschap. Aangetoond werd dat in
het gebied naast de door Vittucci gekarteerde, en vrijwel exclusief uit de Romeinse periode daterende,
monumentale resten ook veel kleinere en minder opvallende Romeinse sites lagen; bovendien bleek er
een door haar in het geheel niet geregistreerd maar intensief gebruikt pre-Romeins (Archaïsch en postArchaïsch) landschap aanwezig te zijn. De bewoningsgeschiedenis van deze landschapseenheid (het
‘noordelijk colluvium’, inclusief de proto-urbane nederzetting Caracupa/Valvisciolo) lijkt hiermee meer
overeen te komen met die van het kerngebied van de Latiale samenleving in de Albaanse heuvels, dan
met dat van de ‘marginale’Pontijnse vlakte waarvoor pas vanaf de mid-Republikeinse periode intensieve
bewoning en landgebruik wordt verondersteld. Voor de Romeinse periode beginnen nu, met het
herkennen van verschillen in grootte, ligging, en status van de vindplaatsen, in het Ninfa gebied
bovendien de contouren van een meer gedetailleerde nederzettingshierarchie op te doemen. Op
methodologisch gebied gaf de survey vooral inzicht in de vertekeningen inherent aan de oude
topografische manier van onderzoek, en in de noodzaak om in latere surveys methodes te ontwikkelen
voor de registratie van continue verspreidingen van aardewerk over het landschap (in tegenstelling tot
discrete verspreidingen, in de vorm van ‘sites’).
4
SAMENVATTING
Dit laatste zou inderdaad in latere surveys een belangrijk punt van aandacht worden, waarmee voor het
eerst werd geëxperimenteerd tijdens de eerste Fogliano-survey (1998). De uitvoering en registratie van
het veldwerk werd gericht op geografische eenheden ('blokken' van circa 1 hectare) in plaats van op
archeologische (sites) of landbouwkundige (percelen), en het probleem van de selectieve waarneming
werd bestreden door alle oppervlaktevondsten per 'blok' te verzamelen en pas achteraf door een expert te
laten beschrijven. Het Fogliano-veldwerkgebied werd gekozen omdat het representatief werd geacht voor
het op grond van zowel klassieke als meer recente historische bronnen als ‘marginaal’beschouwde
kustlandschap van de Pontijnse regio. De resultaten van de survey, gepresenteerd in hoofdstuk 10, wijzen
er op dat dit ook inderdaad tot aan de laat-Republikeinse periode het geval is geweest. Resten van
waarschijnlijk extensief landgebruik in dit landschap van fossiele strandwallen, valleien en kustlagunes
waren incidenteel aanwezig voor de gehele ceramische periode tot en met de IJzertijd, waarna het aantal
vindplaatsen begint te groeien in de Archaïsche periode. Voor de post-Archaïsche en mid-Republikeinse
periode blijft de intensiteit van het gebruik van dit landschap, vanwege het ontbreken van diagnostische
aardewerktypen, vooralsnog onduidelijk, maar er lijkt een duidelijk contrast aanwezig te zijn met de
opvallende groei van het aantal rurale villas in de laat-Republikeinse periode (200 – 0 v. Chr.). Aangezien
deze groei voornamelijk plaatsvindt in het centrale deel van het gebied, waar zich ter plaatse van het
moderne dorpje Borgo Grappa een groter, relatief vlak en aaneengesloten gebied met zandige bodems
bevond, kunnen we voor deze periode spreken van de groei van een ruraal dorp – een sociaaleconomische ontwikkeling die misschien in verband moet worden gebracht met de productie en
bovenregionale handel in vis en visproducten die in deze periode langs de Pontijnse kust op gang kwam.
Deze rurale bloei duurde niet langer dan 2 eeuwen, want geen van de vindplaatsen lijkt na de vroege
Keizertijd nog in gebruik te zijn geweest; deze neergang wordt in verband gebracht met de algemene
trend tot schaalvergroting en extensieve exploitatie die in deze periode zijn intrede doet in het
expanderende Romeinse rijk.
Naast het gegeven dat de geschiedenis van nederzetting en landgebruik in beide gebieden zich beter laat
begrijpen wanneer we haar analyseren in termen van relatief kleine fysieke landschapseenheden dan
wanneer we proberen eenzelfde ontwikkeling te schetsen voor de Pontijnse regio als geheel, bracht de
Fogliano survey ook een problematiek aan het licht die eerder al bij de analyse van surveys in het gebied
van de Egeïsche zee was herkend door Britse archeologen, namelijk dat van de interpretatie van de vaak
zeer lage ‘off-site’vondstdichtheden uit verschillende perioden die nagenoeg het gehele Mediterrane
landschap kenmerken. In latere surveys is door middel van het herbezoeken van vindplaatsen getracht
meer greep te krijgen op de factoren die de kans op het doen van oppervlaktevondsten beïnvloeden, en
op grond hiervan meen ik met anderen dat ook een enkele vondst onder bepaalde omstandigheden
opgevat moet worden als aanwijzing voor het bestaan van een lokaal ondergronds reservoir (vindplaats).
Factoren die het overleven en de zichtbaarheid van aardewerk in de ploegvoor beïnvloeden wisselen
bovendien sterk in belang met de productiewijze en ouderdom van het materiaal, reden waarom de kans
op het terugvinden van bijvoorbeeld protohistorisch aardewerk veel lager gesteld moet worden dan van
het klassieke Romeinse en, zoals we hieronder zullen zien, Hellenistische aardewerk.
Het veldwerk in 1999 uitgevoerd nabij het stadje Ostuni in de Salentijnse Murge had, net als dat van
Fogliano, ten doel een voordien niet of nauwelijks onderzocht ‘marginaal’deel van het archeologische
landschap intensief te karteren. Zoals in hoofdstuk 11 wordt uitgelegd, vormde het kalksteenplateau van
de Murge van oudsher de landschappelijke marge van de urbane samenleving die zich vanaf de vroegHellenistische periode in de Salentijnse Isthmus ontwikkelde. Waar het onderzoek zich vanuit de
Universiteit van Lecce en de Vrije Universiteit van Amsterdam had geconcentreerd op de centrale
plaatsen en hun achterland, bood de Ostuni-survey voor het eerst de kans om de gebruiksgeschiedenis
van een deel van zowel de hoge Murge zelf als de overgangszone naar de Adriatische kustvlakte in detail
te karteren. Op methodologisch vlak werd de survey, naar aanleiding van de ervaringen met eerder
veldwerk, uitgevoerd met een hogere geografische resolutie (eenheden van 0.25 hectare) en een
consequenter doorgevoerde registratie van zichtbaarheidsfactoren. Voor de Hellenistische en Romeinse
perioden bevestigde de survey enerzijds het marginale karakter van landgebruik, waarbij individuele
boerderijen op ongeveer een km afstand van elkaar lagen, anderzijds was het verrassend dat vroeg
Hellenistisch aardewerk, bouwmaterialen en –stijlen reeds zo diep in de Murge waren doorgedrongen. De
groei van een inheems-Hellenistische urbane samenleving op de Isthmus ging dus gepaard met een
5
V A N L EUSE N : P A T T E R N T O P ROCESS
gelijktijdige Hellenisatie, mogelijk zelfs kolonisatie, van zelfs de meest afgelegen gebieden, een aanwijzing
dat de totale bevolking bij dit proces betrokken was.
Met betrekking tot de landschapsgeschiedenis van de protohistorie, de klassieke en de archaïsche
perioden in de Salento bevestigde het Ostuni-veldwerk door de nagenoeg volkomen afwezigheid van
vondsten uit de late Bronstijd tot en met de 4e eeuw dat de samenleving gedurende die periode
waarschijnlijk sterk gecentraliseerd leefde in strategisch (kliffen, heuveltoppen) gelegen nederzettingen.
De veranderlijkheid van dergelijke protohistorische strategieën werd onverwachts aangetoond door het
karteren van zeer algemene voorkomende vondstspreidingen van zeer homogeen impasto-aardewerk uit
de midden-Bronstijd. Deze worden geïnterpreteerd als de resten van een slechts enkele eeuwen in gebruik
geweest zijnd systeem van mobiel landgebruik, ‘shifting cultivation’, waarbij de bevolking in
familieverband gedurende korte perioden steeds weer nieuwe of geregenereerde delen van het landschap
bewoonde en ontgon, en er dus geen sprake was van permanent bewoonde grotere nederzettingen.
Voor het veldwerk in de Sibaritide (2000) werd ervoor gekozen de survey te richten op het toetsen van de
bestaande grootschalige kartering door Lorenzo Quilici uitgevoerd in de jaren ’60, en tegelijkertijd een
deel van het achterland van de protohistorische nederzetting en cultusplaats op de Timpone della Motta
nabij Francavilla Marittima in meer detail te leren kennen. Ook hier was weer een belangrijke
methodologische component in het veldwerk, waarover in hoofdstuk 7 uitgebreid verslag gedaan wordt.
Uit de kartering door Quilici was, net als andere surveys uitgevoerd in de stijl van de Forma Italiae, een
intensief bewoond maar vrijwel exclusief klassiek (Hellenistisch-Romeins) landschap naar voren
gekomen, waarin bovendien een clustering in (aan hypothetische doorgaande routes gekoppelde) ‘dorpen’
viel waar te nemen. Ons veldwerk was erop gericht de correctheid van dit patroon te toetsen door middel
van een veldverkenning van een representatief transect door de voetheuvelzone, waaruit dan eventuele
chronologische en ruimtelijke vertekeningen naar voren zouden moeten komen. In dit geval bevestigde
het veldwerk het merendeel van Quilici’s resultaten: een slechts zeer sporadische aanwezigheid van
protohistorisch materiaal wijst erop dat deze zone voorafgaand aan de archaïsche expansie van de
Griekse kolonie Sybaris niet in intensief gebruik was, de vele Hellenistisch/Romeinse vindplaatsen
bleken merendeels in de vroeg Hellenistische periode te zijn ontstaan, en te clusteren rond plateauranden. Anderzijds werd dit kaartbeeld wel genuanceerd door de ontdekking dat de Hellenistische
vindplaatsen ook in andere landschappelijke settings wel voorkwamen, en dat de meeste niet aantoonbaar
in de Romeinse periode gecontinueerd of uitgebreid werden.
DEEL IV: GIS-TOEPASSINGEN
In het vierde deel van dit proefschrift wordt een vijftal toepassingen van GIS gepresenteerd die niet
alleen dienen ter beantwoording van specifieke onderzoeksvragen, maar ook ter illustratie van de
methodologische studies in deel II. In hoofdstuk 13, bijvoorbeeld, wordt de vorming, vergelijking en
interpretatie van gegevensbestanden op regionale schaal (hoofdstuk 2) toegepast op de Pontijnse regio.
Daarbij wordt onder andere ingegaan op de gewenste structuur van een regionale relationele
archeologische database, op de noodzaak van een eenduidige definitie van archeologische entiteiten
(zoals de ‘permanent habitation site’), en op het opvallende gebrek aan standaardisatie van
veldwerkmethoden en –publicatie waardoor zelfs vergelijkingen op het meest eenvoudige plan tussen
archeologische bestanden mank gaan. In hoofdstuk 14 wordt de door recent en subrecent landgebruik
veroorzaakte vertekening in regionale archeologische bestanden (besproken in hoofdstuk 4)
gedemonstreerd aan de hand van voorbeelden uit het Wroxeter Hinterland (ingeleid in hoofdstuk 3); en
in hoofdstuk 17 wordt ditzelfde thema uitgewerkt in een studie naar de veranderingen in landvorm die
vanaf de late jaren ’20 van de vorige eeuw door de fascistische en latere landverbeteringen in de Pontijnse
vlakte hebben plaatsgevonden, en de invloed daarvan op de resultaten van de veldverkenningen bij
Fogliano (waarvan in hoofdstuk 10 verslag gedaan wordt). Vooral uit deze laatste twee hoofdstukken
blijkt duidelijk hoe sterk de meeste kleinschalige patronen in regionale archeologische bestanden
gecorreleerd zijn aan de combinatie van recent en subrecent landgebruik, en lokale onderzoeksmethoden
en –interessen.
6
SAMENVATTING
In hoofdstuk 15 worden ruimtelijke modellen voortkomend uit de problematiek van centralisatie,
urbanisatie en kolonisatie (besproken in hoofdstuk 2) aan de hand van voorbeelden uit de Pontijnse regio
en de Sibaritide gepresenteerd – centralisatie en de vorming van territoria in de late Bronstijd,
urbaniserende inheemse ‘peer polities’ in de IJzertijd/Archaische periode, en vroege en midRepublikeinse kolonisatie van de Lepijnse marge. In hoofdstuk 16 wordt aan de hand van voorbeelden
uit het Wroxeter Hinterland project de toepassing van zichtlijnen- en kostenoppervlakten analyse
(hoofdstuk 6) uitgewerkt. Zichtlijnen worden benut om de potentiële mate van controle over de centrale
Severn-vallei vanuit IJzertijd-hillforts en, later, het Romeinse legerkamp te Wroxeter te modelleren en te
visualiseren; kostenoppervlakten worden gebruikt om eerst de begaanbaarheid van het gebied ten tijde
van de Romeinse invasie te modelleren, en op grond daarvan de ligging van potentiële knooppunten in
de lokale infrastructuur.
Resultaten, argumenten, en conclusies uit al het voorgaande worden tenslotte in hoofdstuk 18 nog eens
samengevat, waarbij de onderlinge relaties tussen de in het inleidende hoofdstuk uitgezette
onderzoekslijnen nog eens benadrukt worden.
7
INTRODUCTION
C HAPTER 1
INTRODUCTION
1
AIMS AND BACKGROUN D
“The greatest challenge of inter-disciplinary landscape archaeology in the Mediterranean in the
coming years will be how to bridge the divide between the ecological approaches of the natural
sciences to past landscapes, on the one hand, and the concerns of social archaeologists on the
other with the interface between human actions and landscape.”
- Graeme Barker and David Mattingly, in their introduction to the POPULUS
series of conference proceedings on the archaeology of Mediterranean
landscapes (1999/2000: vii)
AIMS
In the archaeology of Italy from the Bronze Age to the Roman period, the study of the internal
development of indigenous Italic societies and landscapes has remained a relatively underdeveloped area
due to the emphasis on explanations relying on external factors (the influence of non-Italic cultures),
dominant historical processes (the Greek and Roman colonizations), and a simplistic culture-historical
view of society (stages of growth, flowering and decline). Much attention has been lavished on the
influence on regional Italic cultures of foreign artefacts and manufacturing techniques during the
‘Mycenean’and ‘international’periods, when contacts of trade and exchange ranged throughout the
Mediterranean. Similarly, interest in Greek and Roman colonization, mainly based on historical sources,
has dominated the study of the role of native culture to the organization of regional Italic societies and
landscapes. This one-sided approach has led to the view that the early urbanization of central and
southern Italy has been a relatively homogeneous process, in which the role of international impulses and
colonization movements has been paramount. Accordingly, the core aim of the Regional Pathways to
Complexity (RPC) project has been to demonstrate both the much more complex nature of
archaeological reality, and the decisive role played by the perspective offered by regional archaeological
landscape study, by comparing the development of indigenous societies in central and southern Italy
through the 1st millennium BC and into their incorporation into the Roman state, with the emphasis on
the processes of centralization, urbanization, and colonization.
THE REGIONAL PATHWAYS TO COMPLEXITY (RPC) PROJECT
The RPC project started in the summer of 1997, was carried out by staff at the archaeological institutes of
the University of Groningen and the Free University of Amsterdam, and ran for an initial period of four
years until 2001. Accounts of its aims and context are provided elsewhere (most recently: Attema et al.
1998, Attema et al. (eds) forthcoming), and will not be repeated in detail here. The project as originally
proposed for funding under the Netherlands Organization for Scientific Research (NWO) research
program 'Settlement and Landscape in Archaeology' (Attema 1996) defines the aim of the project as "the
analysis of the process of urbanization that took place in large parts of the Mediterranean world in the 1st
millennium BC, through a study of long term developments in settlement behavior, land use and
technology".
1-
1
V A N L EUSEN : P A T T E R N T O P ROCESS
In the project proposal Attema (1996:13-15) argued that, in the protohistoric and classical archaeology of
Italy, the internal dynamics of the indigenous Italic societies and landscapes have remained
underemphasized in favor of the influence of dominant external cultures and historical processes and the
culture-historical paradigm of the growth and decline of great powers. Attention has been lavished on the
presence of foreign artefacts and technologies in Italy in the Mycenean and 'international' periods, when
networks of trade and exchange extended across the Mediterranean. Similarly, archaeological study of
indigenous Italian cultures has been overshadowed by a focus on Greek and Roman colonization based in
historical sources. This lopsided interest has led to the early urbanization of central and southern Italy
being cast as a relatively homogeneous process, in which external impulses and colonization movements
were the prime moving factors. In contrast, regional archaeological research conducted from the 1980s by
the institutes participating in the RPC project has shown that reality is much more complex; and a project
that focuses on the variability of the Italian landscape and the persistence of indigenous regional
traditions in land use, technology, and settlement behavior, has the potential to contribute significantly to
a well-founded interpretation of early Italian urbanization processes.
Trajectories towards social complexity in protohistoric Italian regions are most clearly expressed in the
Italian landscape through the centralization of settlements and through forms of early urbanization.
Excavations and surveys have shown that in this process Greek and Roman colonization were potent
forces of change. But social complexity was not solely brought about by colonists - some regions already
had complex societies at the time that they were colonized, others less so. In addition, the nature and
intensity of indigenous contacts with the colonial presence differed, with some regions less directly
involved than others. The RPC project studies and compares three areas that show such diverging
trajectories. These are the Pontine Region in Central Italy, the Salento Isthmus in Puglia and the Sibaritide
in Calabria, all of which have a tradition in Dutch archaeological fieldwork. In these areas the research
team of the RPC project compares the modes of interaction between the Italic peoples and Greek and
Roman colonialism.
The central question of the project is "how did colonialism affect the Italian regional pathways to social
complexity?" Comparative diachronic research of landscape and settlement dynamics is the way in which
the project attacks this question over a time span of a millennium from the late Bronze age to the early
Roman Empire. A long term perspective (1400 BC - AD 400) is adopted in order to observe this process
in its full duration and regional variability. The chosen method of long term comparative study aims to
identify connections between early urbanization in the three regions, as well as to gain a fuller
understanding of the elements playing a role in this process - landscape, settlement systems, land use,
technology, and tradition. Four Ph.D. students are being employed by the project to study each of these
elements; the thesis that now lies before you has addressed mainly the element of settlement systems.
THE WROXETER HINTERLAND PROJECT (WHP)
The Wroxeter Hinterland project, funded by the Leverhulme Trust and running from 1994 to 1997 at the
University of Birmingham Field Archaeology Unit (BUFAU) under the direction of Dr. Vince Gaffney. It
is introduced in more detail in chapter 3 of this thesis. Although it is concerned with a much later period,
and an area far removed from Italy, the Wroxeter Hinterland project’s study of the transformation of the
late pre-Roman British tribal society of the Cornovii into a Romano-British Civitas centering on
Viroconium Cornoviorum (Wroxeter) focuses on the very same processes studied in the RPC project as
well. The juxtaposition of examples and studies from both projects therefore provides a welcome broad
canvas against which to place these processes.
PATTERNS
The emphasis throughout this thesis on the recording and analysis of pattern is in the intellectual tradition
of the New Archaeology (Trigger 1989:310-312). I am disturbed by the current emphasis in both
archaeological theorizing and teaching, on the historicist, post-processual approach. The academic
1 -
2
INTRODUCTION
practice of archaeology can only be justified, in my view, if the basic rules of scientific research,
publication, and debate continue to be taught and adhered to. As the pendulum swings from the
scientific, rule-finding approach of the ‘60s and ‘70s to the humanistic, historicist approach of
postmodernism in 1980s and 1990s archaeology, the straw man of ‘environmental determinism’was set
up to decry and distance oneself from.
Regional or landscape studies, whether research- or CRM-oriented, with or without the use of GIS, are
generally concerned with discerning and interpreting patterns of archaeological land use and settlement.
And undoubtedly the archaeological record is patterned in various ways, but so are other factors
influencing our knowledge of that record – erosion and deposition, land use, and research bias, to name
the most important. How can we interpret the archaeological record if we cannot separate the effects of
these different types of pattern? And if human patterning is only one of a number of factors determining
the archaeological patterns found by us, what does that say about the models (predictive or otherwise)
that we produce? We can relate slope, distance to water, and other environmental variables to the
occurrence of certain site groups, but the correlation may run via the effects of erosion and the
differential visibility of archaeological materials on the surface, rather than directly.
Different types of patterning occur at different spatial and temporal scales. This thesis is about the
detection, description, and explanation of such patterns. Since we must first understand how the ways in
which we study the past affect our understanding of it, much of this thesis is devoted to methodology how do we collect and record data, how do we analyze its structure, and how do we attach explanations
to such structures? The goal of this work is to find ways of studying the past which can either avoid
recording biases, or which allow formal corrections for such biases to be made. Methodological themes
investigated in this thesis include: the design and execution of archaeological surveys (field walking);
methods of analysis of both regional ‘site’databases and local ‘non-site’survey data using Geographical
Information Systems; and methods of dealing with the tension between description, explanation, and
extrapolation in archaeological location / allocation studies.
LANDSCAPE ARCHAEOLOGY AND THE ROLE OF GIS
The case studies involving field work and GIS analysis are all based on the theoretical premises of
landscape archaeology – that human actions may occur, and leave an essentially continuous ‘blanket’of
traces, anywhere in the landscape, that the resulting surface record is a palimpsest of such traces through
time, and that patterns in this record may be explained in part by the in turn limiting and enabling
qualities of the landscape. These principles are extended into the realm of spatial extrapolation and
cultural resource management using the theory of spatial sampling (which says that properties of a
properly selected sample have a specific likelihood of also being properties of the parent population).
Archaeological input for GIS analysis comes in two forms. Firstly, traditional archaeological records often
collected and enhanced in the form of a ‘topographic’or desktop survey; and secondly, field walking
surveys. The former are site-oriented without exception, and are based on archive research and a limited
amount of field work; the latter come in several forms (urban, rural) and intensities but share the same
theoretical basis.
Most analytical GIS use is based on the analysis of spatial or statistical patterns, hence on the quantitative
aspects of the archaeological data. Before we can use either data type as GIS layers, a stage of source
criticism (see Chapter 4 on bias modeling) must be applied. Next, methods for recording and describing
the data must be standardized to some extent in order to allow comparison between surveys, and here
again the premises of landscape archaeology influence the approaches chosen (chapter 13).
1-
3
V A N L EUSEN : P A T T E R N T O P ROCESS
2
STRUCTURE OF THIS THESIS
As once Gaul, the body of this thesis is divided into three parts – methodological studies, field work, and
case studies. Most of the individual chapters have been conceived as separate articles, indeed some have
already been published or are currently in publication, so each study carries its own introduction and
conclusion, its own set of bibliographic references and, where appropriate, acknowledgements. Whilst I
have tried to avoid duplication of text and figures as much as possible, and have liberally cross-referenced
between the chapters, I have given precedence to the need to present the chapters in the form of selfcontained units. In order to clarify to the reader how all the parts of this loose structure interconnect and
serve to reinforce each other, it is presented in some detail below, and an attempt at its graphical
representation has been made in figure 1.
Figure 1 - Structure of this thesis. Middle range theory and models derived from archaeological theory
are used to explain patterns in the data being derived from the archaeological landscape; field work
is carried out in order to test and improve the models and develop an appropriate methodology.
The next two chapters in this introductory part of the thesis are intended to provide the broad
archaeological and theoretical context to the research reported in the main body of the thesis. The central
concepts and definitions employed by the RPC project are reviewed in chapter 2, which also defines and
describes broad stages and variations in the processes of centralization, urbanization and colonization in
the first millennium BC in Italy. It provides brief qualitative, quantitative, and spatial descriptions of these
processes, and discusses differences and similarities of the three regions. Since the emphasis throughout
this thesis is on the detection and explanation of spatial patterning on different scales, the potential for
spatial analysis of archaeological data is discussed at two scales – that of the individual survey and that of
the region – and the use of GIS in the modeling of archaeological landscapes is introduced here because it
imposes certain restrictions on the types of analysis that can be handled. The introduction to the
1 -
4
INTRODUCTION
thesis is completed with a chapter on the process of romanization in the Wroxeter Hinterland,
reproduced from White & Van Leusen 1997 (chapter 3).
Following the body of the work, a final chapter (18) considers how the conclusions reached in these
studies affect our view of the settlement dynamics and interregional differences in the study areas, and
what recommendations can be made for future research and development in the field of landscape
archaeology, GIS, and the management of archaeological resources.
2.1
METHODOLOGICAL S TUDIES
My methodological bent is evident in the four chapters that make up this part of the thesis. The
methodological studies conducted for this thesis are all concerned with the use of GIS for landscape
archaeological research, either by understanding existing data sets in terms of their formation (chapter 4),
by reviewing maturing research areas such as predictive modeling (chapter 5) and cognitive landscape
analysis (chapter 6), or by examining its role in driving the evolution of field recording techniques
(chapter 7).
BIAS MODELLING
The first (chapter 4) is about methods for recording so-called bias factors, i.e. factors that distort our
picture of the archaeological record, and methods for correcting these distortions. It is based on an article
published in 1996 (Van Leusen 1996a), which established my aims and general approaches to this subject
in the context of the Wroxeter Hinterland Project. The chapter is about recent post-depositional and
research biases in the kind of data that form the basis for archaeological landscape reconstruction and
settlement history – site-based data collated from desktop study and older surveys, and land parcel-based
data coming from modern surveys. Recent post-depositional biases are nearly exclusively related to
human changes to the landscape and its use; research biases are those biases that have occurred in the
past, and still occur, during the construction of the archaeological record; I specifically exclude biases
occurring during the site formation process. My main point is that recent post-depositional and research
biases can not just obscure, but also create patterns in the archaeological record. This has two
consequences: firstly, if significant biases in the data we work with are not dealt with, then our
reconstructions based on those data will be significantly flawed; secondly, comparison of the
archaeological records of the three RPC Project study regions is predicated on the assumption that such
records are, or can be made, comparable. Two case studies (chapters 14 and 17) demonstrate my
approach in practice at two geographical scales, one concentrating on the role of ‘discovery’biases in the
regional site database used for the WHP; the other on the effect of subrecent large-scale landscape
changes on survey results in the Pontine Region.
PREDICTIVE MODELLING
My second chapter on method (chapter 5) consists of a paper on the methodology of predictive modeling
of archaeological site distributions, which is based on the extrapolation of geographical patterns and
correlations in order to describe and predict typical locations where specific types of archaeological
remains may be expected to occur. The chapter grew out of my earlier review of Dutch approaches to
archaeological predictive modeling (Van Leusen 1996b), but many aspects were further developed in
subsequent irregular meetings and discussions in the period 1998-2000 with members of the ‘bath-house’
group: Harry Fokkens, Hans Kamermans, Jos Deeben, Daan Hallewas, Jan Kolen, Ronald Wiemer, Eelco
Rensink, Philip Verhagen and Milco Wansleeben. Philip, Milco and I co-authored and presented a
previous incarnation of this paper at the 4th international conference 'Archäologie und Computer'
(Vienna 1999, published as Verhagen et al. 2000), and the group as a whole recently submitted a
successful proposal for an in-depth study of the role of predictive modeling in archaeological resource
management to the Dutch Foundation for Scientific Research (NWO, Kamermans 2001). The chapter
presented here, however, substantially reflects my own personal research and opinions with regard to
predictive modeling. In it, I argue that many improvements are necessary to the current predictive
modeling methodology as practiced by Dutch and international modelers, before the method can be
1-
5
V A N L EUSEN : P A T T E R N T O P ROCESS
labeled either reliable or useful; and several likely avenues for future research and development are
outline.
COGNITIVE MODELLING
The third chapter (chapter 6) is concerned with visibility and accessibility modeling, two geographical
analysis techniques only recently made feasible and popular among archaeologists by the spread of GIS.
Both techniques are being used in attempts to model the social / cognitive, rather than the physical /
economic, landscapes of the past. The text of this chapter is substantially enlarged and updated from an
article I published earlier (Van Leusen 1999), and critically reviews the majority of accessible
archaeological studies based on viewshed and cost surface analysis over the decade 1990-2000. The two
techniques are discussed together because of certain similarities in methodology and underlying
theoretical principles, which express an emphasis on the human experience of being and moving in the
landscape; not surprisingly, they have been at the center of processual – postprocessual debate almost
from the beginning. This chapter should be read in conjunction with the case studies presented in
chapters 15 and 16, which investigate aspects of ‘dominance’, territory, and accessibility arising from
current archaeological thinking about the role of Late Iron Age hillforts and markets in the Wroxeter
hinterland, of early Roman colonies on the Lepine Margin, and of protohistoric settlements in the
Sibaritide.
FIELDWORK METHODS
The fourth chapter on methodology (chapter 7) deals with the very practical question of how to conduct
a field walking survey with a minimum of effort and error, and discusses experiments conducted during
the SIBA2000 campaign with the use of self-locating digital handheld computers. This text has been
presented by my co-author, Dr Nick Ryan, at the annual conference of CAA (Visby, Sweden, 2001), and
will be published in the CAA proceedings for 2002 (Ryan & Van Leusen, forthcoming). The article
describes ongoing development of the FieldNote system at the Department of Computer Science,
University of Kent at Canterbury, UK. A new version of this portable system for self-location, mapping,
and note-taking during archaeological fieldwork, was field tested in October 2000 during a systematic
survey in the Sibaritide (Calabria, Italy; see chapter 12). In these tests, the system was used for wide-area
mapping tasks for the first time, and proved to be very useful in mapping field boundaries, highland
transhumance routes, and archaeological sites to a specified accuracy and in the absence of detailed up-todate topographic maps. It also proved useful in navigation and in re-locating archaeological sites mapped
in the 1960's. The article presents the results of these field tests and discusses their significance for future
survey design and methodology, emphasizing the trade-off between speed of operation and accuracy.
Lines for further development of the system, including improvements to both interface and functionality
are set out as well.
2.2
FIELD WALKING CA MPAIGNS
In this part are collected the preliminary reports of the four field surveys which I conducted between
1998 and 2000 with other members of the RPC project in all three Italian study regions. The reports
themselves have already been published or are in press, and are preceded here by a chapter (8) reviewing
the aims, approaches, and results of the RPC field work. The Lepine Margin: Ninfa 1998 (chapter 9, Van
Leusen 1998) was published in Assemblage, on-line journal of the graduate students of the Archaeological
Institute at the University of Sheffield; The Pontine Margin: Fogliano 1998-9 (chapter 10, Attema et al.
2001) is published in Palaeohistoria, the annual journal of the Groningen Institute of Archaeology; The
Salento Margin: Ostuni 1999 (chapter 11, Attema et al. forthcoming) will be published in Studi di Antichitá;
and the preliminary report on the Sibaritide 2000 survey appearing here as chapter 12 has been submitted
for publication in Palaeohistoria as well.
It may be asked why field work was included at all in a synthetic project such as the RPC project. There a
two reasons for this. Prior to the start of the my research, I was unfamiliar with the archaeology and the
landscapes of central and southern Italy. The fieldwork has been essential in providing first-hand
1 -
6
INTRODUCTION
experience of the landscape, its scale and characteristic features, the climate, and myriad circumstances
which currently preserve or destroy, hide or present archaeological remains. The second and more formal
reason is that in providing methodological control over poorly understood regional data sets, the
fieldwork represented an essential phase of ‘source criticism’without which any synthetic work would
have to appear of doubtful value.
2.3
CASE STUDIES
The case studies in this part of the thesis explore and demonstrate several of the issues raised in the
methodological studies presented in chapters 2, 4, and 6, and most have already been referenced there.
BIAS MODELS
The case studies in chapters 14 and 17 explore the extent of modern land use and research biases in
regional archaeological records (cf. chapter 4) at two different spatial scales. Chapter 17 presents an
investigation of changes in land form associated with the land improvement scheme to which the Pontine
plain was subjected in the late 1920’s and early 1930’s, and their influence on the results of the Fogliano
survey (cf. chapter 10). This was the subject of an elective study under my supervision by graduate
student Hendrik Feiken during 1999-2000, subsequently presented as a joint paper to the CAA 2000
conference and published in its proceedings as Feiken & Van Leusen 2001. Chapter 14 explores the
correlation between land qualities, land use and land cover on the one hand, and the formation of the
regional archaeological site record of the Wroxeter Hinterland on the other (see chapter 3 for an
introduction). It is found that, whereas most large-scale patterns in this record are strongly correlated to
the opportunities for discovery afforded by the combination of land use / land cover (LULC) and
research methods in the 2nd half of the 20th century, a historical LULC reconstruction suggests that
Wroxeter’s territory has been extremely stable in the longe durée – indicating that the subrecent pattern of
land use in the region might be similar to that of the late Roman period.
COGNITIVE AND LOCATIONAL MODELS
The case studies presented in chapters 15 and 16 demonstrate issues currently at the heart of the debate
on GIS predictive modeling (chapter 5), namely the modeling of ‘cognitive’aspects of the landscape as
opposed to the usual physico-economic modeling. Cases taken from both the RPC project area (chapter
15) and the Wroxeter Hinterland (chapter 16), and ranging in time from the Bronze Age to the Roman
period, are used to explore the concepts of dominance and accessibility through the GIS techniques of
viewshed analysis and cost surface analysis (chapter 6). These concepts are closely related to those of
centralization, urbanization, and Romanization presented in chapter 2.
INTRA-REGIONAL COMPARISON
The discussion of the theoretical basis of, and potential approaches to, interregional comparison of
settlement and land use histories in chapter 2 further explored in a case study concentrating on the
Pontine region (chapter 13). Two aspects of such comparison are explored in depth: the creation of a
GIS-enabled regional archaeological database, and the quantitative comparison of data sets collected
within the same, adjacent, and more widely separated landscape units. An attempt is made to assess the
extent to which the comparison of field survey results can throw light on the three core processes studied
by the RPC project - centralization, urbanization, and colonization – in southern Italy.
REFERENCES
Attema, P 1996
Naar een archeologie van vroege urbanisatievormen in het Italische cultuurlandschap: Midden- en ZuidItalische kustlandschappen tussen internationalisering en kolonisatie, exploitatie en marginalisering (1400
v.Chr - 400 n.Chr) [Towards an archaeology of early forms of urbanization of the Italic culture landscape:
Central and South Italian mediterranean coastal landscapes caught between internationalization and
1-
7
V A N L EUSEN : P A T T E R N T O P ROCESS
colonization, exploitation and marginalization (1400 BC - 400 AD)]. Project proposal, 3rd version, dated
10/02/96.
Attema, PAJ, G-J Burgers, M Kleibrink & DG Yntema 1998
Case studies in early Italian centralization and urbanization: a Dutch perspective, European Journal of
Archaeology 1(3):326-381.
Attema, PAJ, G-J Burgers, E van Joolen, PM van Leusen & B Mater (eds) forthcoming
Regional Pathways to Complexity. Proceedings of a conference on the archaeology of early Italian
urbanization held at Groningen (April 2000). BAR Int. Ser.
Attema, P, G-J Burgers & M van Leusen, forthcoming
Walking the Murge: The Ostuni Field Survey Campaign (Apulia, Italy). Studi di Antichitá 11 (2001).
Attema, P, E van Joolen, & M van Leusen 2001
A Marginal Landscape: Field work on the beach ridge complex near Fogliano (South Lazio). In:
PaleoHistoria 40/41.
Feiken, H & M van Leusen 2001
Interpreting Field Survey Results In the Light of Historic Relief Change: the Fogliano beach ridges (south
Lazio, Italy), in Stancic, Z & T Veljanovski (eds), Computing Archaeology for Understanding the Past.
Proceedings of the CAA2000 conference (BAR International Series 931): 205-211. Oxford: Archaeopress.
Ryan, N & M van Leusen, forthcoming
Educating the Digital Fieldwork Assistant, paper presented at CAA2001.
van Leusen, PM 1993
Cartographic Modeling in a Cell-Based GIS, in Andresen, J, T Madsen & I Scollar (eds), Predicting the
Past. Computer Applications and Quantitative Methods in Archaeology 1992: 105-124. Aarhus: Aarhus
University Press.
van Leusen, PM 1995
GIS and Archaeological Resource Management: a European Agenda, in Lock, G & Z Stancic (eds), GIS
and Archaeology: a European Perspective: 27-42. London: Francis & Taylor.
van Leusen, PM 1996a
Unbiasing the Archaeological Record, in Archeologia e Calcolatori 7:129-36
van Leusen, PM 1996b
GIS and Locational Modeling in Dutch Archaeology: a review of current approaches, in Maschner, HD
(ed), New Methods, Old Problems. Geographic Information Systems in Modern Archaeological Research
(Southern Illinois University Center for Archaeological Investigations Occasional Paper 23):177-197.
Carbondale: Center for Archaeological Investigations.
van Leusen, PM 1998
Archaic Settlement and Early Roman Colonization of the Lepine Foothills, in Assemblage 4 (1998),
http://www.shef.ac.uk/~assem/4/.
van Leusen, PM 1999
Viewshed and Cost Surface Analysis Using GIS (Cartographic Modeling in a Cell-Based GIS II), in
Barceló, JA, I Briz & A Vila (eds), New Techniques for Old Times. CAA 98. Computer Applications and
Quantitative Methods in Archaeology. Proceedings of the 26th Conference. BAR International Series
757:215-223.
van Leusen, PM in press
Intensive Surveying in a Marginal Landscape, in Darvill, T & M Gojda (eds), One Land, Many Landscapes.
BAR int ser. Oxford: Archaeopress.
van Leusen, PM & VL Gaffney 1996
Extending GIS Methods for Regional Archaeology: the Wroxeter Hinterland Project, in Kamermans, H &
K Fennema (eds), Interfacing the Past. Computer Applications and Quantitative Methods in Archaeology
1995 (Analecta Praehistorica Leidensia 28):297-305.
Verhagen, P, M Wansleeben & M van Leusen 2000
Predictive Modeling in the Netherlands. The prediction of archaeological values in Cultural Resource
Management and academic research. In: Harl, O (ed), Archäologie und Computer 1999.
Forschungsgeselschaft Wiener Stadtarchäeologie 4: 66-82.
White, RH & Ph Barker 1998
Wroxeter. Life & Death of a Roman City. Stroud: Tempus.
1 -
8
C HAPTER 2
PAT T E R N S A N D P R O C E S S E S
1
INTRODUCTION
The archaeological themes investigated in this thesis are to a large extent those of the individual projects I
have been part of since 1994. The nature of changes in settlement and land use patterns spanning the
Iron Age / Roman transition in an outlying part of Britain has been the subject of the Wroxeter
Hinterland project; the comparison, from the late Bronze Age to the Roman Empire, of long-term
processes of centralization, urbanization, and colonization in the three Italian regions of the Pontino, the
Salento Isthmus, and the Sibaritide has been the aim of the Regional Pathways to Complexity project.
Such aims require the careful definition and detailed description of the core concepts and processes
involved, of the available general explanatory frameworks for socio-political change, and a justification of
the chosen methodology to approach this aim. The interpretation of the available archaeological and
historical evidence in terms of processes of centralization, urbanization, and colonization (Hellenization,
Romanization) taking place in ancient Italian societies must rely heavily on appropriate middle range
theory – the theory concerning the cultural transformations and formation processes giving rise to the
archaeological record. Our methods of collecting and interpreting the raw data themselves must be
studied in order to assess the tendency of the methodology employed both by us and by previous
researchers to produce spurious patterns in those data; we cannot avoid taking this step if we want to feel
secure in our interpretations.
The current chapter is therefore devoted to a review of the concepts, theories, and methods employed
later on in the case studies. A crucial part of the theoretical discussion also concerns the basis for
interregional comparison. What makes us think that the historical trajectories of the three study regions
can be compared at all? If we feel they can profitably be compared, what yardsticks are we going to use?
2
REGIONAL SETT LEMENT DYNAMICS
A review of the current consensus among Anglophone researchers regarding the settlement dynamics of
the three study regions provides the basis for a discussion of the core concepts and terms being used to
describe and understand the presumptive societal processes operating in 1st Millennium BC. We begin
with a brief chronological review of the main settlement dynamics in the three study areas. In a second
section, the interrelating concepts and processes central to the RPC project are reviewed; this is followed
by a consideration of the continued significance of the traditional concepts of ‘Hellenization’ and
‘Romanization’.
2 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
2.1
SETTLEMENT DYNAM ICS OF THE STUDY ARE AS
THE PONTINE REGION
The study area of the Pontine Region Project is located about 60 km south of Rome and comprises part
of the volcanic landscape of the Alban Hills, the limestone range of the Monti Lepini and the coastal plain
of the Agro Pontino bordering on the Tyrrhenian sea. Surveys in this area focus on the 1st millennium BC
(Latial protohistory and Roman Republican period), though the Bronze Age and the Imperial period are
considered as well. In the first five years of the project (1987-1993) research mainly focused on the preRoman landscape. In addition to extensive transect surveys and environmental research, three
protohistoric settlement areas were intensively surveyed. In a follow-up program (1994-1997) the impact
of early Roman colonization on the protohistoric landscape was investigated in three sample areas. The
final publication for this is in preparation. The research of the PRP was carried out in close collaboration
with the Latial Pottery Research Group and the Satricum excavation team, both at the GIA.
Figure 1 – The Pontine Region, with indication of the main surveys and excavations
which had been undertaken by the Groningen Institute of Archaeology before the start
of the RPC project.
The Tyrrhenian coast of central Italy is generally believed to be the area where the earliest urbanization
within Italy took place. Settlements begin to nucleate during the final Bronze Age and early Iron Age,
with differentiation in grave goods indicating the growth of social hierarchies (Peroni 1994: 221-5). A this
early stage, the settlements are ‘proto-urban’ in the sense that many will later develop into urban
settlements and the cores of early states. Attema (1993:217) suggests that Satricum and Caracupa in the
Pontine region began their existence in the early Iron Age as gathering places with cultic and territorial
marker functions for transhumant groups claiming rights in the winter pastures of the Pontine plain, and
2 -
2
PATTERNS AND PROCESSES
only later developed into settlements1. In this scenario, urbanization in south Lazio only began in the 8th
century BC and culminated with the large Archaic centers of the 6th century. Urbanization and large scale
trade in agricultural products developing in the Archaic period are preceded by an approximately threefold population growth during the Orientalizing period (7th century BC), allowing Etruscan settlements to
be classified into a hierarchy of types (Perkins 1999:104-6).
The term ‘colonialism’ may under some circumstances be interpreted as indicating a conscious
movement, a policy. The early (5th and 4th century) colonies in southern Latium may have been part of
such a strategic movement on the part of the early Roman state, aiming to secure the disputed
borderlands of Latium Vetus, and was certainly represented by later writers (esp. Livy) as such. The
strategic value of a colony was expressed with admirable clarity by Machiavelli in The Prince (translation
Bondanelli 1984:10)
“The other and better solution [to securing new territory] is to send colonies into one or two places that
will act as supports for your own state (… ) Colonies do not cost much, and with little or no expense a
prince can send and maintain them; and in doing so he hurts only those whose fields and houses have
been taken and given to the new inhabitants, who are only a small part of that state; and those that he
hurts, being dispersed and poor, can never be a threat to him, and all others remain on the one hand
unharmed (and because of this, they should remain silent), and on the other afraid of making a mistake,
for fear that what happened to those who were dispossessed might happen to them.”
But Livy may have been telescoping events that took place almost two centuries apart, and we should not
discount the possibility that the reality of the earliest Romanization of Latium Vetus was much more
haphazard and unplanned, and contingent on historical circumstances than he suggests. In contrast,
Roman colonies established in the later 4th century BC appear to have been more systematic and “urban”
ventures. According to Attema (1993:13) they expressed a territorial organization totally different from
the Archaic patterns of settlement and land use, involving such structural innovations as the putting into
place of a system of rural villas for olive culture on the Lepine footslopes, centuriation of agricultural land
near Terracina, and drainage and road-building (the Via Appia through the Pontine marshes). Olive
culture requires a large investment in establishing plantations, but perhaps the villas could exploit an
existing (less intensive) agricultural system by the Archaic and post-Archaic Latial peoples2. These colonial
towns must each have had rights over parts of the Lepine upland and the Pontine plain, up to the Via
Appia or even beyond, possibly even including fishing and fowling rights in the coastal environment. But,
since recent intensive surveys in marginal landscape units within the Pontine region indicate that a large
measure of settlement continuity may have been present between the Archaic and Republican periods
(see chapters 9 and 10), the impact of 4th century colonization may have been restricted to selected ‘core’
parts of the region. Especially strategic locations on the Lepine margin.
In this scenario, two temporal gaps remain to be accounted for. Firstly, what happened in between the
historical establishment of the colonies around 500 BC and their first archaeological appearance around
the middle of the 4th century BC? And secondly, why was the apparent establishment of systematic olive
culture in the Lepine margins delayed for more than a century after the 4th century colonization? There is
no archaeological evidence for the presence of Roman colonists in this region in the Archaic or postArchaic periods. It appears that the early process of colonization by Rome was either largely unplanned
and long drawn-out, or it ran into unexpected difficulties and was discontinued. However, recent field
surveys indicate that a densely settled Archaic (and, less clearly, post-Archaic) landscape existed in at least
1 In view of the new radiocarbon dates and improved understanding of site formation processes at these sites, this process must
be dated some 50 – 100 years earlier.
2 Pollen evidence indicates that olive culture on a substantive scale began in the 3rd century BC in this region (Haagsma, in
Attema 1993).
2 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
two ‘marginal’parts of the Pontine region, with no evidence for the displacement of this indigenous
population by Romans in the later 3rd century BC (chapters 9 and 10).
Judging from the urban architectural remains at Setia (Sezze) and the late 4th century infill of its immediate
rural surroundings (ager), Roman colonization here took place as much as two centuries earlier than in the
coastal landscape near the Fogliano lagoon. Clearly, if there was a planned or sustained Roman policy of
agricultural colonization in the 4th century BC, it extended only over limited and specific areas. If aerial
photographic indications for a 4th century centuriation of the central part of the Pontine graben between
Sezze and Terracina, predating the construction of the Via Appia about 326 BC, are correct (Cancellieri
1990), then the area colonized would have mostly consisted of previously marginal land – the macchia of
pre-Roman Setia. Agricultural production in the Pontine region became less important to the expanding
Roman empire after 200 BC anyway, and there are indications that the coastal area may at that time have
specialized in fish farming for the market at Rome (Attema et al. 2001).
Later instances of colonization by Rome took place at a much greater pace and left abrupt changes in the
archaeological record; it is generally believed that, once the expanding Roman Republic had acquired
sufficient experience from its earlier attempts at colonization, it was capable of planning and executing a
rapid colonization policy ‘package’. But even then, local circumstances would influence the speed and
success of the process. In the case of Wroxeter, after a very brief military phase in the middle of the 1st
century AD, local elites were co-opted into the political and administrative structures of the Roman state
such that the processes of centralization, urbanization and romanization were essentially complete within
the next 25 years (chapter 3).
THE SALENTO ISTHMUS
Research in the Salento Isthmus (in the heel of Italy) was started in 1981 by Prof. Dr. Joh. Boersma and
Dr. D.G. Yntema of the AIVU, in close collaboration with the Scuola di Specializzazione in Archeologia
Classica e Medievale of the University of Lecce. Its aim was to elucidate the development of regional
settlement patterns in the Brindisi region in the context of the integration of native society into the
Roman world. AIVU surveys have covered a total area of some 90 square km, incorporating various
environmental zones. Between 1981 and 1983 surveys were concentrated on an area of circa 63 square
km around the town of Oria, the roots of which go back to well within the Bronze Age. In 1989 and 1990
another team conducted a field survey in an 18 square km large transect between the Adriatic and the
ancient fortified site of Valesio. The AIVU has been engaged in a complete survey and partial excavation
of its walled area since 1984; this being the first systematic urban survey in the area.
In 1991 the Brindisino research was extended to include the Taranto plain, and renamed the Salento
Isthmus Project, under the direction of Drs. Gert-Jan Burgers and subsidized by NWO. Four complex
settlement areas and their surroundings were incorporated into the regional survey project - Muro
Tenente, Muro Maurizio, Li Castelli di San Pancrazio, and Cellino San Marco. Through intensive total
coverage surveys of these more or less urban settlements their chronology, extent, lay-out, occupational
density and nature was studied. From 1993 onwards the AIVU has been engaged in large scale
excavations at Muro Tenente.
2 -
4
PATTERNS AND PROCESSES
Figure 2 – The Salento Isthmus, with indication of the main survey and excavation projects
which had been undertaken by the Archaeological Institute of the Free University of
Amsterdam before the start of the project. For legend, see figure 1.
In the Late Bronze Age (14th-12th centuries BC) of the Salento peninsula, a preference for coastal
locations is suggested by the presence of large enclosed promontory settlements such as the one at
Masseria Risieddi (see chapter 11). These settlements are thought to be located here for easy participation
in overseas communication networks (D’Andria 1991:403). On the basis of the lack of finds in surveys
conducted by the AIVU, and the predominantly coastal location of the known sites of the FBA and Early
Iron Age (11th-9th centuries BC), Burgers concludes that the pattern of a relatively empty inland Salento
landscape continued (1998:173), although some inland sites are known from this period (Oria and Monte
Salete).
The question of continuity or discontinuity in settlement dynamics at the end of the Late Bronze Age and
the beginning of the FBA has been a subject of intense debate. Some have emphasized a continuous
process of increasing complexity (notably Peroni 1979), while others insist on radical cultural disruption
and a subsequent invasion of tribal elements from Illyria, typified by violent destruction, abandonment or
restructuring of Late Bronze Age settlements (De Juliis 1988: 9-19). Still others recognize a relative
continuity in interregional networks, emphasizing instead the collapse of overseas exchange with the
Mycenean world which could have been the cause of a decreasing complexity of society in general
(Yntema 1990:38-39, Yntema 1993:154). Coastal communities at this time ‘were probably autarchic, while
the inland may have been exploited for extensive pastoralism, if at all’(Burgers 1998:174 and note 95).
In the Iron Age (9th and 8th centuries BC), native communities were ‘engaged … in settlement expansion,
territorial reorganization, demographic growth, increasing rural use of the interior of Salento, and
overseas contacts’ in order to ‘enhance internal power positions’ (Burgers 1998:296). Demographic
growth and influx from the Balkan area are thought to have driven a strong development of the village
system and a gradual occupation of all available agricultural land in southern Puglia (D’Andria 1998:108).
This crystallization of a settlement system developing since the 9th century BC was interrupted on the
western side of the Isthmus at the start of the 7th century by the establishment of the colony of Taras
which carved out its chora. The origins of the early Hellenistic fortified sites in the Brindisi region and the
larger Salento peninsula could still be traced by Burgers (1998:293-6) to the Iron Age, because ‘the earliest
diagnostic artefacts found are Iron Age matt-painted ceramics’. The founding of these settlements is
thought to be an expression of a larger gradual process of landscape reorganization, accelerating from the
late 8th century BC to include the interior of the peninsula and other outlying regions such as the area
around Ostuni (chapter 11). By the 6th century BC a two-level hierarchy (or, following Semeraro
2 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
1997, ‘hieratic system’) had come into existence, with three large towns (Oria, Cavallino, and Ugento)
surrounded by a larger number of small villages.
Urban features, such as the growth in size of single dominant settlements, social and economic
hierarchies, and the construction of monumental buildings, were all appearing in southern Italy between
the 6th and the early 4th century BC, a process of urbanization which accelerated in the early Hellenistic
period (later 4th and 3rd centuries; Lomas 1993, Burgers 1998:293). A settlement system emerged that was
dominated by a series of fortified towns. In the later Hellenistic period this system disintegrated as the
region became increasingly involved in supra-regional conflicts. Integration of the Salento into the
expanding Roman state started with the defeat of the Tarentine/Messapian allies in the first half of the 3rd
century BC, and was spearheaded by the colony of Brundisium (founded 245 BC). 3rd Century wars, and
especially the 2nd Punic war, resulted in massive disruption of the old native/Greek culture, but recent
research indicates that it was not everywhere replaced by an exploitative large-scale colonial economy –
instead, differences can be attested within the region (Burgers 1998:30-31). For the native elites of the
Salento Isthmus, close association with the Roman state and way of life became the means to further
oneself. Towns away from the central axis of the Via Appia decline, while Brundisium becomes the focus
of Roman and native surplus production. The increased market orientation in the production of olive oil
and wine caused rationalization and concentration of farm labor both locally and, regionally, near the Via
Appia (Burgers 1998:303-7).
THE SIBARITIDE
The Sibaritide, consisting of a coastal flat and its surrounding hills, has been the subject of research by the
GIA since 1990. The main object of study has been a system of Bronze Age hilltop sites, one of which,
near the town of Francavilla Marittima, consists of a low hill called 'Timpone Motta' with a sanctuary on
top and settlements of huts and houses on three lower 'plateaus'. It also includes the general catchment of
this hill and connected areas such as the adjacent Iron Age Macchiabate necropolis, part of which was
excavated in the 1960s by Paola Zancani-Montuoro. Following a 20 year gap, the project was revived by
M. Kleibrink with test campaigns in 1990/91 in order to research the problem of native Enotrian versus
Greek settlement and colonization activities - particularly the relationship with the nearby Greek colony
of Sybaris. This new research brought to light a number of huts from the Middle Bronze Age and early
Iron Age, as well as a 8th - 7th century cult activity area and a 6th century 'colonial' village on the lower
slopes of the Timpone Motta (Maaskant Kleibrink 1993; for recent publications containing further
references, see Kleibrink 2000 and Kleibrink & Sangineto 1999).
The recent mapping and discussion of pre- and protohistoric settlement in northern Calabria by Peroni
and Trucco (1994) has served to modify the old ‘textbook’view of a largely pastoral society in the
Apennine period (Puglisi 1959) 3. One set of more elevated settlements, situated on calcareous outcrops
and connected most likely with specialized transhumant pastoralism since they lie along routes still used
today to reach the higher mountain ranges, is now thought to have been "under the control" of another
set of larger, lower-lying settlements4. Following Barker (1985), Peroni points to the spread of dry farming
in the Middle Bronze Age period in the higher valleys as the causative factor for the development of this
latter set. Because many of its settlements are situated on old fluvio-marine terraces, consisting of
conglomerate and/or sand, Peroni could plausibly argue that settled Middle Bronze Age society preferred
well
Peroni proposed the following chronology for the Bronze Age and early Iron Age: Middle Bronze Age 1600 – 1300 (300
years), Recent Bronze Age 1300 – 1150 (150 years), Final Bronze Age 1150 – 900 (250 years), and Early Iron Age 900 – 700 (200
years).
3
This is an unfortunate turn of phrase in my view, which is also used by other authors. A more neutral expression would be: “are
part of the same system as”.
4
2 -
6
PATTERNS AND PROCESSES
Figure 3 – The Sibaritide, with indication of the main survey and excavation projects which had been
undertaken by the Groningen Institute of Archaeology before the start of the project. For legend, see
figure 1.
defensible hills and terraces on which to practice dry farming of cereals and vegetables (Peroni & Trucco
1994:37). In the excavations at Broglio di Trebisacce Peroni saw signs of increasing demographic pressure
in the course of the protohistoric period. The presence, in the Broglio excavations, of Mycenean wares
dating to the 15th – 12th centuries BC has been seen as an indication that Greek potters were present in
the Sibaritide at that time, which would argue for an early hierarchization of indigenous society, in which
elites maintain long-distance contacts and channel tribal surpluses into the acquisition of prestige goods.
In the Late Bronze Age, Torre Mordillo became a major settlement with a defensive wall (agger) encircling
the highest plateau (Arancio et al. 1994), Late Bronze Age layers at Torre Mordillo, Broglio and
Francavilla Marittima contain Italo-Grey ware of different fabric qualities indicating local ceramic
production, and evidence for wider use and cultivation of the olive tree (Peroni & Trucco 1994:45). This
evidence for Late Bronze Age exchange of objects and technology indicates that overseas contacts with
the Aegean must have been frequent (Peroni 1994:24) and presumably profitable for both sides. The Late
Bronze Age in the central Mediterranean is therefore seen as ‘a clear case of expansion of people,
technology, ideas and products’(Kleibrink forthcoming par. 3.2), with autonomous settlements occupying
all viable river-delimited territories in the Sibaritide, and powerful leaders controlling the means of
production and redistribution (idem, par. 3.6/7).
As in the Salento, overseas and possibly supra-regional contacts broke down during the FBA, but the
settlement system continued to crystallize. The occurrence of pairs of defensible hilltop sites of unequal
size and agricultural potential has been interpreted by Peroni in terms of a defensive strategy, with the
smaller and higher of the pair having little agricultural potential and a role in defending the lower
settlement and its agricultural riches in which a large section of the population would have lived. One
example of such a pair, brought to light by surveys, is the following:
2 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
1. Monte San Nicola, a site overlooking both the Raganello valley and the Sibari plain, at about 500 m asl.
Potsherds were found here on a plateau of circa 1 ha; the encircling terraces lower down are much
larger and together form an area fit for cultivation of circa 25 ha (Peroni & Trucco 1994, no 31).
2. Monte Spirito Santo, a site situated about 1 km northwest of Monte San Nicola at an elevation of circa
510 m asl; it is located on a much smaller plateau and without a view over the plain. The site is
connected with the steep slopes of the Raganello near Cività and on its southern end with similar
steep terraces of the Eiano (Peroni & Trucco 1994, no 30).
By the beginning of the Late Iron Age, the two most important settlements in the Sibaritide appear to
have been located at Torre Mordillo and at Francavilla Marittima, and it is thought that these were
respectively the economic and cultic centers of the region.
Large sections of the coastal landscape may have been of marginal significance to the indigenous tribal
societies, so that the establishment in the late 8th century BC of a Greek trading emporium which later
evolved into a colony need not be seen in terms of conflicting interests at all. The earliest indications of
classical ‘colonial’influence in the Sibaritide are the Aegean style temples built on the Timpone della
Motta presumably with the active help of early Greek traders (Kleibrink, in Attema et al. 1998:127), but it
is only in the mid-7th century that Greek colonists would finally claim the sanctuary as theirs by rebuilding
it in a fully Greek style. From about 640 BC, Greek and colonial pottery became the dominant gift both
at the cult center on the Timpone della Motta of Francavilla Marittima and in grave inventories of the
nearby Macchiabate necropolis. It is noteworthy that the first indications for the rural spreading of Greek
pottery out into the Sibaritide foothills date to the 6th and 5th centuries BC, making it likely that the chora
of Sybaris did not yet extend beyond the coastal plain around 600 BC (see chapter 12). If Kleibrink is
correct in dating the latest rich burials at Macchiabate to the first decades of the 6th century, then we may
suppose that the local elite by that time found it opportune to relocate themselves to Sybaris – a sign that
the nearest indigenous polities of the foothills were rapidly being absorbed by the colony by then.
Historical sources claim that Sybaris, in its (6th century) heyday, ruled over four tribes and 25 towns
(Strabo VI, 1, 13). Certainly it was instrumental in founding a further colony as far away as Metapontion,
and dominating others as far away as Laos on the opposite coast of the Calabrian peninsula. By the
middle of the 6th century BC the town shows an enormous expansion from Stombi to the outer walls, and
regular plan buildings must have appeared as at Amendolara and elsewhere. Pottery production became
standardized. For example, there is no difference between the local soft ware productions at Sibari,
Amendolara and Francavilla (Attema et al. 1997/98) and, whilst the 7th century BC matt-painted tradition
at Francavilla still showed a peculiar and original development, the 6th century BC brings standardization
in both architecture and pottery. By then, over one century of acculturation between natives and Greeks
must have created a new social structure in which Sybaris (and later Thurioi) became the regional urban
administrative and economic center for an expanding rural hinterland of villages and isolated farms with
associated rural sanctuaries and cemeteries. While the settlement history of the plain must remain largely
unknown because of later substantial alluviation, surveys indicate that farmsteads began to appear in the
foothills, some 15 km away from the urban center itself, sometime in the 6th or 5th century. Clear evidence
for settlement expansion in the foothills becomes available only with late 4th century Hellenistic fine wares;
by the 3rd century, isolated Hellenistic farms even occur far into the highlands at elevations up to 1000 m asl.
It would appear that the inhabitants of the Timpone della Motta managed to adapt to the presence of the
powerful Greek colony without losing their social structure. The site was only finally abandoned in the 5th
century BC, possibly because it lost its function as an extra-urban sanctuary when Sybaris was destroyed
by the Crotonides in 510 BC. Since its pan-Hellenic successor colony Thurioi was not established until
443 BC, and surveys have barely been able to identify any material from this period, something of a ‘dark
age’lurks between the late 6th century BC and the onset of the Hellenistic rural expansion phase in the
late 4th century BC.
As around Poseidonia, and unlike the Metapontino where the process happened already in the 6th century,
a major increase in settlement density appears to occur in the Sibaritide only in the late 4th century
2 -
8
PATTERNS AND PROCESSES
BC. De Neef (1998:105, 110) suggests that this may in part be related to the marshy nature of both plains,
the large scale drainage of which would only have been possible in the 4th century, but it seems equally
likely that the rural colonization of the chora of Sybaris had already begun in the 6th century but was
interrupted by the conflict with Croton before it had time to reach the archaeologically visible foothill
zone. Large-scale drainage of the plain in the 4th century, directed from Thurioi, could still be invoked as
an early phase in the Hellenistic colonization process, leading to a late 4th century rural expansion into the
foothills and explaining the preponderance of Greek material culture from that period and area (see also
chapter 12).
Historians report that, during the 4th century BC, there was constant warfare with the Lucanians and
Bruttians, and Thurioi became a voluntary Roman dependency. Like the Salento, it then became involved
in the 3rd century struggles of the Romans against Pyrrhus and Hannibal and their Italian allies. After the
2nd Punic war the Romans attempted to revive the town by replanting it as the colony of Copiai in 194
BC, but according to the ancient historians it was not successful and quickly dwindled to insignificance
(Appianus, BCiv. 5.56). Even less is known of a second Roman colony at Interamnium, depicted on the
Peutinger Map and identified with the modern town of Castrovillari on the upper Coscile5.
2.2
CORE CONCEPTS AN D TERMS
It is obvious that the three processes of Centralization, Urbanization and Colonization cannot be seen as
wholly independent of each other. Centralization and urbanization are two ways in which early societies
can become more complex; colonization comes into play at a later stage when societal structures become
organized at a wider (interregional) scale, and bring other forms of urbanism. However, the terms
centralization, urbanization, and colonization have been used to encompass such a wide and ill-defined
range of processes, that they can not provide a reliable basis for comparing regional histories. For
example, Burgers (RPC in prep.) regards pre-Roman urbanization as part of a supra-regional process of
social differentiation involving most of the Mediterranean basin and affecting all of the landscape
including ‘marginal’areas. Perhaps wisely, therefore, Attema (1993:17) did not even attempt to provide a
formal definition of urbanization in his discussion of the settlement history of the Pontine region.
Neither, in my opinion, can 'urbanization' or 'colonization' be regarded as neutral concepts. Urbanization
and urbanism play a central role in the expansion of the Greek colonial and, later, Roman polities in Italy,
because the town was the focus of the classical conceptual landscape (Laurence 1994:139). Both the later
Greek colonists in Magna Graecia and the expanding Roman state were possessed of a mindset in which
the town was the center of life – and colonization would, ipso facto, have meant urbanization. To accept
the classical definition of urbanism is therefore tantamount to accepting a teleological view of history, in
which the historical urban forms of classical Greek and Roman society represent the ideal or standard by
which archaeological reality is measured. Such a view ignores the potentially different evolution of
indigenous settlement systems. Similar objections may be raised with regard to the use of the term
‘colonization’, which is colored not just by its origins in and treatment by the ancient historians but also
by 19th century western colonialism seeking justification of its 'civilizing' activities elsewhere, and assumes
that there is a fundamental inequality between the external colonizing party and the indigenous 'receiving'
party. To describe the 1000 years of history of settlement and land use in the study regions in these terms
is therefore to invite a 'classicist' bias.
These issues have recently become part of a wider ‘revisionist’trend regarding the role of the Greeks and
Romans in initiating and dominating ‘civilizing’processes in early historic central and southern Italy. For
instance, regarding the relation between colonization and urbanization, opinions expressed in the recent
literature range from McIntosh’s (1991) “urbanization was a development initiated by the colonizing
Greeks” to Van Dommelen’s (1997, 1998) “indigenous urbanization was, in some instances, influenced
by colonists”. However, most importantly, we should not expect any of the three processes to lead to
5
In my view, this identification is incorrect; see my argument in chapter 12, note 9.
2 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
uniform results irrespective of the time, place, and local history of the study area. Attema, Burgers,
Kleibrink and Yntema (1998:131), in their presentation of case studies introducing the RPC project,
noted:
‘[how] important the indigenous perspective is in the study of centralization, early urbanization and
colonization processes in Central and Southern Italy, and how regional cultures and landscapes
underwent these processes at different points in time, in different ways, with different intensities and with
different results. It is clear that the natural environment, technological level, subsistence and ideological
strategies of the local populations and the degree of early external contacts and colonization movements
were all important influences on internal regional developments.’
CENTRALISATION
The concept of centralization is applicable to societies from the tribal stage onwards, and thus plays a role
in both the protohistoric and the early historic periods. ‘Centralization’in the context of such early
societies may be defined as the process by which societal functions, and the power and control derived
from them, become geographically concentrated at a small number of locations and, socially, in a small
number of individuals. Since it is such an all-encompassing concept, archaeological evidence for
centralization may be derived in almost any context, from evidence for social differentiation in cemeteries
to evidence for the spatial concentration of settlement across a wide region. The latter type of evidence
provides a direct link with the concept of urbanization, which may be viewed as a particular type of
centralization. For example, to account for the lack of identifiable settlements in the Early and Middle
Bronze Ages of both Greece and southern Italy, a system of shifting cultivation has been proposed for
these periods; whereas the Late Bronze Age (Bronzo Recente, Bronzo Finale) saw the population move
to live together in villages, each with its field system and associated manuring spread.
The development of centralized settlement during the later Bronze Age and early Iron Age (but beginning
already in the Middle Bronze Age in some areas) cannot be explained by simple ‘external’monocausal
factors such as overseas contacts with a higher (Mycenean, Phoenician, Greek) civilization resulting in a
core-periphery transfer. Not only would this by itself have been unlikely to cause such a fundamental shift
in the organization of prehistoric societies, but dating evidence as well indicates that developments in Italy
were not lagging (sufficiently) behind those in the eastern Mediterranean to allow for this kind of
causation. Another factor which has been advanced to explain the development of centralized settlement
is that of ‘defensibility’. In my view the significance of this should not be overstated; although it is true
that Late Bronze Age centralized settlements occur in relatively defensible positions (e.g., ‘capes’and
hilltops), such locations also have other desirable qualities such as exposure to cooling breezes and large
viewsheds. It is to be considered unlikely a priori that a society could experience (or survive) such a long
period (3 to 5 centuries) of insecurity as to base its settlement structure on it.
One can imagine a competitive ‘big man’or chiefdom society in which defense is needed from the
ritualized raiding familiar from the Irish sagas… but in such a case the settlements should be interpreted
in terms of strength – as strongholds expressing the wealth of a tribal unit and prowess of its leaders –
rather than in terms of defense against attacks. In such a perspective, what could be the reason for Late
Bronze Age settlements in the Salento peninsula to be located on the edge of the Murge rather than
within it, as in the Early Iron Age? It is possible that the competitive structure based on pastoral wealth,
which required territories composed of coastal as well as upland components, was gradually replaced by
one in which agricultural wealth played a larger role. This could have been expressed in a progressive
‘carving up’of the Murge upland zone into agro-territories centered on hilltop settlements; an idea
supported by the more or less regular distances of circa 12 km separating Early Iron Age settlements in
the Murge and other south Italian upland areas, as recently mapped by D’Andria (N.D., fig. 9; see also my
chapter 11, fig 1).
2 -
10
PATTERNS AND PROCESSES
A completely different type of argument about centralization stresses the effects of research and
discovery biases, and suggests that the currently known settlements represent only the most obtrusive
remains of a much more extensive and complex settlement system. Prospection and research has tended
to focus on relatively easily discovered sites with substantial structural remains, even if these are of a later
date, while equally large sites lacking such features and lying outside areas of interest remain undiscovered
unless a systematic survey happens to hit on them (as with the several hectares of clearly visible Middle
Bronze Age impasto discovered during the Ostuni99 survey, chapter 11). If it is accepted that this effect
plays an important role, then our whole frame of reference changes to one where we try to understand
why the rank-size hierarchy of protohistoric settlement developed in the way it did.
URBANISATION
Definitions of the concept of ‘urbanization’(the process) in the archaeological and human geographic
literature abound. Many of these make no clear distinction with the related concept of ‘urbanism’(the
state). Since we are here concerned solely with the diachronic process of town formation (Formazione delle
città), we will use with minor modification the definition given by McIntosh (1991:208): urbanization is ‘a
process of regional transformation by which a rural landscape of undifferentiated villages and hamlets
with homogeneous populations transforms into a settlement network in which an agricultural hinterland
supports a few population agglomerations to which specialists are attracted.’
Note that this definition gives both quantitative (the number and size of agglomerations) and qualitative
(the presence of specialists) characteristics by which to recognize the process of urbanization; in my
discussion of methodology I will specify how these – and additional – characteristics are employed to
model the dynamics of the settlement history in the three study regions. The process of urbanization is
characterized by the growth of large(r) population centers, economic specialization, markets, and services;
the state of urbanism is characterized by the physical presence of a town, of its symbols (towers, walls,
temples), and of an urban hierarchy (plebs, skilled workers, and political/religious elites). The question of
whether something is a town must be therefore be decided by the application of physical and social
functional criteria (such as economic specialization, hierarchization, spatial differentiation, central market
and religious function). The term ‘proto-urban’has been widely used to designate nucleated settlements
during the early stages of the urbanization process, when it may be expected that a candidate
archaeological site does not meet all the criteria set out above. The related term ‘semi-urban’has also been
used to describe settlements that do not (clearly) meet a sufficient number of these criteria. Since such
pre-urban settlements may exhibit a whole range of divergent forms (McIntosh 1991), no single set of
characteristics can be given to define this concept with.
The use of the term ‘proto-urban’carries two potential dangers. Firstly, with hindsight it is tempting to
see processes of centralization and urbanization in proto- and early historic Italy as inevitable. There is
thus a potential teleological element to the use of the term ‘proto-urban’,suggesting that urbanization is a
natural progressive development for any society and that settlements that are not completely urban are in
some sense ‘not there yet’or even ‘failed’. We should instead use the term to indicate that no clear urban
character could be established. Secondly, the term can be used to ‘demote’native settlements which do
not display the characteristics of Roman or Hellenistic urban forms (Lomas 1996:142); it is therefore
important to guard against in-built Greek or Roman biases in the definition of what constitutes ‘urban’.
Of course urbanization is a form of centralization (both physical and social), and hence carries
implications for the level of organization of a society. Since in true towns social organization is no longer
exclusively based on kinship, early urbanization can only take place when societies are moving from a
segmented, tribal structure to a hierarchical, early state (McIntosh 1991). Pre-Greek urbanization in the
south, and pre-Roman urbanization in central Italy, would therefore indicate an indigenous development
toward early states in a manner at least partly independent of external forces. A priori it is unlikely, for
example, that (proto-) urban development in the Iron Age and the Archaic of south Lazio was restricted
to Rome itself, although it may well have been limited to those coastal areas which had sufficient external
contacts to generate wealth from trade.
2 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
COLONISATION
A widely used definition of colonialism by Prochaska (1990, quoted in Van Dommelen 1997:306) sees
“the presence of one or more groups of foreign people in a region at some distance from their place of
origin (the ‘colonizers’) and the existence of asymmetrical socio-economic relationships of domination or
exploitation between the colonizing groups and the inhabitants of the colonized region” as its two
fundamental characteristics. Note that the characteristic of asymmetry is here linked to a dominant role
for the colonizers. Both the presence of ethnic foreigners and the asymmetry of economic relationships
are very difficult to prove as long as all arguments are based on the material culture of the colonizer rather
than that of the native population. The dominant role of the colonizer, at least in the early stages of the
process, appears to be more of an assumption then an argument based in fact.
In contrast, the term ‘colonization’has been used to describe a number of disparate processes occurring
in different archaeological and historical contexts. When we speak of Greek and Roman ‘colonization’of
parts of central and southern Italy, we lump together processes ranging from the undirected ‘internal’
cultivation of previously uncultivated land to the plantation of colonists whose activities form an integral
part of the economic and military strategies of the state. The historical process of colonization, by Greeks,
Phoenicians, and Romans, has traditionally been a focus of interest and study within Mediterranean
archaeology. Since it began to be studied at a time (the mid-19th century) when western powers were
colonizing many other parts of the world, it was naturally seen from the colonizers’perspective as a
benign, civilizing process. For the same reason, the terms ‘colonialism’and ‘colonization’are often used
with the implicit connotation of asymmetric power relations between the colonizer and colonized, with
the connotation that the ‘native’can only act within a space determined by the colonial power (Rowlands
1998:329); but there is no reason to assume such asymmetry in many actual instances of Greek and
Roman colonization. Guzzo (1982), for example, argues that the Greek colonies of southern Italy were
established in coastal areas that were only marginally exploited by the indigenous populations, whose
settlement system had developed in the hilly hinterland. In the first years, contact with nearby indigenous
people would have been crucial for the development of the incipient colony, while increased trade with
and through the colony would cause the focus of the indigenous settlement pattern to shift gradually
toward the colony in later decades and centuries6.
Even if the asymmetric character of the colonial situation is without doubt, the indigenous party may
perceive it as providing opportunities rather than posing restrictions. In an article on aspects of
romanization in the hinterland of the Roman civitas capital at Wroxeter (Shropshire) for example, Roger
White and I proposed a detailed model for the relatively smooth changeover from the decentralized preRoman Iron Age tribal society of the Cornovii to the urbanized Roman civitas (Van Leusen & White
1997; chapter 3). In this instance, the colonizer opened up political and economic possibilities previously
inaccessible to a land-locked indigenous Cornovian society.
It appears likely that no single definition of either colonialism or colonization can be applicable across the
three study regions and the five centuries between 800 and 300 BC, and the nature of both must therefore
be defined separately for each instance. Within the context of the RPC project the goal has not been to
define Greek and Roman colonization as such, but rather to gauge the influence of these processes on
native development in the three Italian study regions. Were the Greek and Roman colonies and urban
developments dominant forces in this development (creating center-periphery relations), or did they
create equal polities integrated within indigenous society (creating peer polity interaction)? Herring, in his
study of socio-political change in the south Italian Iron Age and Classical period (1991), argues for the
existence of the latter up to, at least, the 6th century BC, while Whitehouse and Wilkins address the former
in their review of the archaeological evidence for Greek - native relations in south-east Italy (1989). In
general, it is easy to overemphasize the colonizers’perspective, because of the near exclusive role of late
classical authors in providing historical sources about the colonization process. Similarly, the highly
identifiable remains of classical cult may have led De Polignac (1994, 1995) to put too much stress on the
However, it has been pointed out (eg, by Carter 1993) that much of the archaeological proof for this kind of development is
very sensitive to the dating and interpretation of archaeological remains.
6
2 -
12
PATTERNS AND PROCESSES
use of ‘cult politics’ in claiming and securing colonial territory – more detailed investigation as at
Francavilla Marittima (Kleibrink 1997:69) has highlighted weaknesses in his argument.
If the strategies of ancient colonizers were, at times, less coherent than previously imaged, then
comparisons with early modern instances may be instructive. Attema (1999) suggests that a fruitful
comparison may be made between some instances of early Greek and Roman colonization, and the 18th
century colonization of Australia, where the presence of ‘debatable land’– that is, land on which no clear
ownership claims rested – allowed colonization by the English not because those lands were unused but
because the natives’and colonizers concepts of ownership differed too widely. A similar scenario may be
invoked to model early (7th century BC) Greek colonization in southern Italy; there, the colonies and their
initial agricultural base were established in a part of the landscape – alluvial deposits in the coastal plain –
considered of marginal importance by the indigenous populations whose societies focused on the low
hills and plateaus surrounding the plain (see chapter 12).
Another instructive comparison may be made with the early 16th and 17th century colonization of the
North American east coast by Europeans, where colonists depended upon trade with the indigenous
tribes for their survival. Native tribes had some decided advantages when it comes to living off the land:
“England’s attempts to establish colonies on the mainland in the sixteenth century failed completely. In
the early seventeenth century, the English succeeded only because neighbouring Indians assisted the
newcomers. The English colonies prospered by learning to grow such unfamiliar crops as corn and tobacco
and by developing extensive trading relations with Native Americans. … To achieve their first goal –
feeding themselves – they had to adopt agricultural techniques suited both to the new crops and to an
alien environment. Their second goal – maintaining lucrative trade networks – required them to deal
regularly on a more or less equal basis with people who seemed very different from them and who were far
more familiar with America than they were.”(Norton et al. 1991:2)
Obviously the historic instances of colonization by the Greek city states and by Rome did not take place
in an environment that was very different from what the colonizers were used to at home, nor were
communications with the homeland as difficult and time consuming, but one may still wonder (on the
one hand) how much help the colonists needed before they could fend for themselves, and (on the other)
how the relative freedom from the mother country might have afforded the colonists room for
experimentation with the organization of society (Rowlands 1998:330).
In other instances the process of colonization proceeded along very different lines. For example, the
Romans arrived in the Salento only after their military defeat of the Tarentines and their associates, while
in Lazio they progressively dominated a pre-existing urban structure and established new colonies in
strategic areas; a spontaneous process of urbanization during the Iron Age and Archaic became directed
urbanization under the Romans.
Whereas traditionally ‘colonization’has been a concept inextricably linked to the superior and civilizing
role of Greeks and Romans in the indigenous societies of Italy, research over the last decades has made
clear that the earliest phases of Greek and Roman presence in indigenous landscapes were characterized
by unobtrusive and small contingents of trader/settlers (Burgers 1999, Yntema 1999). There is no reason
to assume any kind of conflict existed between Greeks and natives of the Salento and Sibaritide at this
stage, and no reason to assume that their role in society became preponderant until the late 7th century.
But the Greeks arrived in a society which had already established a settlement system centering on large
defended villages in the foothills by the Late Bronze Age, and the Romans took over a landscape already
2 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
substantially settled by the native Archaic Latial tribes. If ‘colonization’is taken to signify the bringing
into productive use of a previously wild landscape, these natives were the ones who colonized Italy, and
who had established certainly by the late Iron Age a system of ‘proto-urban’top-ranking settlements in
places considered as marginal as the Salentine Murge.
The traditional view as, for example, formulated by Piero Guzzo and François de Polignac, denies this
native (‘Oenotrian’) dynamic of settlement and land use. In Guzzo’s view the Greek colonies, all founded
in coastal plains, led the Oenotrians to live in villages on the hilltops around, not the other way around
(Guzzo 1983:14-151). In De Polignac’s view the Greeks founded sanctuaries along the border of their
colonial territories to offer indigenous societies an opportunity to display wealth and influence, thereby
civilizing them (De Polignac 1984). Any ‘native’progression to urbanized life and organized cult in this
view only emerged under Greek domination and coaching. The archaeological evidence, however,
indicates that the indigenous society came under the domination of Sybaris only much later when, in the
6th century BC, the historically attested ‘imperium’of Sybaris became fact (e.g., Greco 1996:236).
The driving force behind both the native and the Greek colonization of the coastal plains of southern
Italy could well have been the same - demographic increase caused by the growing importance of
agricultural and arboricultural land uses since the Late Bronze Age and the restoration of supra-regional
contacts in the Early Iron Age, both of which led to a flowering of native elite-led redistributive societies
organized as chiefdoms (De Neef 1998:101-105; cf. Crielaard 1999).
The odious (because supremacist) terms of 'hellenization' and 'romanization' can, as we have seen, be
avoided in all this by the use of a neutral term such as 'acculturation'. While this begs the question of
what, precisely, was the relative contribution of both sides to this process, it has the advantage of lacking
an in-built pro-Greek or pro-Roman bias in the interpretation of the archaeological evidence.
3
TOWARDS INTERREGIO NAL COMPARISON
The aim of the RPC project is to compare the settlement dynamics of the Pontine region, the Salento
Isthmus, and the Sibaritide from late protohistory until the early Roman empire. I intend to approach this
goal via a circuitous route, first discussing the theoretical basis for making such comparisons, then
deriving a methodology from that, and lastly pointing forward to the case studies implementing them
(chapter 13).
3.1
WHY, HOW AND WHAT TO COM PARE ?
Before we begin our attempt to compare the settlement dynamics of the study regions, we need to
establish a theoretical and methodological basis for such a comparison, answering the questions of why
we think these regions can be compared at all; and, if they are comparable, of how to compare them.
Implicit in both these questions is the definition of comparanda, that is, which are the things we will be
comparing?
A starting point for answering the question of why the study regions should be comparable may be found
in similarities and differences in their respective physical landscapes. To begin with what all the regions
have in common, they all consist of coastal plains with a mountainous hinterland and transitional foothill
zones; each of them is a self-contained physical geographical unit of sorts. In a general sense the regions
are also similar because they are relatively close to each other – a well-known maxim in Geography says
that ‘everything is related to everything else but, all things being equal, closer things are more strongly
related to each other’.As to the differences between the regions, all three are situated on different coasts of
2 -
14
PATTERNS AND PROCESSES
the Italian peninsula, although the Salento and Sibaritide are very much closer together and the Pontine
region is both very near Rome, and beyond the farthest point reached by Greek colonists. Unlike the
relatively open geography of the Pontine region and the Salento Isthmus, the physical geography of the
Sibaritide (an isolated alluvial plain surrounded by foothills) may be compared to that of the plain of
Poseidonia (Paestum; De Neef 1998:101-105), neither of the latter two has any evidence for substantial
human activity in the plain until the 7th century BC.
One can also compare the regions by looking specifically at the cultural processes taking place in each of
them, and this is the approach taken within the context of the RPC project. During the 1st millennium BC
all three regions witnessed processes of settlement nucleation and urbanization (Formazione delle città), and
were subjected to one or more colonizing movements emanating from outside the region itself. On the
other hand, the Pontine region came under the early influence of Rome, and both were part of the same
Latin culture, whereas the early colonizers in the south were Greeks bringing a much more foreign
culture. Why did the Greek colonization of the Salento take a different route to that of the Sibaritide?
Were processes of centralization (urbanization) running similar courses in each region before they were
altered/truncated by outside factors? Culturally, the Sibaritide was also much more isolated from
neighboring areas than the Latial tribes in the Pontine region, and this might have affected the rate and
direction of cultural evolution within these two regions.
COMPARANDA
In reply to the question of what to compare and how, I take the phrase ‘dynamic settlement models’to
imply the modeling of processes rather than states; this is a strong definition, akin to that of simulation. A
less strict definition, more descriptive of current practice, is one whereby static models are placed within a
dynamic narrative.
It should be noted here that interregional comparison entails the explanation of differences between regions,
not the explanation of societal change itself. Thus, Bintliff’s (1997:17-33) models for regional
development provide explanations for what has been the outcome of the comparison between regions of
Greece. Can we indeed explain why things went differently in the longe duree / conjoncture of the three
study regions, without being forced to explain the fact of the occurrence of the processes themselves as
well? Perhaps what we should be explaining is the phenomena in terms of the underlying processes, e.g.
the hypothesis of craft specialization can explain certain archaeologically visible phenomena in terms of
an underlying process of centralization of society.
As the operation of the core processes in our study regions cannot be attested directly by archaeological
proof, centralization and urbanization might as well be seen in the way biological evolution was seen in
the past, i.e. as stages in a progressive development of societies (for example, Guidi writes about the
general evolution of chiefdoms into early states in the Alban hills in the central Tyrrhenian area from 1200
to 700 BC)? From a pragmatic point of view, processes encompassing all three study regions may be
argued to be outside the scope of this study, and need not be explained here either. This then leaves us
free to concentrate on the comparison, between the regions, of archaeological expressions of these
underlying processes.
ISSUE OF SCALE
The potential for any type of question, including geographical analysis, of archaeological data is
dependent on the scale at which that data was collected. At the regional scale, unless extensive surveying
has taken place, the bulk of the record will consist of point-like observations relating mostly (in Italy) to
easily observable structural remains of the Classical period. By their very nature such data present a
mainly classical Roman and Hellenistic landscape of towns, roads, necropoli, and villas, overlying a
protohistoric landscape of hillforts and tombe principesche. Such data are exemplified by the early volumes in
the Forma Italia series of publications; later volumes are much more detailed and include mapping of nonvisible zones (Cambi & Terrenato 1994:152).
2 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
3.2
EXPLANATORY MODE LS OF SOCIO-POLITICAL CHANGE
A combination of demographic growth and technological change may be assumed to be the ultimate
driving factor behind many of the processes of change in early Italy, but as we shall see explanatory
models of socio-political change tend to concentrate on more proximate factors. Among these,
quantitative/geographical models have gone out of favor since the 1970s, and have since been replaced
by sociological models closer to the humanistic outlook of most archaeologists. However I would like to
reiterate that, at sufficiently coarse geographical and temporal resolutions, physical parameters such as
geographical boundaries and the availability of natural resources may provide sufficient explanation for
the historical outcome of the processes we study (cf. Diamond 1998). Thus, biogeographical similarities
explain why crops, animals, and lifestyles could be communicated across the Mediterranean basin with
relative ease. Matthews (1999), in a volume dedicated to a discussion of models of late prehistoric and
Romano-British rural settlement in North West England, provides some useful pointers to nonarchaeological models, in particular to anthropological models of population density and forms of
organization (Kosse 1990, Bekker-Nielsen 1989) and Christaller’s geographical models of settlement
hierarchy (Collis 1986). He argues that such models have unjustly gone out of favor with the demise of
New Archaeology.
While archaeological explanations of socio-political change in classical Italy were originally driven by a
classicist perspective, the current literature, dominated by Italian prehistorians and Anglophones, is
almost exclusively expressive of the ‘revisionist’trend that emerged in the 1970s. The ‘new archaeology’is
seen by many (e.g., Trigger 1989:294-303) as the force that finally put native populations on a par with the
conquering military and administrative population, and authors such as Herring, Whitehouse and Wilkins,
and Burgers, accordingly reject the traditional ‘literary’Hellenophile approach. Recently indications are
mounting that the pendulum has reached its maximum: Jones (1997), in his review of the historiography
of Roman Imperialism, describes the rise of revisionism and its recent slowing down as ambiguities in the
archaeological record are recognized (see also Rowlands 1998:328).
In the archaeological literature about the structure of protohistoric and early historic societies of southern
Italy (and generally of Europe), two types of explanatory models were advanced in the mid-1980s to
replace the earlier ‘advance of classical civilization’model. The Peer Polity Interaction model, introduced
by Renfrew and Cherry (1986), sees change brought about by the interaction and competition between a
large number of independent and initially approximately equally matched polities; the Core-Periphery
model advanced at almost the same time by Rowlands (Rowlands et al. 1987), describes change as driven
by unequal interacting parts of single systems. In the two sections below, the specific application of these
two explanatory models to the south Italian evidence is given; however, it is not my intention here to
argue for or against either – rather, these models act as the backdrop against which testable hypotheses
about the patterning of the archaeological record in our three study regions may be developed. The value
of explanatory models, as is particularly clear from the discussion of the Core-Periphery model by
Whitehouse and Wilkins, is in how well they are able to predict the occurrence of certain archaeological
evidence; my purpose here is to derive specific testable hypotheses about the spatial scale and distribution
of archaeological features. The tests themselves are then conducted by transforming these hypotheses
into a form conducive to GIS analysis.
PEER POLITY INTERACTION
Herring (1991:35-6, 42-9), discussing peer polity interaction in the south Italian Iron Age and Classical
period, aimed ‘to show how communications (on all levels) between the different communities could
have been a major dynamic to socio-political change’.He argues that the Greek colonies were themselves
tribal societies and therefore peer polities to the native tribes, and attempts to fit the available
archaeological evidence into Renfrew’s (six) characteristics of peer polity interaction. Herring concludes
that the peer polity model ‘works well between the late 8th century to sometime in the 6th century’, while
later on one can see two systems of peer polity interaction (one indigenous, one colonial Greek) existing
in southern Italy at the same time. The change-over being brought about by ‘exogenous change’. Herring
struggles to avoid lapsing into the old hellenization idea when arguing that in the later period, while
2 -
16
PATTERNS AND PROCESSES
new ideas and products came from Greece through the Greek colonists to the natives, this was an
exchange between equals rather than one between a dominant and a subservient partner.
Peer polity models for the organization of pre- and protohistoric societies have enjoyed a growing
popularity with students of Bronze Age and Iron Age hillforts, especially those working with GIS,
because the underlying assumption of equality between hillforts belonging to the same system allows the
application of a number of spatial analytical techniques. Hillforts formed the top of the settlement
hierarchy in the tribal societies of the Bronze Age and early Iron Age in large parts of Europe, but very
little is known about the living systems they were part of. Since many of them were only investigated from
a topographic point of view, and dates are in many cases only available where diagnostic surface finds
could be made, assumptions must be made about their contemporaneity in order to be able to treat them
as the foci of interacting peer polities. The polities themselves might best be viewed as tribal subdivisions
or ‘cantons’with a population ranging from a few hundred to a few thousand, as in the case of the tribal
area of the Cornovii in north-western Britain (Van Leusen & White 1997).
CORE – PERIPHERY INTERACTION
Since the original volume edited by Rowlands and others, the core-periphery model has been applied to
many comparative studies by prehistorians (see, for example, the volume edited by Champion 1995).
Whitehouse and Wilkins (1991) employ its theoretical framework to study the south Italian archaeological
evidence, discussing in turn the archaeological evidence, the Greek cities, the relationship between Greek
and natives, and changes in the native communities. These authors conclude (1991:123) “We have tried to
demonstrate that an analysis of the development in south-east Italy in terms of a center-periphery model
can give important insights into the precise forms of social and economic relations through which the
native communities were brought into contact with the Greeks, and we have highlighted the specific
importance of prestige goods for the transformation of native economies and social organization.”
However, as noted in a review by Yntema (1996), there is an in-built assumption that the native
hinterland must perforce be the periphery to the Greek coastal colonial centers; the reverse is not being
investigated. Whitehouse and Wilkins (op. cit.: 107-115) tabulate the expected differences in evidence
between the rival hypotheses of Greek control and co-existence, but their discussion of the artefactual
evidence is marred by ignoring the strong biases that are present. They come down clearly in favor of the
co-existence model, and explain the occurrence of predominantly ‘Greek’sanctuaries in the native area by
reference to a prestige-goods system, in which the sanctuaries functioned as emporia for the trading of
native products (wool) for Greek prestige objects.
When discussing direct Greek control of Italian territory in the Archaic period, Whitehouse and Wilkins
want to emphasize the small amount of land held directly by the colonies – only 15 by 15 km in the case
of Metapontion, and perhaps four times as much for Taras. However, this just about uses up all of the
available coastal plain and raises the interesting question of just what might have been the role these
plains could have played in the native economies – a question that is also relevant to the Sibaritide in the
Archaic period! The native settlements outside this area were perhaps ‘allowed to continue to flourish’,
but if so they would have to do it on a different economic basis from that of the colonies themselves…
MODEL AND REALITY
Both Peer Polity and Core-Periphery explanatory models for dynamic interaction predict specific regional
and supra-regional patterns of archaeological correlates, but only for idealized homogeneous spaces and
societies. It is therefore doubtful that they can be tested through landscape archaeological fieldwork.
Similarly, static regional locational models such as gravity or central place models apply only to certain
types of interaction within hierarchical settlement patterns. For example, a successful application of
central place models is possible only in the study of retail distribution within physiographically
homogeneous regions such as river valleys (Crumley 1979:152). Such models are not applicable to
dynamic and open societies, which have a heterarchical organization.
2 -
17
V A N L EUSEN : P A T T E R N T O P ROCESS
3.3
DETECTING MACRO -ARCHAEOLOGICAL QUANT ITATIVE PATTERNS
Human action is structured in space and time; hence the archaeological record is patterned in space and
time. Unfortunately the processes by which the archaeological record is preserved, recovered and
described (constructed) are themselves also non-random; hence the result is a second set of patterns
superimposed on those within the archaeological record itself. The general theoretical approach followed
in this thesis is that of the French Annales school as applied in archaeology (Bintliff 1991, Knapp 1992);
this is most suited to deal with the nature (mostly undiagnostic surface material) and temporal scale of the
material studied (from about 1400 BC to about AD 300). The case studies involving field work and GIS
analysis are all based on the theoretical premises of landscape archaeology – that human actions may
occur, and leave an essentially continuous ‘blanket’of traces, anywhere in the landscape, that the resulting
surface record is a palimpsest of such traces through time, and that patterns in this record may be
explained in part by the in turn limiting and enabling qualities of the landscape. These principles are
extended into the realm of spatial extrapolation and cultural resource management using the theory of
spatial sampling (which says that properties of a properly selected sample have a specific likelihood of
also being properties of the parent population).
A model for interregional quantitative pattern analysis is provided by Bintliff’s (1997) comparison of
diachronic site counts from surveys in Greece. He shows that more intensive surveys yield larger numbers
of Geometric and Archaic sites. It is therefore possible that part of the geographical shift in the period of
demographic take-off is due to the fact that areas where, at best, only extensive surveys took place will
show a sudden increase in site numbers for the period in which the first easily recognizable material is
present – that is, the classical/Hellenistic period. By the same reasoning, inasmuch as the more intensive
surveys have concentrated on the heartland of classical Greece, these areas will show the clearest evidence
of protohistoric settlement activity. Bintliff’s hypothesis of shifting demographic take-off and conclusion
(1997:14) that, ‘by and large, the “evidence on the ground” is broadly comparable to political history’is
therefore, I feel, not well supported by the evidence he adduces and must be strengthened by a more
detailed study of potential biases. From his very brief statement of method (1997:2) it is not clear how
successful he was in ‘unbiasing’his data.
SAMPLING IN ARCHAEOLOGICAL SURVEYS
Approaches to pattern detection in archaeological data are contingent on the methods used to collect
those data. The study of questions relating to how to collect sample data in order to allow generalizations
about a sampled population with a specific degree of certainty is the domain of sampling theory. Once data
have been collected, other approaches are needed to determine whether patterns may be present. A very
popular approach, especially with the advent of GIS, has been to measure the correlation between the
sample data and explanatory variables, but this has generally been tied up with the discrete, site/non-site
approach to the archaeological record (for a further discussion of these methods, see chapter 4).
Geostatistics and signal processing theory are two alternative approaches to consider when studying
continuous survey data.
The type of sample taken during archaeological surveys depends on the type and scale of patterning one
expects in the archaeological record, but is restricted by available time and funding. Thus catchment
studies may sample nearly all of the immediate surroundings of some central place or village in order to
establish the existence of a microregional pattern, or they may sample a small part of a physiographic
region such as a watershed basin in order to establish regional patterns of settlement and land use.
SOPHISTICATION IN SPATIAL ANALYSIS
One of the most popular approaches in archaeological pattern detection and explanation since the 1960s
has been to demonstrate a correlation between two or more variables - for example, the presence of farm
sites and the agricultural productivity of the soil. The bulk of GIS work in the 1990 has concerned exactly
such work. Unfortunately, the analysis of spatial (geographical) data has always been complicated by the
fact that real-world spatial variables exhibit a large amount of autocorrelation (near things are more
2 -
18
PATTERNS AND PROCESSES
similar), and variables are also spatially correlated. A measure of spatial correlation between two variables
has been developed from the formula for autocorrelation. This issue is dealt with in more detail in chapter
5.
Geostatistics are a body of theory and methods designed for the analysis of spatially correlated,
geographical variables. Despite the reservations expressed by Barceló and Pallarés (in their discussion of
the theory and method of social space, 1998: 65) that geostatistical methods do not perfectly fit
archaeological purposes because social action and, with it, social space is discrete rather than continuous,
I believe that the construction of geopedological units on the basis of point measurements (corings) and
areal observations (geomorphological units) is sufficiently similar to the construction of meaningful
archaeological entities (e.g., site catchment areas and urban manuring zones) on the basis of excavations
and surveys to warrant a further exploration of the potential of geostatistics including such spatial
extrapolation techniques such as Kriging. Any underlying assumptions (e.g., the normal distribution of the
variables) should of course be born in mind when applying geostatistical methods to archaeological data.
BIAS MODELING
The theory behind bias modeling is that the factors and processes causing bias can be measured and
modeled, thus taken into account when analyzing and interpreting survey data. This line of research is the
subject of chapter 5 of this thesis.
SIGNAL PROCESSING
Recording archaeological surface material as one or more gridded surfaces of counts, weights, or densities
of object or material types results in a data type – gridded continuous variables – that is, as we have seen,
conducive to storage and processing in a GIS. However, it also suggests that principles and techniques
from the field of signal and image processing may be applicable as well. If, as many authors argue, we
should no longer employ the concept of ‘site’at the heart of our survey designs, field procedures, and
analysis, we must identify an alternative concept.
If, as is argued by most landscape archaeologists, we can regard the surface archaeological record as an
essentially continuous variable, though varying in density, then perhaps we can escape from some of the
confines and confusions of established archaeological terminology (‘site’, off-site, non-site, etc) by
temporarily adopting the vocabulary and some of the techniques of a related field – that of signal
processing. In order to do this effectively, I would like to argue first that we should view the surface
record not as one single signal or variable, but as a potentially large number of overlapping signals, each
representing the land use intensity of a particular period. This is in agreement with the generally accepted
view of the surface record as a palimpsest of past uses. Each of these signals in itself may be continuous.
Rather than predefining the nature of these signals, for example as representing archaeological activity
types (‘subsistence farming’would be one example), we can adopt more pragmatic methods that allow us
to analyze the signals without establishing their nature first.
If the surface record can be viewed as a series of overlapping signals of varying frequency and amplitude,
then we may be able to use frequency filtering techniques such as Fourier analysis in our attempt to define
and separate these signals. And we may be able to employ non-archaeological definitions of what
constitutes a signal (and therefore of what constitutes noise) to aid in the quantitative processing of data
generated by surface survey. These two possibilities are discussed below. A number of fairly strong
assumptions about the type of patterning present in the archaeological record underlie the following
discussion, and I will refer to these where-ever necessary.
Frequency filtering
Although the surface record is an essentially two-dimensional data set, assumptions of isotropy (see
section 3.1) allow us to consider a one-dimensional ‘slice’through this surface first. The total density of
the record is the signal, and it is made up out of a number of patterned human activities. We postulate
that these are distributed regularly in space (frequency) and in density (amplitude); for example, an
2 -
19
V A N L EUSEN : P A T T E R N T O P ROCESS
area settled by small farmers in the Roman Republican period will, other things being equal, be
represented in survey data by a regularly spaced occurrence of ceramic scatters of some typical density.
Each of these scatters in turn has a typically bell-shaped density distribution. If a data set consists of a
number of signals of differing frequency and/or amplitude, standard signal processing techniques can be
used to separate these signals.
Fourier analysis of discrete two-dimensional data is a well-known technique in image processing,
discussed for example in the handbook by Lillesand & Kiefer (1994:563-566). The data are
mathematically described by a combination of sine and cosine functions transformed into two arrays
containing the real and imaginary components of its frequency space, which can then be masked so as
remove unwanted frequencies or select frequencies of interest7. Its uses within archaeology have so far
been limited mainly to noise removal or reduction in aerial and satellite imaging, and its potential for the
analysis of archaeological survey data has, as far as I have been able to establish, not been investigated. In
theory, GIS software containing tools for Fourier analysis can be used to split up the results of a survey
into a series of maps, each representing a different frequency8. The criteria for this splitting can be purely
empirical, that is, based on signal strength only, or they can be derived from field measurements of
observed frequencies such as those of plough furrows and ridge-and furrow, or they can be derived
theoretically. For example, if an archaeological landscape is hypothesized to consist of modular units
centered on a single family farmstead, the theoretical diameter of the modules (e.g., 400 m) may be used
as the frequency of interest. In a further step, a frequency or set of frequencies may be subtracted from
the original data, leaving only the residuals for further modeling and analysis.
Signal vs. Noise
Much of the debate has centered on the question of ‘signal vs. noise’or, in archaeological terms, site vs.
off-site. Although it is now generally agreed that all surface finds can contribute to our knowledge of the
archaeology of an area (and in that sense there is no noise9), the use of statistical noise levels can help to
distinguish meaningful variation from meaningless variation in the surface record. For example, a ‘site’
might be defined as any area that has a significantly higher surface sherd density than its surroundings
(e.g., Gallant 1986), with ‘significant’indicating that a certain level of random variation will be present in
the data and should not lead to the definition of a site. Similarly, the definition of what constitutes a
‘manuring spread’will have to include some way of dealing with statistical noise levels (Wilkinson 1982).
The concept of statistical variance must be introduced here, because it provides a means of specifying the
probability that data constitute ‘noise’rather than a ‘signal’.
These issues can best be illustrated by examining Bintliff’s recent reanalysis of the results of the 1980s
survey campaigns conducted by him and Snodgrass around the Greek city of Thespiae (Boeotia), in
combination with the results of the RPC project’s recent surveys (Bintliff & Howard 1999). The Boeotia
project team is now looking for low-density patterns that may have been hidden by the general late
Archaic/early Hellenistic ‘blanket’of ceramics resulting from intensive manuring practices during the
demographic heyday of the city. Among their early results are the discovery of a prehistoric landscape
consisting of Early/Middle Bronze Age small short-lived farmsteads, and the realization that localized
areas of low ceramic density, if combined with finds of certain pottery types, indicated the presence of
cemeteries; of especial interest here is their use of numeric correction procedures to re-assess the
presence of Greek and Roman sites in the light of expected ‘off-site’finds densities in the chora of Thespiae.
The approach taken by Bintliff and Howard is to use GIS to model the distribution of identifiable off-site
material (the manuring scatter emanating from Thespiae itself, and the scatters surrounding rural farm
sites) using a cost surface based on terrain slope and distance to historically known trackways. They then
subtract the ‘expected’manuring finds density from the recorded finds density in order to interpret the
7
Low frequencies are in the centre and high frequencies near the edges of these arrays.
8
GRASS 4.1 uses the fast forward and inverse Fourier transforms.
9
A position taken, for example, by Kuna (2000:42).
2 -
20
PATTERNS AND PROCESSES
residuals. However, the recorded finds densities result from material collected during uncontrolled
gridded site surveys (i.e., neither the type nor the number of finds collected is necessarily representative of
the area surveyed) and it is therefore unclear that the residuals reflect actual variation in the density of
archaeological materials on the surface. Furthermore, Bintliff and Howard’s assumption (1999:60) that
manuring will, over time, result in a smooth blanket of ceramics seems untenable; rather, there will be
peaks and troughs (noise) that are the accidental by-product of the historical manuring process. The
expected pattern of manuring may also be different from the one modeled by Bintliff and Howard,
leading to a different pattern of residuals; for example, might there be an inverse relation between soil
fertility and manuring practice? For such methods to become generally accepted, we must develop better
ways of calculating expected off-site densities and their residuals, perhaps using standard deviations and
confidence levels rather than absolute densities. And our trust in the correctness of the residuals and
‘noise’levels should be expressed by establishing formal statistical levels of confidence.
If we compare this to the RPC project approach to low finds density, two comments may be made.
Firstly, the re-survey in 1999 of parts of the 1998 Fogliano survey area has demonstrated that very low
finds densities cannot be disregarded as ‘noise’or ‘off-site’– they may well be related to low visibility
conditions and that revisits/excavations have the potential to substantiate the ‘one sherd can indicate a
site’idea. Low finds densities therefore have to be problematized. On the other hand, it should be
recognized that even intensive ‘general’surveys cannot hope to collect representative samples of low
density, low visibility categories such as the prehistoric impasto in the Sibaritide 2000 survey (chapter 12);
this will need a targeted re-survey. Finally, the recording of low densities can also be an artefact of the
classification process; some periods can only be recognized if diagnostic forms or decorations are present.
If these are rare or absent, finds will be classified into broad undiagnostic categories, or even as
‘indeterminate’.
Secondly, our interpretation of the protohistoric off-site ceramic ‘carpet’during the Ostuni survey is
different from Bintliff’s interpretation of the carpet of classical material around Thespiae. The
protohistoric (Middle Bronze Age) impasto carpet of OST99 is explained by assuming a relatively short
period (150 years) of shifting cultivation which transformed a large percentage of the area into a ‘site’; in
contrast, the Classical period carpet around Thespiae is explained as the result of a relatively brief period
of manuring from the town.
4
DISCUSSION
The brief review of current archaeological opinion regarding the settlement and land use history of the
three Italian regions, presented in section 2.1 above, immediately tempts us to make comparisons and
define similarities and differences. However, the question arises of whether these similarities and
differences are properties of the regional archaeological records themselves, or of the explanatory
frameworks being used. Since this allegation has been made repeatedly of the traditional classicist Greekand Roman-centered narrative, it is incumbent on the current generation of researchers to demonstrate
that they have not just replaced this with a 'native-centered’explanatory framework. The following should
therefore be regarded as a preliminary comparison only:
• In both central and southern Italy, a nucleation of permanent settlement out of a previous
(presumed) system of pastoralism and shifting cultivation becomes archaeologically apparent by the
Late Bronze Age. Whilst the same set of causative factors has been adduced to explain this
phenomenon in both areas, I have argued above that this may be caused by our use of ‘catch-all’
explanations when we have a poor understanding of the true causes.
• Both the Greeks in the Late Iron Age Salento and Sibaritide, and the Romans in Archaic southern
Latium, encountered a developed indigenous tribal society which had already colonized all but the
most marginal landscape units and who lived in a range of settlements from isolated farms to ‘proto-
2 -
21
V A N L EUSEN : P A T T E R N T O P ROCESS
urban’settlements. In both cases the process of acculturation took some 150 years before the native
culture was fully submerged into the new supra-regional material culture of the Hellenistic and
Roman Republican periods.
It would be better if we could make comparison of quantitative aspects of the regional archaeological
records, and here the example of Alcock’s (1993) and Bintliff’s (1997) work on Classical to Roman
Greece provides a valuable point of departure. However, as noted in section 3.3 above, this approach
must be refined in order to allay doubts about their methodological justification. The greatest potential
for resolving these doubts lies in limiting the geographical scope of our comparison to an infra- or even
micro-regional level; this potential is further explored in chapter 13.
In the end, one must agree with Bintliff’s (2000:214) opinion that
Field survey is an incomplete guide to regional settlement systems, but it is an illusion to suppose that
excavation or historical source control is a firmer basis – these approaches are probably even more
inadequate for regional settlement reconstruction than large-scale intensive survey. In combination however
I believe that these three approaches can create Piggott’s ‘cumulative credibility’; many of the more
intractable problems of settlement and population reconstruction and interpretation may be assisted
considerably through a dialectic in the field involving information from all three sources of regional
information.
This sentiment is supported by Millet’s (2000) convincing argument that surface (surveyed) and stratified
(excavated) assemblages are complementary in content, and therefore have a complementary role in
archaeological explanation.
REFERENCES
Arancio, ML, VC Buffa, I Damiani & F Trucco 1994
Torre Mordillo, in Atti Taranto 1992: 145-163. Napoli.
Alcock, SE 1993
Graecia Capta: the landscapes of Roman Greece. Cambridge: Cambridge UP.
Attema, PAJ 1993
An Archaeological Survey in the Pontine Region. A contribution to the early settlement history of south
Lazio 900 – 100 BC. Groningen: Rijksuniversiteit Groningen.
Attema, P 1999
Koloniale geografieën in de Mediterrane wereld, TMA 21 (1999):27-37.
Attema, P, G-J Burgers & M van Leusen forthcoming
The Ostuni Field Survey Campaign 1999. Studi di Antichitá 11 (2002).
Attema, P, G-J Burgers, M Kleibrink & D Yntema 1998
Centralization, early urbanization and colonization in a regional context, dutch excavations and landscape
archaeology in central and southern Italy, Saguntum (PLAV) 31:125-132.
Attema, P, E van Joolen, & M van Leusen 2001
A Marginal Landscape: Field work on the beach ridge complex near Fogliano (South Lazio), PaleoHistoria
40/41 (1998/1999):149-162.
Attema, PAJ & PM van Leusen 2000
Kern en periferie in het RPC-project (1); de Doganella di Ninfa-survey in de Pontijnse regio (MiddenItalië), in PaleoAktueel 10: 25-30.
Barceló, JA & M Pallarés 1998
Beyond GIS: The archaeology of social spaces, in Archeologia a Calcolatori 9:47-80.
Barker, G 1981
Landscape and Society: Prehistoric Central Italy. London etc: Academic Press.
Bekker-Nielsen, T 1989
The geography of power: studies in the urbanization of Roman north-west Europe (BAR S477).
Bintliff, J (ed) 1991
The Annales School and Archaeology. Leicester, London: Leicester UP.
2 -
22
PATTERNS AND PROCESSES
Bintliff, J 1997
Regional Survey, Demography, and the Rise of Complex Societies in the Ancient Aegean: Core-Periphery,
Neo-Malthusian, and Other Interpretive Models, Journal of Field Archaeology 24:1-38.
Bintliff, JL 2000
The concepts of ‘site’ and ‘offsite’ archaeology in surface artefact survey, in Pasquinucci, M & F Trément
(eds), Non-Destructive Techniques Applied to Landscape Archaeology (The Archaeology of Mediterranean
Landscapes 4): 200-215.
Bintliff, J & Ph Howard 1999
Studying Needles in Haystacks – Surface Survey and the Rural Landscape of Central Greece in Roman
Times, Pharos 7:51-91.
Boardman, J 1980
The Greeks Overseas: Their early colonies and trade. 3rd revised edition, London: Thames and Hudson.
Bondanelli, P (ed) 1984
Nicolò Macchiavelli: The Prince. The World’s Classics, Oxford: Oxford University Press.
Burgers, G-JLM 1998
Constructing Messapian Landscapes (Dutch Monographs on Ancient History and Archaeology 18).
Amsterdam: Gieben.
Burgers, G-J 1999
Antieke koloniale situaties in Zuid-Italië: een regionaal perspectief, TMA 21(1999):19-28.
Cambi F & N Terrenato 1994
Introduzione all’archeologia dei paesaggi. Studi Superiori 203. Roma: La Nuova Italia Scientifica.
Cancellieri, M 1990
Il territorio pontino e la via Appia, Archeologia Laziale 10 (1) (QuadAEI 18:61-72).
Carter, JC 1993
Taking possession of the land. Early Greek colonization in Southern Italy, in Eius virtutis studiosi.
Classical and postclassical studies in memory of FE Brown: 342-367. Washington/Hanover: University
Press of New England.
Champion, TC (ed) 1989
Center and Periphery; comparative studies in archaeology. London: Unwin Hyman.
Collis, J 1986
Central place theory is dead: long live the central place, in E Grant (ed), Central places, archaeology and
history:37-9. University of Sheffield Department of Prehistory and Archaeology.
Crielaard, JP 1999
Eerzucht en krijgshaftigheid als discours voor vroeg Grieks kolonialisme, TMA 21(1999): 12-18.
Crumley, CL 1979
Three locational models: an epistemological assessment for anthropology and archaeology, in ???: 141-173.
D’Andria, F 1998
Ricerchi recenti sugli insediamenti indigeni di Puglia e Basilicata, in La Forma della Cittá e del Territorio,
Esperienze metodologiche e risultati a confronto. Atti dell’Incontro di studio – S. Maria Capua Vetere 2728 novembre 1998, pp 103-118. Napels: Seconda Universitá degli Studi di Napoli.
De Neef, W 1998
Patronen van Verandering. Landschapsgebruik en menselijke activiteit in de Metapontino (Basilicata) en
de Vlakte van Paestum (Campanie) in de Ijzertijd en de Griekse kolonisatieperiode (1000-400 v. Chr.).
Unpublished M.A. thesis, University of Groningen.
De Polignac, F 1984
La naissance de la cité grecque. Paris.
De Polignac, F 1994
Mediation, competition, and sovereignty: the evolution of rural sanctuaries in Geometric Greece, in Alcock,
S & R Osborne (eds), Placing the Gods. Oxford: Clarendon Press.
De Polignac, F 1995
Cults, Territories, and the Origin of the Greek City State. Chicago/London: University of Chicago Press.
Diamond, J 1998
Guns, Germs, and Steel. London, etc: Vintage.
Gallant, TW 1986
‘Background noise’and site definition: a contribution to survey methodology, JFA 13:403-418.
Greco, E 1996
La città e il territorio, in Pugliese Caratelli (ed) 1996: 233-242.
Guzzo, PG 1982
2 -
23
V A N L EUSEN : P A T T E R N T O P ROCESS
Modificazioni dell’ambiente e della cultura tra VIII e VII secolo sulla costa ionica d’Italia, in Dialoghi
d’Archeologia 2 (1982): 146-151.
Guzzo, PG 1983
La Sibaritide e Sibari nell’VIII e nel VII sec. a.C, Atti del Convegno Internazionale, Grecia, Italia e Sicilia
nell’VIII e VII sec. a C. , Atene (5-20 ott 1979), Annuario della Scuola Archeologica di Atene e Missioni
italiane in oriente 60 n.s. 44 (1982): 237-251.
Herring, E 1991
Socio-political change in the south-Italian Iron Age and Classical periods: an application of the peer polity
interaction model. Accordia Research Papers 2 (1991):33-54.
Johnson 1998
Jones, GDB 1997
From Brittunculi to Wounded Knee: a study in the development of ideas, in DJ Mattingly (ed) 1997:185200.
Kleibrink, M 1997
Dark Age or Ferro I? A tentative answer for the Sibaritide and Metapontine plains, in Caeculus III(19961997):63-90. Groningen: Institute of Archaeology.
Kleibrink, M 2000
Early Cults in the Athenaion at Francavilla Marittima as Evidence for a Pre-colonial Circulation of nostoi
Stories, in: V. Gassner, et al. (eds), Die Ägäis und das westliche Mittelmeer, Beziehungen und
Wechselwirkungen 8. bis 5. Jh. V. Chr.:165-184. Wien: Verlag der Österreichischen Akademie der
Wissenschaften.
Kleibrink, M forthcoming
The sacred landscape of the Sibaritide – the veneration of ancestors, nymphs and deities, in Attema, PAJ,
G-J Burgers, E van Joolen, PM van Leusen & B Mater (eds), Regional Pathways to Complexity.
Proceedings of the RPC conference, Groningen 2000 (BAR Int Ser).
Lillesand, TM & RW Kiefer 1994
Remote Sensing and Image Interpretation. 3rd Edition, New York etc: John Wiley & Sons.
Maaskant Kleibrink, M 1993
Religious Activities on the "Timpone della Motta", Francavilla Marittima, and the identification of Lagaria,
BaBesch 68:1-49.
Kleibrink, M & M Sangineto 1999
Insediamento enotrio su Timpone della Motta I, la ceramica geometrica dal edificio V, Francavilla
Marittima, BaBesch 74:1-61.
Knapp, AB (ed) 1992
Archaeology, Annales, and Ethnohistory. Cambridge: Cambridge UP.
Kosse, K 1990
Group size and societal complexity: thresholds in the long-term memory. Journal of Anthropology and
Archaeology 9, 275-303.
Laurence, R 1994
Roman Pompeii: Space and Society. London: Routledge.
Lomas, K 1993
The city in south-east Italy. Ancient topography and the evolution of urban settlement, Accordia Research
Papers 4:63-77.
Lomas, K 1996
Greeks, Romans, and Others: problems of colonialism and ethnicity in southern Italy, in Webster, J & N
Cooper (eds), Roman Imperialism: Post-colonial perspectives. Leicester Archaeology Monographs 3: 135144.
Matthews, KJ 1999
Rural settlement in Roman Cheshire: a theoretical view. In M Newell (ed), Living on the Edge of Empire
(Archaeology North West 3(13) (1998)): 27-34.
Mattingly, DJ (ed) 1997
Dialogues in Roman Imperialism. Power, discourse, and discrepant experience in the Roman Empire.
Journal of Roman Archaeology Suppl. Series, vol 23.
McIntosh, RJ 1991
Early urban clusters in China and Africa: the arbitration of social ambiguity, in Journal of Field
Archaeology 18:199-212.
Millett, M 2000
The comparison of surface and stratified artefact assemblages, in Pasquinucci, M & F Trément (eds), Non-
2 -
24
PATTERNS AND PROCESSES
Destructive Techniques Applied to Landscape Archaeology (The Archaeology of Mediterranean
Landscapes 4):216-222.
Norton, MB, DM Katzman, PD Escott et al. 1991
A People and a Nation, a history of the United States. 3rd Brief edition, Houghton Mifflin: Boston etc.
Perkins, Ph 1999
Reconstructing the Population History of the Albegna Valley and Ager Cosanus, Tuscany, in Gillings, M,
D Mattingly & J van Dalen (eds), Geographical Information Systems and Landscape Archaeology (The
Archaeology of Mediterranean Landscapes 3):103-115. Oxford: Oxbow books.
Peroni, R 1994
Introduzione alla protostoria italiana. Roma/Bari: Editori Laterza.
Pugliese Caratelli, G (ed) 1996
Magna Grecia. Milano: Electa.
Puglisi, S 1959
La civiltà appenninica. origine delle comunità pastorali in Italia, Firenze
Renfrew, C & JF Cherry (eds) 1986
Peer Polity Interaction and Sociopolitical Change. Cambridge: Cambridge UP.
Rowlands, M 1998
The archaeology of colonialism, in Kristiansen, K & M Rowlands (eds), Social transformations in
archaeology: global and local perspectives: 327-333. London/New York: Routledge.
Rowlands, M, M Larsen & K Kristiansen (eds) 1987
Center and Periphery in the Ancient World. Cambridge: Cambridge UP.
Semeraro, G 1997
En neusí. Ceramica greca e societá nel Salento arcaico. Lecce-Bari.
Trigger, BG 1989
A History of Archaeological Thought. Cambridge: Cambridge UP.
Van Dommelen, P 1998
On Colonial Grounds. Archaeological Studies Leiden. Leiden: University of Leiden.
Van Dommelen, P 1997
Colonial constructs: colonialism and archaeology in the Mediterranean. World Archaeology 28:305-23.
Van Leusen, PM 1998
Archaic Settlement and Early Roman Colonization of the Lepine Foothills, Assemblage 4,
http://www.shef.ac.uk/~assem/4/.
Van Leusen, PM & PAJ Attema 2000
Kern en periferie in het RPC-project (2); de Fogliano-survey in de Pontijnse regio (Midden-Italië),
PaleoAktueel 10: 31-35.
Van Leusen, PM & RH White 1997
Aspects of Romanization in the Wroxeter Hinterland, in K Meadows, C Lemke & J Heron (eds), TRAC 96.
Proceedings of the Sixth Annual Theoretical Roman Archaeology Conference, Sheffield 1996:133-43.
Oxford: Oxbow Books.
Whitehouse, RD & JB Wilkins 1989
Greeks and natives in south-east Italy: approaches to the archaeological evidence, in Champion, TC (ed),
Center and Periphery; comparative studies in archaeology:102-126.
Yntema, DG 1990
The Matt-Painted Pottery of Southern Italy. 2nd Enlarged and corrected edition. Galatina.
Yntema, DG 1993
In Search of an Ancient Countryside. The Free University Field Survey at Oria, province of Brindisi, Soth
Italy (1981-1983). Amsterdam.
Yntema, DG 1996
Review of Champion, TC (ed) 1995, Center and Periphery; comparative studies in archaeology (One World
Archaeology 11), in Tijdschrift voor Mediterrane Archeologie 17:65-6.
Yntema, DG 1999
Kolonisten en migranten, TMA 21(1999): 50-6.
2 -
25
CHAPTER
3
INTRODUCTION TO THE
WROXETER HINTERLAND PROJECT
This chapter consists of two previously published articles which, together, provide the backdrop against
which the case studies on land use / land cover bias (chapter 14) and visibility and friction (chapter 16)
should be read.
The first article, ‘Extending GIS Methods for Regional Archaeology: the Wroxeter Hinterland Project’,
co-written with Vincent Gaffney, was originally presented to the 1995 Meeting of CAA (in Leiden, The
Netherlands), and was published the following year in its proceedings (Kamermans and Fennema 1996).
The article introduces the project itself, its main research goal of explaining the anomalous existence of
Wroxeter itself through a study of both the town and its hinterland, the models for urbanisation and
Romanisation that were to be tested with the project data, and especially to detail the geographical
approaches that were to be developed and applied for this study.
The second article, ‘Aspects of Romanization in the Wroxeter Hinterland’, co-written with Roger White,
was originally presented at the 6th annual Theoretical Roman Archaeology Conference (TRAC, Sheffield
1996) and appeared in its proceedings the following year (Meadows et al. 1997). It provides a preliminary
analysis of the evidence and arguments explaining the paradoxical existence of a large and thriving
Romano-British town in the middle of a rural landscape almost devoid of any evidence for indigenous
wealth, centralisation, or significant acculturation to Roman lifeways. Wroxeter is the only large Roman
town in Britain which does not appear to have a substantially Romanised hinterland. Previous
hypothetical explanations for this phenomenon have included a hostile stance of the local population, an
over-ambitious civilian foundation of Wroxeter, and /or its economic underdevelopment. The second
idea could be laid to rest given the results from the aerial photographic and geophysical study of the town
itself (Van Leusen 1999b, Gaffney & Gaffney 2000); we argue here that the first and third idea can now
be discounted as well because the archaeologically attested success of the town implies that the hinterland
must have been wealthy. We conclude that the paradox is mainly due to the near invisibility of the forms
of wealth which the native Cornovians would have commanded and to the lack of a systematic study of
late pre-Roman Iron Age and Roman patterns of settlement and land use in the Wroxeter hinterland.
The congruence in both the core processes being studied (centralisation, urbanisation, and Romanisation),
the general theoretical stance (landscape archaeology), and the emphasis on the use of methods of
geographical analysis, suggests that a comparison of the results of the WHP and RPC projects may be
fruitful.
REFERENCES
Gaffney, V & C Gaffney (eds) 2000
Special Issue: Non-invasive investigations at Wroxeter at the end of the Twentieth Century, Archaeological
Prospection 7(2).
van Leusen, PM 1999b
The Viroconium Cornoviorum Atlas: high resolution, high precision non-invasive mapping of a Roman
civitas capital in Britain, European Journal of Archaeology 2(3):393-405.
CHAPTER
4
D E A L I N G W I T H R E C E N T P O S TDEPOSITIONAL AND RESEARCH
BIASES IN ARCHAEOLOGICAL
LANDSCAPES∗
1
INTRODUCTION
This chapter builds on a previous publication on bias modelling (Van Leusen 1996) and is about recent
post-depositional and research biases in the kind of data that form the basis for archaeological landscape
reconstruction and settlement history – site-based data coming from desktop study and older surveys, and
land parcel-based data coming from modern surveys. Recent post-depositional biases are nearly
exclusively related to human changes to the landscape and its use; research biases are those biases that
have occurred in the past, and still occur, during the construction of the archaeological record; I
specifically exclude biases occurring during the formation process.
My main point is that recent post-depositional and research biases can not only obscure, but also create
patterns in the archaeological record. This has two consequences: firstly, if significant biases in the data
we work with are not dealt with, then our reconstructions based on those data will be significantly flawed;
secondly, comparison of the archaeological records of the three RPC study regions is predicated on the
assumption that such records are, or can be made, comparable.
1.1
AIMS & DEFINITIO NS
When archaeologists collect data, they nearly always do so non-randomly. In many cases this is
intentional, but it also happens under circumstances where the archaeologist is unaware of, or
underestimates, the selectiveness of his or her sample. Even in modern field surveys, where the data
collection method is intended to give the archaeologist a representative and un-biased sample of the
archaeological material present on the surface, biases introduced by differences in land use, survey
conditions, collection methods, and differences between the individuals taking part in the survey may
conspire to create a highly non-representative data sample instead. My aim in this article is to outline
∗
Much work on this article was done while I was working with the Wroxeter Hinterland Project at the Birmingham (UK)
University Field Archaeology Unit. My thanks go to Vince Gaffney, for allowing me to develop my interest in GIS-implemented
bias modeling, and to Roger White and his volunteers for providing me with some of the most systematically collected survey
data imaginable. Similar thanks should go to the surveyors – too numerous to mention here - for the various RPC project
campaigns in south-central Italy, and to Peter Attema, for sharing forthcoming survey data. Luke Dalla Bona’s kind invitation to
attend his Ontario workshop on predictive modelling in spring 1997 provided a second and very welcome push to work on the
problem of biases; and as always, members of the GISARCH discussion list were extremely helpful with their comments on
various draft versions.
4 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
approaches toward dealing with the significant research biases which do occur in our data sets, and to
substantiate this with several case studies.
The study of processes transforming the archaeological record is a natural consequence of the theoretical
stance of the New Archaeology. Originally, interest centred on depositional and taphonomic processes
because these provided the link between archaeological assemblages and their
anthropological/ethnographic parallels. As processual archaeology took hold in the later 1970s, a more
general interest in formation processes emerged because it was hoped that if these could be modelled in
sufficient detail, the ‘original’record might be reconstructed (‘behavioral archaeology’; Schiffer 1976).
This extreme processualist stance led to the emergence of the post-processual movement in the early
1980s.
Unfortunately the latter movement has given rise to a general rejection of ‘rule-finding’approaches in
archaeology, thus throwing away the baby with the bath water. This has been particularly clear in current
criticism of GIS-based landscape studies as being ‘environmentally deterministic’ (see discussion in
Gaffney & Van Leusen 1995). In view of this, I should stress that this article is not about the modeling of
past or current taphonomic processes, but rather about past and current research processes - that is, the
processes by which we study the archaeological record1. These have long been recognised as a separate
set of distorting processes (eg, Daniels 1972:202) but have had most of their impact north of the
Mediterranean region (see esp. Hamond 1978, 1980).
A subsidiary point I should make here is that biases operate at all spatial and temporal scales and
resolutions, but that different biases may become significant at different scales and resolutions. For
example, although land use and pedological conditions for the fields surveyed in the WHP were relatively
uniform, the distribution of these conditions within the larger WHP survey transect and 30 by 40 km
study area was decidedly non-random (Van Leusen & Gaffney 1996:303); good examples of the
occurrence of biases at larger scales (smaller distances) are provided by the RPC surveys: historic earth
movement in the Fogliano area, and the ploughing-up of pebbles from underlying conglomerates in the
Francavilla area. I will refer to the issue of scale and resolution throughout this article, especially during
the discussion of GIS-based landscape modeling.
In the remaining part of this introductory section, I will review the background and history of the
treatment of biases in Mediterranean landscape archaeology (concentrating on Italy), in general
methodological studies, and in GIS-based landscape studies. I will then briefly introduce the concept of
bias modelling, which will be treated in detail in section 2. This will be followed by a concluding
discussion (section 3) which also serves to establish a context for the case studies. These were conducted
to demonstrate, firstly, the relevance of bias factors to the interpretation of survey data and of landscape
archaeological data in general; and secondly, approaches to the inclusion of bias factors in geographic
models of archaeological landscapes. The first of the case studies addresses biases occurring at a regional
or supra-regional scale where, usually, ‘sites’are the analytical unit, concentrating on the effects of
differential land use / land cover history in the hinterland of the Roman civitas capital at Wroxeter
(Shropshire, England). The second case study addresses research biases occurring at a local or regional
scale where land parcels are the analytical unit, and concentrates on visibility-related biases in RPC project
surveys in central and south Italy. The third case study again concerns the history of land use, this time in
the context of 20th century land improvement of the Fogliano area of the Pontine region (chapter 17,
published as Feiken & Van Leusen 2001).
1
I prefer this to the alternative name ‘discovery processes’which stresses accidental effects.
4 -
2
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
1.2
HISTORY & TREATM ENT OF BIASES IN MED ITERRANEAN LANDSCAPE
ARCHAEOLOGY
RESEARCH TRADITIONS
The treatment of biases is intimately tied up with more general traditions of archaeological research.
Terrenato (1996) began his discussion of landscape archaeology in the Italian regions of Etruria and
Umbria by remarking that ‘it did not succeed in standardising itself in the period 1960 – 1980’(nor later;
see also Cambi & Terrenato 1994), and that, in order to compare the settlement histories of these regions,
we must first understand what methods were used in each. He defined two methodological traditions: the
Forma Italiae approach originating in the early 20th century (including such non mapsheet-based projects as
the South Etruria Survey), and the systematic sampling designs using transects that began in the late 1970s
(such as the Ager Cosanus, Ager Tarraconensis, and Agro Pontino projects; Terrenato 1996:217-220).
The latter all shared one important design trait - their stratified sampling designs were a-spatial, in the
sense that they aimed to answer questions about properties of the sampled population as a whole, rather
than to record spatial variables. Their aim was to establish differences and similarities between the strata
which could be interpreted as differential use of landscape units (see Orton 2000 for a detailed review of
methodology). In the late 1980s it became clear that neither approach produced satisfactory results2, as
ongoing research began to show that vegetation cover and recent geopedological phenomena played an
important role in the process of finding sites in Central Italy (e.g., Malone & Stoddart 1994:5).
In the early 1980s the general recognition of the existence and importance of biases led to a phase of
experimentation (continuing at a slower pace today) aimed at determining the precise effect of the biases.
An important publication because of its effect on survey work carried out by Dutch teams in Italy was
Shennans East Hampshire study, which aimed to solve “many problems concerning the significance of
information derived from surface survey”, in particular observer effects, differential visibility effects, and
the effects of sub-recent landuse (Shennan 1985:2).
From the early 1990s onwards British and Dutch archaeologists in Italy also began to record off-site
densities (for example, in the Pontine Region, Gubbio Basin, and Rieti projects) while Italians stuck to the
site-based approach developed for the Forma Italiae series. The increasing research intensity required by
off-site approaches, with its emphasis on high resolution data collection and quantification based on land
parcels, has forced many researchers to recognise the importance of coming to terms with biases in the
archaeological record. This point was reiterated recently in the influential (but sadly delayed) POPULUS
project edited volume on ploughsoil assemblages (Francovich & Patterson 2000). It was recognised at the
time (circa 1995) that a practical standardised approach to bias correction would have to be developed
(Millett 2000:92-4). On the other hand, the category of ‘site’ remains a mainstay of archaeological
interpretation in both research traditions, allowing the issue of bias to be ignored to a greater or lesser
extent. Since both of the research traditions mentioned above have contributed data to modern GISbased regional archaeological records created for landscape approaches to archaeological analysis and
interpretation, the challenge now is to conceive of approaches which will be able to deal with biases in
wide area, site-based data as well as in local area, parcel-based data.
GIS-BASED ANALYSIS OF ARCHAEOLOGICAL LANDSCAPES
The emergence, in the early 1990s, of GIS as data management and analytical tools in landscape
archaeology has had a marked influence on both data collection and analytical methods, including those
that have a bearing on biases. The need to collect and standardise data on a regional basis has highlighted
gaps, biases, errors and uncertainties. Not surprisingly, low archaeological data quality has been widely
identified as a fundamental problem for the application of cartographic modelling techniques in
archaeology. Solutions have been sought in two directions: firstly, to accept low data quality and find a
2
Most of these projects still await full publication.
4 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
lowest common denominator in point-based archaeological site records; and secondly, to study ways in
which data quality can be improved.
Published cartographic models of archaeological landscapes (including my own) have, without exception,
been based on point-based records rather than parcel-based distribution maps. However, as I have argued
elsewhere (Van Leusen 1996) the set of point observations (‘sites’) resulting from a typical desktop-based
survey of an archaeological landscape is equally, though differently, biased by the non-random effects of
recent post-depositional research processes. This point will be substantiated in a case study on biases
caused by land use, land cover, and research methods in the Wroxeter Hinterland. GIS analyses of pointbased archaeological records have also suffered from two deficiencies which have prevented them from
dealing adequately with biases, namely the absence of ‘non-site’ observations, and the fact that
cartographic models cannot be tested with the aid of independent control samples. The absence of ‘nonsite’observations means that the landscape characteristics of sites cannot be compared to those of nonsites; the lack of independent control samples means that we cannot test cartographic models for biases in
our input data.
In order to explain why these two are of such great significance when it comes to avoiding biases, I must
briefly digress to explain some principles of locational (location-allocation) modelling in archaeology.
Archaeological locational modeling in GIS is the modeling of the probability of the presence of
archaeological site locations based on the properties of each topological unit (polygon, cell) in the study
area; or in other words, the prediction of the chances of finding a site in any particular location on the
basis of models of the ancient landscape. Many such applications have derived landscape ‘predictors’by
studying the known record of archaeological sites and by assuming that this input (the ‘test sample’) is
more or less representative of all archaeological sites (the ‘population’). Exactly what that ‘population’is,
almost always remains undefined, but it is implicitly assumed to be either a) original distributions of
archaeological sites, b) distributions of sites modified through taphonomic processes, or c) distributions
of sites modified by both taphonomic and discovery processes (Hamond 1980, Orton 2000).
In order to find which landscape characteristics are predictive of site location, and which are not, a
statistical analysis of site characteristics is performed. However, most such methods require that ‘site’
characteristics be tested against ‘non-site’characteristics, that is, the test sample should include data about
site absence as well as site presence. But records of such negative observations were never kept, and thus
the preconditions for such statistical tests cannot be met. To get around this, it has been argued that
random observations, given the scarcity of sites, are likely to be non-site observations; and therefore a set
of such random observations may be used as a stand-in for the missing non-site records. This method
implicitly assumes that the intensity of research has not varied across the study area, but we know in fact
that this is not true – both the quality and the intensity of research display huge variations as soon as the
scale of the study area transcends that of a single project! Thus, the absence of ‘non-site’observations in
regional archaeological records is one cause of the tendency to ignore spatial variation in taphonomic and
discovery processes as a factor in archaeological landscape analysis.
A second way lies in the manner in which models are tested. There are two criteria by which to measure
the quality of any prediction: specificity (yield) and accuracy. Models with a low specificity are
uninteresting; models with a low accuracy are wrong. The specificity of an archaeological locational model
is usually expressed as the ratio of the percentage of input sites and the percentage of total study area. For
example: if the model can describe 75% of its input sites by characteristics occurring in only 10% of the
landscape, its yield is said to be 75/10 or 7.5. Once a model has been made, its specificity can be
calculated; but its accuracy can only be assessed by testing the model - applying it to a set of independent
observations which were not used to construct the model in the first place. Because such independent
observations cannot be collected in a reasonable amount of time and money by doing new fieldwork, this
‘control sample’nearly always derives from the same data set of known archaeological sites from which
the test sample was taken. Since, ipso facto, it has been subject to the same biases as the test sample of sites,
it cannot be used to remove those biases. Therefore, the typical locational ‘predictive’model is in fact
only a descriptive model, and what it describes are the characteristics of the known archaeological sites.
4 -
4
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
The second major approach to problems with archaeological data quality in landscape studies has been to
improve matters either by collecting ‘better’data or by correcting existing data. Both approaches entail
greater attention to the understanding, recording and correction or avoidance of post-depositional and
research biases, usually within the context of intensive ‘non-site’or ‘off-site’surveys. In contrast to the
site-based analyses discussed above, GIS has been applied to the spatially continuous distributions of
archaeological data collected in such surveys at a much more local scale (usually not more than a few
square km). Since most of these studies – among them the analyses of the WHP- and RPC project
surveys - are not yet published, there has not yet been a thoroughgoing discussion of approaches to
dealing with biases in such data. The recent publication of Bintliff’s reanalysis of the Boeotia survey data
and the discussion that followed in the pages of the Journal of Mediterranean Archaeology (Bintliff et al
1999; Bintliff & Howard 2000; JMA 13) may be regarded as the opening shots in this debate. GIS in
general are eminently suited for the management and analysis of archaeological data collected in the form
of quantitative measurements per land parcel, but the higher spatial resolution required by modern
intensive surveys does present the problem that increasingly low densities of artifact are being recorded,
which have a lower diversity, a lower diagnosticity, and are relatively more sensitive to stochastic effects.
The increased emphasis on high resolution quantitative processing has led researchers to devise
increasingly formalised and standardised methods for collecting and recording survey data. To avoid
slowing down the fieldwork too much, these processes are now being automated through the use of
digital field equipment. The potential of portable GPS attached to handheld field computers in particular
is now being explored, and recent experiments suggest that future surveys will operate new, powerful and
versatile navigation, recording and communication tools which are likely to replace all paper-based
recording practises (Van Leusen & Ryan forthcoming, see chapter 7). The gain may be put to good use
for increased attention to biases and methods for dealing with them.
TYPES OF BIASES
As indicated in section 1.1, I define biases not as distortions of some idealised archaeological record, but
distortions in the way we go about obtaining knowledge about that record, and in that sense all biases are
‘research’biases. The point I wish to make in the current section is, that research biases can (and do)
occur at any stage from the definition of research aims through the design and execution of the fieldwork,
and into the analytical and interpretative stages. In order to limit confusion in what follows, I will
distinguish the following bias types occurring during these research stages: Conceptual bias, Visibility bias,
and Observer bias.
Conceptual biases are biases caused by the classification of data under preconceived concepts. These play
an important role at all scales and stages of archaeological enquiry and include the tendency to study only
some geographical, typological, and chronological parts of the available archaeological record.
Paraphrasing the quote made famous by Cherry, Mediterranean archaeologists have been like frogs
around a pond, with very little study taking place in inland and highland areas. Until very recently
archaeological interest in southern Italy for example, was mainly targeted at the Hellenistic colonies there.
Following Livy (XXV: I, 1), indigenous settlement was regarded as uninteresting. A comprehensive and
systematic overview of indigenous settlement in southern Italy was therefore unavailable until recently,
when the University of Lecce began building a regional archaeological site database containing
settlements from the 9th (EIA) to the 3rd century BC. This highlighted both the previously ignored
presence of a dense network of indigenous settlements contemporary to the Hellenistic colonies, and that
of major geographical hiatuses in research, for example in central Puglia and Basilicata (D’Andria 1999).
It also highlighted the typological concentration of research (especially for the Archaic, Classical, and
Hellenistic period) onto necropoleis, wall circuits, and cult places. Similar conceptual biases occur at the
local scale as well – in the ‘judgmental’choice of the study area for example, based on landscape
physiography and on the character of the known archaeological record. Many surveys are targeted at
exploring the ‘catchment’of some known highly monumental site – a town, hillfort, or large villa – and
therefore the results of such surveys are biased in favour of central places and inadvertently stress the the
role of settlement continuity and hierarchisation in the landscape.
4 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
Another form of conceptual bias occurs during classification – the process of sorting our observations
into sets of mutually exclusive classes both in the field and during finds processing. Among the most
important high-level classes we use in the field are those of ‘site’and ‘non-site’. An extensive literature
exists on the theory of the siteless survey and why we should not employ the concept of ‘site’in the field
(Dunnell & Dancey 1983; Ebert 1992); in a very practical sense the continued definition of ‘sites’in the
field (as opposed to during the interpretation) may cause more problems than it solves. If we increase the
‘intensity’of our survey once we have defined a ‘site’,not only does the density (numbers per unit area) of
our samples increase, but so does their diversity (number of distinct material categories). Whereas the
quantitative effect has been noted by many survey practitioners in the past, the more subtle effect on
diversity has escaped general notice. For example, Small (1998:349) noted that the most significant
concentrations of paleolithic flint in his Basentello surveys occur at ‘gridded’sites, where the likelihood of
recovery would be much higher; he did not note whether the diversity of flint types within those
concentrations increased as well. It is to be hoped that the recent publication of re-analyses of the Boeotia
survey data will kindle more interest in the subject (Gillings & Sbonias 1999:45; Bintliff & Howard 2000;
see also below in section 2.1).
A second important set of conceptual biases comes into play during the collection, recording, and
processing of individual finds. The distinction of ‘diagnostic’and ‘indeterminate’finds gives rise to a
tendency to concentrate collection and recording practises on those parts of the surface archaeological
record which are amenable to functional or chronological interpretation. Many past and current survey
methodologies have relied to a greater or lesser extent on the walkers’selection, and the ceramicists’
subsequent preferential processing, of ‘diagnostic’sherds. Besides the admittedly practical reasons – a
wish to limit the volume of finds to be processed - even the value of collecting non-diagnostic types of
material is sometimes doubted. Thus Yntema, in his introduction to the diagnostic pottery types of the
Brindisi area, writes “A field survey with a very limited quantity of diagnostic wares or without expert
knowledge of regionally current ceramics of the periods on which the survey centres offers no sound
basis for regional studies and is to be considered a waste of time and money” (Yntema 1993:29).
In the RPC survey areas in central and southern Italy the concentration on diagnostic wares and forms
leads to biases in favour of the better preserved and better studied pottery types, which are
chronologically tied to dated sequences outside the regions themselves (sometimes even outside Italy).
This introduces the further danger of assigning too strict a dating range to particular types of ceramics
(see Bintliff & Howard 2000). In this, survey ceramic specialists are perforce dependent on similar
practises in excavation. Malone and Stoddart (2000:95-6) note that the tendency to date excavated
features from lower frequency diagnostic elements (which are rare in surface deposits) is especially
marked in the later prehistoric metal producing periods. Since the progressive degradation of prehistoric
pottery leads to a relative dearth of diagnostic sherds anyway (Malone & Stoddart 2000: fig 11.1), and the
recognisability of ceramics varies wildly over time with low recognisability occurring at phases of
important reorganisation within societies (Malone & Stoddart 2000: fig 11.2), our reliance on diagnostic
material indeed introduces strong biases.
All non-total collection strategies seem to me to contain built-in conceptual biases of this type because
they require the field team, or even the individual walker, to make decisions about which finds are
‘diagnostic’and ‘indeterminate’. Under typical survey conditions, not even a team consisting of period
specialists could reliably make such decisions.
Visibility biases are research biases caused by regional and local variations in the visibility of the
archaeological record. The term ‘visibility’has been used to describe a number of things, but in the
current context I will equate it to the retrieval rate - the probability that an artefact lying within a walker’s
transect will be recorded. On the widest scales this type of bias is obviously mediated by land use and land
cover, which circumscribe the outcomes of different types of regional research. Examples of this are the
peculiar sets of sites generated by aerial reconnaissance in the Wroxeter Hinterland (LULC bias, see
chapter 14) and by ‘topographic’surveying in the Pontine Region, Etruria, and more generally in the
Mediterranean (see chapter 13). On the scale of an individual survey, visibility bias is mediated by local
variations in a large number of factors, of which traditionally the amount of vegetation covering the
4 -
6
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
soil surface has been seen as the most important one. As I will argue in more detail below, this is a
simplistic view and the understanding of local visibility biases will require much more research effort.
Observer biases, as distinct from visibility biases, are concerned with the ability of the observer to record
information which is available in principle. Again, examples can be adduced for both the regional and the
local scale of enquiry. In taking aerial photographs, and again in mapping from aerial photography,
researchers have consistently selected interpretable shapes such as circles, squares, and linear features. In
surveys, many observers have noted the different abilities of individual walkers to ‘see’different artefact
types such as flint, but opinions differ as to how significant such biases are. My own opinion, which I feel
confident would be backed-up by a survey of the relevant psychometric literature, is that the differences
between observers are much larger than most students of the problem realise, and extend to such areas as
general ability to recognise shapes and colours, and to concentrate for any length of time.
Besides individual differences, there are also generalised observer effects for which many examples can be
adduced from the RPC surveys. The height of the observer is one such effect: the closer one comes to the
surface, the more detail (including finds) one sees. ‘Hands and knees’inspection of limited areas yields
finds where normal upright survey does not; it follows that small people (all other things being equal) will
find more stuff than tall people. Another effect is distraction, where one or more finds categories may
become ‘invisible’because the eye is distracted by more prominent features of the soil surface. For
example, even though prehistoric ceramics may have lain on the surface in parts of the SIBA2000 survey,
the presence of overwhelming numbers of similar-coloured and –shaped fragments of ploughed-up
conglomerate rock made it impossible to pick up and inspect many fragments in detail; in other survey
areas a similar role may be played by large numbers of more ‘attractive’Hellenistic/Roman or recent
ceramics.
Observer biases are treated here as a separate category despite their links with both conceptual and
visibility biases, because they highlight weaknesses in one particular link in the whole research process –
our ability to make reliable observations – which has been largely ignored so far. It is precisely in high
intensity surveys of off-site areas that even minor observer biases can have major consequences on results
and, eventually, interpretations.
Summarising all of the above, archaeological research can be biased by conceptual, visibility, and observer
biases. Some of these we cannot avoid; others we may not even wish to avoid. But we must study these
biases if we want to be aware of the systematic distortions they create in our archaeological record; and if
the distortions can be quantified we can attempt to correct for them (see section 2).
2
BIAS MODELLING
Bias modelling is the cartographic modelling, usually with the help of a GIS, of the presence and value of
factors influencing the discovery rates of parts of the surface archaeological record, and of the nature of
that influence, with the aim of correcting the outcome of primary research results. The following sections
detail the stages involved in this (with examples), ending with a discussion of current issues.
2.1
DEALING WITH BI ASES
Archaeologists’responses to the question of how to deal with biases have been three-fold: firstly, to
ignore it because the problem is thought to be insignificant or insoluble; secondly, to avoid it; and thirdly,
to model and attempt to correct for it. Before detailing my own approach to this issue I will need to say a
few brief words about the first two types of response.
In spite of the arguments expounded in section 1 above, some will disagree with either the need or the
feasibility of dealing with biases. As noted in section 1.2, while many researchers involved in
4 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
surveying in the Mediterranean will agree that there is a vast underestimation of the impact of distorting
effects on the quantification of surface material, some believe that the effects of these biases are too
complex to be modeled and corrected (Fentress and Ammerman, cited in Cambi & Terrenato 1994:16871; Terrenato & Ammerman 1996:93-5; Fentress 2000). If taken literally, from this belief it would follow
that the current practice of recording bias factors is largely a waste of time. It is difficult to judge the value
of this belief without the benefit of having detailed published arguments available, but it may be pointed
out we are not seeking to model all the distorting processes (which would indeed be impossible) but only
their effects.
Bias modelling is an essential step in the production of locational models in any study that uses biased
data sets – and there aren’t any others in archaeology. It is similar in this respect to other historical
sciences, which employ source criticism to trace and correct for distortions in their sources. Yet despite
the occasional research into methods needed to correct survey data for known biases, current everyday
practice seems very little affected.
One can think of both general and specific reasons for this: archaeologists (like other people) are
generally overconfident when they estimate the reliability of their results; and having to deal with biases in
our data adds to an already high workload. Archaeologists would rather spend their fieldwork time
collecting fresh archaeological data than on what they perceive to be mere methodological niceties.
The second response type – to avoid biases altogether – is clearly a valuable approach if it can be realised;
but so far I have been able to find only a single example of this. Van de Velde (1996, 2001) claims to have
avoided surface visibility bias during the Riu Mannu survey in Sardinia by conducting, alongside a
traditional extensive survey for diagnostic materials, a systematic gridded point sample survey, in which
two square meters of ground surface were cleared at each sample point ‘to provide quantitative control
without visibility effects’. It is not clear that such a method can be extended to any other factors causing
visibility bias or other types of bias; and Van de Velde himself acknowledged that geopedological effects
cannot be controlled this way. In my concluding section I will return briefly to this issue.
Turning now to the third response type, methodologies for dealing with biases can be described as
following three consecutive steps – recognition, recording, and correction. Each of these is described in
more detail below, and then discussed in the following sections.
RECOGNISING BIASES
A bias factor must first be recognised as such, and its significance must be assessed in a preliminary
manner in order to determine whether it is necessary to continue with the next step. Since the significance
and even presence of bias factors varies over geographical space, it is probable that no regionally or
supra-regionally valid ‘standard’set of bias factors can exist. Instead, significant bias factors must be
recognised locally and for each material category separately. Excellent examples can be cited from RPC
fieldwork experience. Land improvement schemes in the Pontine region had resulted in the localised
burial or removal of ancient land surfaces depending on the terrain morphology and soil type; since the
Fogliano survey area straddled several such terrain units this had an obvious effect on the outcome of the
survey (chapter 17). In the SIBA2000 survey, by contrast, the localised ploughing-up of conglomerate
bedrock proved to be the most significant bias factor in many parts of the landscape unit under
investigation, and the discovery of dull and undiagnostic prehistoric pottery was most significantly
affected (chapter 12: 11). At this stage, surveys may be designed to avoid some biases.
RECORDING BIASES
Once it has been established which bias factors are likely to have a significant effect on the outcome of a
regional or local survey, they must be recorded and assessed. These are two separate things though they
are often confused. To record the bias factor means to measure or estimate the degree of its presence; to
assess it means to measure or estimate the effect of that presence on the retrieval rates of all material
categories.
4 -
8
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
For example, the stoniness of a field may be recorded as a bias factor by estimation using the MOLAS
field guide (Spence 1990: chart 11), or by measurement in the field or from photogrammetry, but the
actual amount of bias due to this factor is not yet known at that stage - 20% stoniness does not equal 20%
less finds. Yet the two are often equated, especially in the case of vegetation cover which is the single
most recorded bias factor in mediterranean surveys. In other words, a linear relation is assumed to exist
between the bias factor and the bias itself, whereas in fact this relation has to be assessed either by
estimation from field observations (the current, though rarely explicit, practise at almost all current
surveys) or by statistical analysis of the correlation between bias factors and retrieval rates, or has to be
measured by conducting field experiments.
Since in most cases there will be multiple significant bias factors present, it may be more relevant to
evaluate their combined effect on the retrieval of each material category (for example by factor analysis or
logistic regression techniques) than to do this for each factor individually. Typically, the end result of this
step is one or more biases expressed on a percentage scale from 0 (no bias; perfect retrieval) to 100 (total
bias; no retrieval).
The distinction between the bias factor and the bias itself, seemingly of academic interest only, may be
seen to be important in the handling of differences between material categories. If the bias is equated to
the bias factor, it becomes an objectified, measurable, ‘environmental’variable rather than one which is
relative to the characteristics of each material category - implying that the bias affects all material
categories to an equal degree, which is patently untrue.
CORRECTING BIASES
Once a bias factor has been recorded and its effect on the retrieval rates of all material categories
assessed, the next logical step is to correct the ‘raw’retrieval rates for this effect. Nance (1983:350) already
advocated the formulation of correction factors in the context of probabilistic sampling. In current
practise, corrections to raw survey data are typically applied by multiplying the counts or weights of
retrieved material categories by the inverse of the relevant bias percentage. A good example of this is
Gaffney’s (Gaffney et al. 1991) correction of raw survey data for surface visibility percentage in the Hvar
project. However, there are problems with this method at both the ‘high’and the ‘low’ends of the
distribution of densities per collection unit, albeit for different reasons. At low retrieval rates, quantitative
correction for biases ignores the problem of statistical diversity and enhances the effects of statistical
noise: multiplying up low retrieval rates will not increase the number of distinct types within the
assemblage, and insignificant density variations can become significant by multiplication. High retrieval
rates typically occur at ‘sites’, where the collection strategy is likely to have been different in various
respects from the one normally used, so that the collected assemblages are no longer representative of
what lay in the field. Multiplication by bias correction factors would then result in completely unrealistic
densities for some material categories.
In other words, current methods will not correct the fact that rare categories will be underrepresented in
areas with a low retrieval rate, while changes in the types of samples taken from areas with high retrieval
rates will tend to lead to overrepresentation of rare and ‘diagnostic’categories, and underrepresentation of
‘indeterminate’or ‘uninteresting’categories. As will be suggested below, the problem of low diversity may
be countered, at the expense of spatial resolution, by merging neighbouring collection units with low
retrieval rates until sufficiently large assemblages have been created. The ‘high end’problem can be
avoided only by a rigorous separation of the ‘standard’samples taken from every collection unit, from the
‘special’samples taken once a ‘site’has been defined.
2.2
IDENTIFICATION A ND ASSESSMENT
IDENTIFICATION
One of the effects of the ‘New Archaeology’has been that, beginning in the late 1970s, many authors
have identified and discussed the impact of factors that bias our knowledge of the archaeological
4 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
record. This is not the place to reproduce long lists of such factors; I will concentrate here on those
factors that are most pertinent to the discussion.
The three factors which, since 1980, have attracted most attention both in the surveying literature and in
the literature on GIS and spatial analysis, are geomorphological processes such as erosion and deposition
(eg, Vermeulen & De Dapper 2000; Allen 1991), land use and land cover (eg, Van Leusen 1996), and
sampling and surveying technique itself (Shennan 1985, Ammerman & Terrenato 1995, Verhoeven 1986).
Geo(pedo-)logical research is increasingly seen to be a requisite part of regional project designs, not just
to map one of the most important factors in determining past land use, but also to map bias in the survey
results caused by natural geomorphological and human processes. While geomorphological processes
work over relatively long time-scales, RPC fieldwork experience in the Pontine and Salento Isthmus
regions confirms that many landscapes have been seriously affected by recent or sub-recent
anthropogenic soil movement and restructuring (see chapters 11 and 17).
The fact that the effects of bias factors vary by region, by period and by material category has been widely
recognised in the last decades, but has so far not led to a systematic approach to the treatment of biases in
Mediterranean archaeology. A major reason for this may be the difference in climates and sedimentary
regimes as compared with north-western Europe. Whereas Mediterranean archaeologists have an
extremely rich and, in the arid climate, well-preserved surface archaeological record to study, the situation
in many parts of north-western Europe is radically different, with sedimentation burying many sections of
the archaeological records and wetter climates giving rise to extensive areas of grasslands.
It is no wonder then, that north-western archaeologists were forced relatively early on to consider bias
factors and how they affect different site types, periods, and regions. Among the published studies of
regional landscape archaeology in northern and western European archaeology, Fokkens’(1991, 1998)
thesis on the settlement history of the north-western Netherlands stands out for its clear and systematic
approach. His regional, site-based study of ‘map formation processes’ stands out as a model still
unrivalled by later GIS-based studies. Fokkens, using a somewhat different bias typology from mine,
recognises post-depositional bias factors and research bias factors. Within the scope of his study, soil
types, land use, and historical land reclamation activity are identified as significant for the former, while
the localised activity of amateur archaeologists is identified as the main significant factor for the latter.
The fact that the Mediterranean surface archaeological record appeared to be generally much more
complete than its northwestern European counterpart has caused Mediterranean archaeologists to take a
different approach to biases, stressing one factor in particular that has a direct, observable, effect on
retrieval rates: land use / land cover (LULC). Cambi and Terrenato (1994:151-2) report that, following
the then recent recognition of the significance of biases by Italian archaeology, new volumes in the Forma
Italiae series of regional site-based surveys have begun to include mapping of non-visible areas. As an
advocate of the ‘Anglo-Saxon’approach, Terrenato himself, in the Cecina survey, considered zones of
geological deposition and zones where land cover largely impeded survey to be the most significant bias
factors (Terrenato & Ammerman 1996). On the local scale, because surveys are targeted towards areas of
optimal land use/land cover to begin with, many practitioners continue to equate ‘visibility’with the
percentage of ground surface not covered by vegetation.
As I have argued above, the tendency to think of the surface archaeological record as something which is
there to be observed, and will in fact be reliably observed by any qualified observer unless blocked by
intervening sediment or vegetation, has one further consequence for the identification and assessment of
bias factors in Mediterranean archaeology - the low retrieval rates of material categories such as flint are
regarded as a consequence of choices made during survey design (in this example, whether a lithic
specialist would be included in the field teams, or not), rather than as a permanent, built-in feature of any
type of archaeological fieldwork. While it is to be hoped that a specialist will attain a higher retrieval rate
for his or her particular material category, it does not follow that his or her assemblages are unbiased.
4 -
10
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
Within Italy, the assessment of the significance of bias factors relative to material categories attracted early
interest through the evaluation of the South Etruria project data. At the time, Di Gennaro and Stoddart
(1982), already noticed the low visibility of prehistoric and medieval ceramics; later on, Malone and
Stoddart (2000, echoed in Kuna 2000:34-6) went on to provide several quantitative arguments to explain
this phenomenon in terms of formation, taphonomic, and recovery processes.
These authors begin their argument by pointing out that the relative abundances of prehistoric ceramics
of different periods may be culture-specific, that is, related to a cultural tendency to dig pits (thus
preserving archaeological materials sufficiently long for their discovery in the present). Kuna calculates
that a prehistoric residential farm in his Bohemian study area may yield only a few ceramic fragments in a
survey at 10% coverage. Although such calculations provide support for an assumption I made for the
interpretation of the Fogliano survey data (Attema et al. 2001), they can only explain why prehistoric
material is underrepresented in the ploughsoil (formation processes) rather than why archaeologists have
trouble recovering a representative sample of the material which is in the ploughsoil. However, Kuna goes
on to demonstrate that the probability of occurrence of a distinctive (diagnostic) fragment within such a
sample varies by period, and decreases with elapsed time. And as we have seen in the discussion of
conceptual biases above (section 1.2.3) the preferential treatment accorded to diagnostic finds implies that
both the variable ratio of diagnostic fragments and its general decline through time constitute significant
bias factors for some material categories.
RPC field survey experience suggests that even this does not come close to identifying all the relevant
biases in regional and local archaeological data sets, as the following brief discussion will show.
At the regional and supra-regional scale, the site-based records compiled by desktop studies or in
‘topographic’survey can be shown to be influenced by differences in accessibility of the terrain. Before
about 1970, the probability of archaeological material being observed, recognised, and reported was
largely due to chance and the activities of local amateur archaeologists. Field observations could only be
made in accessible areas, and the more accessible an area the more likely that it would be visited by a
person able to recognise archaeological features; ‘chance’finds are therefore more likely to occur nearby
modern infrastructure than away from it. Surveys conducted in the ‘topographic’tradition (which requires
a very large area to be surveyed by one or a few persons at most) continue this trend, as is shown clearly
by the Forma Italiae volumes for the Pontine region which were produced between 1920 and 1970 (cf. the
discussion in chapter 13).
At both the regional and local scales, the significance of the influence of subsequent human occupation
and land use on earlier remains appears to be severely understudied by Mediterranean archaeologists3.
Authors such as Shennan (1985), Verhoeven (1991), Van de Velde (1996) and Burgers and Yntema (1996)
do not mention this factor at all, although it is in operation on any but the most short-lived site and can
cause bias both because earlier periods tend to be ‘lost’among remains of a later date, and because the
greater research intensity at sites of a later date can lead to the discovery of low density remains of an
earlier period. Both effects have a bearing on the apparent amount of settlement continuity in a study
region, as is shown by Attema’s (1993, forthcoming; see also Van Leusen 1998) study of a possible
Roman Republican ‘villa colonisation’in the foothills of the Lepine mountains.
My third and final example concerns a previously unidentified bias in the recording of the surface area of
collection units and the subsequent calculation of finds densities on that basis. Typically, the raw counts
and weights per collection unit resulting from a survey will be ‘normalised’to account for any differences
in the size and coverage rate of the collection units. It was found during analysis of the Ostuni survey data
(Attema et al 2001) that the digitised areas of many collection units were approximately 10% smaller than
those mapped in the field on topographic maps at scale 1:10,000. The difference is explained by the fact
that topographic maps generalise and omit features such as paths, wayside berms and scrub, and
buildings; however these features were included in the detailed hand-drawn field maps made of each
3
Fokkens (1998:59) does mention this factor.
4 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
collection unit which formed the basis for the digital map. In the relatively well-mapped Ostuni survey
zone the effect remains fairly small, but in the topographically complex and poorly mapped terrain of the
SIBA2000 survey the areas of some agricultural fields were overestimated by as much as 50%, because the
boundaries of ploughed fields were ‘stretched’to include the surrounding unmapped scrubland as well.
The variable percentage by which the areas of collection units are overestimated indicates how large the
potential errors can be when ‘normalisation’of survey results takes place. This is therefore a potentially
significant bias, especially where it is intended to detect and interpret relatively small variations in finds
densities across the landscape.
Assessing a bias factor means asking the question: Does this factor have a significant effect on the types
or amounts of information I can retrieve from the archaeological record? A simple way to begin
answering this question is by plotting the two against each other in a graph. Examples of such graphs
were produced as far back as 1978, when Fred Plog and his colleagues plotted the number of
person/days per square km surveyed against the number of sites discovered for 12 surveys in the
American southwest (see figure 2; Plog et al. 1978, 389-94 and fig 10.1). Cherry (1983, fig 1) did the same
for a large number of survey projects in Greece, adding a historical perspective by grouping the projects
into broad types as practised since about 1950. In figure 1 I have extended his plot for Italy, including
both extensive regional and intensive local surveys for the Pontine Region. Although Cherry used
surveyed area, rather than the more precise effort in person/days, as his measure of research intensity,
both studies clearly show that more intensive research leads to a higher retrieval rate (in this case, of
sites). While Cherry’s historical perspective uncovers the incredible increase in site density achieved since
the earliest surveys (10-fold even in the last 30 years), Plog’s diagram highlights the fact that there is an
almost linear relation between intensity and result – for every duplication of research intensity, an
approximate 75% rise in the number of sites may be expected.
Figure 1 – Site productivity in relation to surveyed area for different survey types in Greece and
Italy (adapted from Cherry 1983, fig 1). Note that the scale on both axes is logarithmic.
4 -
12
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
On a regional and supra-regional scale, differences in research intensity have been shown to play an
important part in creating spatial variation in site density. As noted already by Nance (1983:311), ‘much of
the variation apparent in data from survey to survey may be explained by differences in survey intensity’.
Moving to more local scales, examples of the assessment of bias factors have been published since about
1990 and have concentrated on the relation of ‘visibility’(however defined) and site density. Again,
Cherry, in his analysis of the results of the Keos survey (Cherry et al 1991, fig 3.6), appears to have been
the first to publish a plot of visibility versus site density (my figure 3), apparently confirming that
vegetation cover is inversely related to site discovery rate. However, such plots presuppose the detailed
recording of the bias factor, which is the subject of the next section, so I will discuss them there.
Four aspects of the examples discussed above are noteworthy: firstly, the relation between research
intensity and site density cannot remain linear - there must be a point of diminishing returns, where
increasing the research intensity will no longer result in the discovery of more sites. It is likely that
modern high-intensity surveys, with crew spacings of less than 10m, have reached this point. Secondly,
measuring research intensity is not trivial. Crew spacing and walking speed are the two most obvious
factors involved, but in almost all cases only the former is recorded. A more reliable proxy measure of
research intensity is therefore likely to be the ratio of the area covered and number of person/days spent,
expressed as either person-days per unit area or area per person/day. Thirdly, measuring the ‘result’of a
survey by counting the number of sites per surveyed unit area becomes increasingly meaningless as the
definition of ‘site’has begun to shift and blur, including smaller and less dense scatters, and is even
ceasing to be used as a unit of discovery. Fourth and lastly, increased research intensity leads not only to
the discovery of more sites, as the examples above have shown, but also to the discovery of different types
of sites: the ones that are smaller and less visible. In my intra-regional comparison of survey data sets of
the Pontine region (chapter 13 section 2) all of these aspects are shown to play a role in confounding
successful quantitative comparisons.
In view of the above, our assessment of research intensity as a bias factor at the regional and supraregional scale must be, that differences in research intensity will not only result in significant spatial
variation in site densities, but also in significant spatial variation in distributions of site types. This
assessment can help us avoid misinterpreting variations caused by the bias factor, with obvious
consequences for settlement chronology and landscape history; but only if it is properly recorded and if
we succeed in quantifying its effects.
4 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
2.3
RECORDING AND EV ALUATION
Whilst bias factors such as vegetation cover can be recorded, their effect on the retrieval rate (ie, the bias
itself) cannot; it has to be evaluated. This section deals with methods for doing both, again with examples
ranging from the supra-regional to the local scale.
RECORDING
Current approaches to the recording of bias factors in the field are based on, and in fact almost identical
to, methods first developed for Mediterranean surveys in the early 1980s. In the Pontine Region,
approaches to the recording of visibility and observer factors developed for the Agro Pontino Survey
project (Voorrips et al. 1991: 82 ff.) were adopted by subsequent Dutch survey projects directed by
Attema, and in turn formed the starting point for the recording experiments carried out during the RPC
project surveys in the Pontine Region and elsewhere.
The most basic method imaginable for recording any bias factor is the ‘binary’approach, in which each
factor can assume only one of two values – present / absent. An example of this is the recording by
Terrenato and Ammerman (1996) of visibility factors during their survey of the Cecina valley, during
which conditions of geopedology and vegetation were classed as favorable or unfavorable (see figure 4a).
While such an approach can be valid for some factors, it does tend to simplify the reality of field
situations, and most other researchers have therefore adopted various simple ordinal rating scales to
record bias factors. Where multiple bias factors have been identified, each is usually recorded separately
on an ordinal scale. In the RPC surveys, five factors are recorded in this manner: vegetation cover,
stoniness, sun/shade, soil weathering, and ploughing conditions. Overall visibility is independently
recorded as a sixth variable. In the Fogliano survey (Attema et al. 2001) overall visibility was recorded as
two related ordinal variables: Field visibility and Block visibility. The former, which rated the overall
visibility of an agricultural field from low through normal and high to optimal, had to be corrected using
the relative visibility rating recorded for each collection unit as low, normal, or high. Low relative
visibility, for example, resulted in ‘demoting’a particular collection unit to a lower overall visibility class.
Figure 4 - relations between visibility and site density for the Cecina survey (from Terrenato
1996). A: site density by visibility class; note that the density in unfavourable geopedology
and vegetation is twice that of the two more favourable central classes. B: site density by
surface visibility for each of 23 square km
The rating of bias factors by estimation along an ordinal scale of measurement is to some (large but
unknown?) extent subjective, so that it is not possible to compare these across projects. It may be
possible to improve ratings using a method called CARS (criterion anchored rating scale) which anchors
such ratings to specific criteria (Suenson-Taylor et al 1999). A rating will then be arrived at by assessing
these criteria much like grain size is arrived at by comparison with standard grain sizes during the
recording of soil properties. This method is claimed to give a proper numeric scale along which to
4 -
14
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
measure the factor, but has not been applied to field archaeology yet as far as I am aware.
EVALUATION
The effect that a specific bias factor has on the retrieval rate of a specific material category can be
evaluated by estimation, measurement, or statistical analysis. Measurement requires the setting up of
controlled field experiments in which know densities of the material are surveyed under different
conditions. Statistical analysis of the influence of bias factors on retrieval rates has only been pursued by
relatively few researchers, among whom Shennan (1985) still takes pride of place. Shennan used
techniques of logistical regression in order to study the impact of a large number of factors on the results
of his experimental survey area in Hampshire; identical methods were later applied to the results of the
Agro Pontino Survey project by Verhoeven (1991).
However, the large majority of past and current practitioners of field survey employs simple, ‘intuitive’
methods for evaluating the effect of bias factors. Because there is no formal method underlying this step,
it receives little or no discussion in the literature. Examples from the RPC field surveys include the
estimation of the overall percentage visibility effect for each of the visibility classes distinguished in the
Fogliano and Ostuni surveys (see chapter 11, page 3). Often the rating for the bias factor is taken to be
the rating for the effect itself, as in the Boeotia survey where the ground surface visibility score is derived
directly from the 10 point rating scale used to rate land use/land cover (Gillings & Sbonias 1999:36). At a
larger scale Fokkens (1998:64-5), too, based his calculations of archaeological visibility indices on
estimations of the bias caused by each of his distorting factors, with 100% indicating optimal recovery
conditions for each.
The statistical approach to evaluating bias factors initially seemed to hold promise of a much more precise
and replicable method. Shennan (1985:38-9) concluded that, although not entirely negligeable, distorting
factors did not seem to have a major impact on the retrieval rates of his three most frequently occurring
material categories (explaining some 17-18% of variation in them). Certainly their effect was less than that
of environmental variables, which he calculated to be on the order of 40-60%4. Verhoeven (1991:87)
found that the effects of weather and field conditions on the results of the Agro Pontino Project surveys
had a similarly limited effect but that ploughsoil conditions (dust/rain) do have a strong influence on the
recovery of flint and obsidian5. Going on to consider less frequent material categories, Shennan found
that light conditions did play an important role (34% of variation explained) in the recording of RomanoBritish pottery.
Going one step further and using his detailed records of who surveyed where, Shennan (1985:40-44)
statistically removed the visibility biases he identified, and went on to look for observer bias in the
residuals. His analysis confirmed that observer bias, though fairly minor, was definitely present. For
chipped stone, his analysis showed a small (3%) but highly significant observer bias; for burnt flint and
post-medieval pottery the variability of the walkers was somewhat greater (explaining 9-10% of the
variation in retrieval rates). Shennan also noted that the differences between walkers in picking up various
materials were quite large.
It is unfortunate that, after these initial attempts at formal evaluation of bias effects, no further research
was published, and practitioners reverted to estimation methods. Two new twists to this approach were
added in the 1990s by, firstly, plotting the effect of bias factors against retrieval rates and, secondly, by
employing simulation studies. Figures 3 and 4 present two examples of site density plotted against
4 This confirms my observation during the Fogliano survey – observer bias and differential visibility differ by finds category. The
fact that Shennan chose to study his most frequently occurring find categories already entails less distortion!
Van de Velde (2001) notes that the method used by Shennan and Verhoeven to calculate ‘field effects’underestimates the true
effect because it disregards ‘zero’observations.
5
4 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
visibility percentage for the Keos and Cecina surveys. Terrenato (forthcoming) also investigated the
effects of surface visibility (geopedology and vegetation) on the recovery of archaeological sites through
simulated survey coverages.
To conclude: what seems to be lacking in the ‘estimation’approach to recording and evaluation of bias
factors, is the presence of a testing phase in which the accuracy and replicability of both the recording
method and of the evaluation ‘rule’can be assessed.
2.4
CORRECTION
Although bias factors are being recorded, the apparent failure of formal evaluation methods has meant
that many researchers used this information only for informal adjustment of the raw survey results. Even
simple map overlays of the bias factors onto the dataset being studied can be very effective in this respect,
and neither Fokkens (1998) nor Attema (pers. comm.) went on to apply formal correction methods to
their data. Three quantitative correction factors, however, are currently being applied in many
Mediterranean surveys because of their computational simplicity – unit area, coverage, and overall
visibility; these will be discussed in the following sections.
What constitutes ‘correction’? Following Shennan, Van de Velde (1996:23) and Verhoeven (1991), believe
that observer effects such as the well-known ‘flint expert’phenomenon can be corrected for by rotating
individual walkers during a field survey. I believe this to be incorrect; while rotating the walkers would
certainly randomise such effects, it does not eliminate them. For example, where-ever a colour blind
observer is put in a field walking line-up, the transects walked by that person will turn up less small sites.
UNIT AREA
Correction for unit area is needed because (ceteris paribus) a larger area will yield proportionally more
finds. This requires the unit area to be measured with a specific precision and accuracy, relating to the
desired degree of confidence that a specific density difference is indeed a significant archaeological
variation rather than a result of random errors in the measured unit area. As we have seen above, such
measurements can indeed be surprisingly inaccurate, and require the adoption of more stringent field
methods (see chapter 7). The correction itself transforms finds counts per unit into finds densities,
expressed as number of finds per hectare within the RPC project.
COVERAGE
The single variable which has been successfully corrected for in surveys (in regional site-based records it
was never recorded in the past), is the coverage (percentage of the collection unit actually observed). This
correction method has even been applied in the field, in cases where the distance between walkers was
adjusted whenever changes in vegetation cover were encountered. Van de Velde (2001) applied this
relatively simple correction based on the % coverage of his survey to arrive at a corrected total number
of sites for one of his survey transects, and reports that this outcome agrees reasonably well with the
number of sites independently discovered by local amateurs within the same area. Such observations are
the first step toward proving that retrieval rates corrected by formal methods are less biased, and form a
better basis for interpretation.
However, coverage is itself composed of two variables: the average distance between walkers and the
swath width. While the former is generally recorded, the latter is set to a notional value (usually 2 m, but
values from 1 to 5 m have also circulated). Seemingly small differences in the swath width (eg, 1.5 vs 2 m)
represent a large percentage change (25-33%) which, if used in a correction formula, will make
comparison between surveys difficult. The correction usually takes place by multiplying the original finds
count per unit by the inverse of the percent coverage for that unit, so that the new finds count represents
what would have been recovered had the unit been fully covered.
4 -
16
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
OVERALL VISIBILITY
The rationale for correcting for visibility is, as we have seen, that the probability of recovering a surface
artefact decrease as the surface itself becomes less visible. Correction takes place by multiplying the
original finds count with the inverse of the visibility score. Gillings and Sbonias (1999:36), for example,
used the 10-point ground surface visibility score recorded by the Boeotia survey to correct raw artefact
counts using the formula
Corrected = recorded * (10 / visibility score)
This, in effect, represents a conversion of the visibility score to a visibility coefficient, which is then used
to multiply raw artefact counts in areas of low visibility. Similar methods were used in the analysis of the
RPC surveys near Ostuni and Francavilla Marittima (chapters 11 and 12).
THE PROBLEM OF LOW DENSITIES
Modern field surveys, with their increasingly detailed and intensive study of ‘off-site’areas, tend to result
in large numbers of collection units containing very few finds. In addition, these finds tend to be less
different from each other – in statistical terms, the diversity of the assemblage is lower. This gives rise to
several interpretative problems. Firstly, the multiplication of finds counts in the course of bias correction,
as suggested in the previous paragraph, will increase the numbers of finds but not their diversity, thereby
causing a relative drop in variability. A method for avoiding this, namely the aggregating of collection
units until a given minimum number of finds is obtained, was suggested as early as 1973 (Dunnell &
Dancey 1983:272), but it remains to be seen how the loss of spatial resolution affects subsequent analysis.
A second implication is that low finds densities can no longer automatically be dismissed as insignificant
‘noise’or ‘off-site material’.Instead, the survey intensity and visibility biases should be taken into account.
Higher survey intensity equates to a higher diversity of the recovered finds assemblage; for example,
Gillings & Sbonias’(1999:46-53) detailed chronological discussion of a single site assemblage of the
Boeotia survey ignores the fact that the site itself was much more intensively surveyed than its
surroundings, and low density unobtrusive find types present on the site could just as well have been
present outside the site grid. Likewise, re-surveying experiments at Fogliano have confirmed that a single
sherd recovered under circumstances of very low visibility can develop into a scatter if the survey takes
place under more favourable circumstances. As it is recognised that even intensive modern surveys
cannot hope to collect representative samples of low density, low visibility categories such as the
(possibly) prehistoric impasto occurring in the Sibaritide foothills (chapter 12), dedicated specialist (re)surveys are required for such materials.
Beyond the purely quantitative aspects of interpreting low density ceramics, there are also some less easily
categorised aspects. Low densities can, for example, be an artefact of the finds classification process, in
that material from some periods can only be recognised if diagnostic forms or decorations are present. If
these are rare or absent, finds will be classified into broad undiagnostic categories, or even as
‘indeterminate’. Finally, the belief that material can be transported very far from its origin by slope
processes or ploughing, prevalent among archaeologists of the Roman school and invoked regularly to
explain away the occurrence of low-density ceramic scatters near urban sites, seems unwarranted in many
cases. Bintliff’s recent work on re-interpretation of the extra-urban ‘blankets’of surface finds from the
Boeotia survey is confronting this issue head-on, and is attempting to distinguish between several
alternative causative processes for low-density distributions6.
Rather than posing a problem, Bintliff and Howard (2000) note that low densities, in combination with certain diagnostic types,
may indicate the presence of a Hellenistic cemetery.
6
4 -
17
V A N L EUSEN : P A T T E R N T O P ROCESS
3
CONCLUDING DISCUSSION
Our ability to record surface archaeological material is not perfect; it is biased by visibility and research
biases. An example of the former is current and historical land use / land cover (LULC; see especially
chapters 14 and 17); examples of the latter are the recording and classification methods used (chapters 13
and 16). Of the two courses of action that have been proposed to remedy the problem, that of avoiding
biases seesm to have little if any potential beyond Van de Velde’s experiments with surface cleaning.
Recording and attempting to correct for biases is the only alternative, but before this can be done with
any degree of confidence in the results, dedicated studies will need to underpin the methodology. Data on
research and visibility biases have so far not been systematically collected at regional scales, although
suitable methods have been developed. Terrenato (1996: 227 –228) underlined the importance of
recording bias factors if we are to attempt ‘the correction, at least partially, of incomplete distributions’,
and advocated ‘a series of methodological experiments dealing with the various aspects of how to
document surface scatters’, e.g. by replication studies, all within a local (regional) context. This research
was then, and still is now, ‘a high priority for survey practitioners in Central Italy’. In fact such
experiments should probably have precedence over the collection of bias data, because we do not yet
understand how best to do the latter.
The issue of bias is almost universally agreed to be an important one, but neither the intensive interest
and study conducted in the early 1980s, nor the constant popularity of surveying or, latterly, the use of
GIS, have so far led to anything resembling a concerted effort to develop a methodology which is valid
across projects. The plethora of biases discussed in this article requires that we return briefly to the
fundamental question raised most recently by Fentress: Could it be that we will not be able to disentangle
the mess, and correct the distortions? I hope I have been successful in arguing that many lines of research
remain to be explored before giving in to such a counsel of despair.
CASE STUDIES
The case studies presented in chapters 13, 14, and 17 were conducted to demonstrate a) the relevance of
bias factors to the interpretation of survey data and of landscape archaeological data in general; and b)
methods by which bias factors can be included in geographic models of archaeological landscapes. At the
regional scale, studies of the data collected by the Wroxeter Hinterland Project and the Agro Pontino
Project (Voorrips et al. 1991) demonstrate this for systematically surveyed data and general archaeological
records; at the scale of a ‘local’survey such as the Ninfa and Fogliano surveys conducted in 1998-9, case
studies demonstrate this for specific visibility and research biases. In the WHP surveys, conducted in
1994-6, the choice of fields was limited by modern land use and land cover (LULC), in particular the
availability of freshly ploughed surfaces. Since these are not randomly distributed over the landscape –
relief, distance to Severn, soil type and hydrology all play a role – they result in the taking of a biased
archaeological sample. In the surveys conducted by the Agro Pontino project, paleosurfaces dating to the
paleolithic period had been covered in some parts of the Pontine plain by more recent alluvial and
colluvial deposits, and similar though less clearly evident biases must have been present for material
dating to later periods.
REFERENCES
Alcock, SE, JF Cherry & JL Davis 1994
Intensive survey, agricultural practise and the classical landscape of Greece, in Morris, I (ed) Classical
Greece: Ancient Histories and Modern Archaeologists, pp 137-170. Cambridge: Cambridge University
Press.
Allen, MJ 1991
Analysing the Landscape: a Geographical Approach to Archaeological Problems, in Schofield, AJ (ed),
Interpreting Artefact Scatters: Contributions to Ploughzone archaeology (Oxbow Monograph 4):39-57.
Ammerman, AJ 1985
Plough-zone experiments in Calabria, Italy. Journal of Field Archaeology 12:33-40.
Attema, PAJ (ed) 1995
4 -
18
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
Nederlandse surveys in Italie en Griekenland: een inventarisatie. Proceedings of a workshop on
Mediterranean surveys, Groningen, 19 May 1995.
Attema, P, M van Leusen & E van Joolen in press
A Marginal Landscape: Field work on the beach ridge complex near Fogliano (South Lazio). PaleoHistoria
40/41 (2001):149-162.
Attema, P, G-J Burgers & M van Leusen forthcoming
The Ostuni Field Survey Campaign 1999. Studi di Antichitá 11 (2001).
Baxter, MJ 1994
Exploratory Multivariate Analysis in Archaeology, 2nd edition. Edinburgh: Edinburgh University Press.
Bintliff, J 1992
Appearance and Reality: Understanding the Buried Landscape, in: Bernardi, M (ed), Archeologia del
Paesaggio, pp 89-140. Florence: All’Insegna del Giglio.
Bintliff, J & Ph Howard 2000
Studying Needles in Haystacks – Surface Survey and the Rural Landscape of central Greece in Roman
Times, in Pharos (Journal of the Netherlands Institute at Athens) VII (1999).
Bintliff, J & A Snodgrass 1988
Off-Site Pottery Distributions: A Regional and Interregional Perspective, in Current Anthropology 29
(3):506-512.
Bintliff, J, Ph Howard & A Snodgrass 1999
The Hidden Landscape of Prehistoric Greece, JMA 12(2): 139-168.
Blankholm, HP 1991
Intrasite Spatial Analysis in Theory and Practice. Aarhus: Aarhus University Press. [13-1]
Bradley, R, T Durden & N Spencer 1994
The creative use of bias in field survey, Antiquity 68 (1994):343-6.
Burrough, PA and RA McDonnell 1998
Principles of geographical information systems. Oxford: Oxford University Press.
Cherry, JF, JL Davis & E Mantzourani (eds) 1991
Landscape Archaeology As Long-Term History. Northern Keos in the Cycladic Islands. Los Angeles:??.
Cressie, NAC 1993
Statistics for spatial data. New York (etc): Wiley.
D’Andria, F n.d. (1999)
Ricerche recenti sugli insediamenti indigeni di Puglia e Basilicata, in La forma della città e del territorio,
esperienzemetodologiche e risultati a confronto (Atti dell’Incontro di studio – S Maria Capua Vetere 27-25
novembre 1998):103-18.
Daniels, SGH 1972
Research Design Models, pp 201-29 in DL Clarke (ed), Models in Archaeology. London: Methuen.
Di Gennaro, F & S Stoddart 1982
A review of the evidence for prehistoric activity in part of South Etruria, Proceedings of the British School
at Rome 50:1-21.
Dunnell, RC & WS Dancey 1983
The siteless survey: a regional scale data collection strategy, Advances in Archaeological Method and
Theory 6:267-87.
Ebert, JI 1992
Distributional Archaeology. Albuquerque: University of New Mexico Press.
Feiken, H & M Van Leusen in press
Interpreting Field Survey Results In the Light of Historic Relief Change: the Fogliano beach ridges (south
Lazio, Italy), in Stancic, Z et al (eds), Proceedings of the annual meeting of CAA, 2000.
Fentress, E 2000
What are we counting for? In Francovich & Patterson (eds) 2000:44-52.
Fokkens, H 1998
Drowned Landscape. The Occupation of the Western Part of the Frisian-Drentian Plateau, 4400 BC – AD
500. PhD thesis, Assen: Van Gorcum.
Francovich, R & H Patterson (eds) 2000
Extracting Meaning from Ploughsoil Assemblages (The Archaeology of Mediterranean Landscapes 5).
Oxford: Oxbow Books.
Gaffney, VL, J Bintliff & B Slapsak 1991
Site formation processes and the Hvar survey project, Yugoslavia. In Schofield, AJ (ed), Interpreting
Artefact Scatters: Contributions to Ploughzone archaeology (Oxbow Monograph 4): 59-77. Oxford:
Oxbow Books.
4 -
19
V A N L EUSEN : P A T T E R N T O P ROCESS
Gaffney, V & PM van Leusen 1995
Postscript: GIS and Environmental Determinism, in G Lock & Z Stancic (eds), GIS and Archaeology: a
European Perspective, pp 367-82. London: Francis & Taylor.
Gillings, M & K Sbonias 1999
Regional Survey and GIS: the Boeotia Project, in Gillings, M, D Mattingly & J van Dalen (eds),
Geographical Information Systems and Landscape Archaeology (The Archaeology of Mediterranean
Landscapes 3):35-54. Oxford: Oxbow Books.
Goovaerts, P 1997
Geostatistics for Natural Resources Evaluation. New York: Oxford University Press.
Hamond, FW 1978
The Simulation of Early Neolithic Settlement Development in the Lower Rhine Basin. Ph.D. Thesis,
University of Cambridge.
Hamond, FW 1980
The Interpretation of Archaeological Distribution Maps: Biases Inherent in Archaeological Fieldwork, in
Archaeo-Physika 7:193-216.
Isaaks, EH & RM Srivastava 1989
An Introduction to Applied Geostatistics. New York (etc): Oxford University Press.
Kuna, M 2000
Surface Artefact Studies in the Czech Republic, in Bintliff, JL et al (eds), The Future of Surface Artefact
Survey in Europe. Sheffield: Sheffield Academic Press.
Leusen, PM van 1993
Cartographic Modelling in a Cell-Based GIS, in J Andresen, T Madsen & I Scollar (eds), Predicting the
Past. Computer Applications and Quantitative Methods in Archaeology 1992, pp 105-124. Aarhus: Aarhus
University Press.
Leusen, PM van 1996
Unbiasing the Archaeological Record, Archeologia e Calcolatori 6:129-136.
Leusen, PM van 1998
Archaic Settlement and Early Roman Colonisation of the Lepine Foothills, in Assemblage 4 (1998), at
http://www.shef.ac.uk/~assem/4/.
Leusen, PM van & VL Gaffney 1996
Extending GIS Methods for Regional Archaeology: the Wroxeter Hinterland Project, in H Kamermans & K
Fennema (eds), Interfacing the Past. Computer Applications and Quantitative Methods in Archaeology
1995 (Analecta Praehistorica Leidensia 28):297-305.
Livy
The War with Hannibal. Books XXI-XXX of the History of Rome from its Foundation. Harmondsworth:
Penguin Classics.
Malone, C & S Stoddart (eds) 1994
Territory, time and state. The archaeological development of the Gubbio Basin. Cambridge: Cambridge
UP.
Malone, C & S Stoddart 2000
The current state of prehistoric ceramic studies in Mediterranean survey, in Francovich, R & Patterson (eds)
2000: 95-120.
Nance, JD 1983
Regional sampling in archaeological survey: the statistical perspective, Advances in Archaeological Method
and Theory 6:289-356.
O’Brien, L 1992
Introducing Quantitative Geography: measurement, methods and generalised linear models. London:
Routledge.
Orton, C 2000
Sampling in Archaeology. Cambridge manuals in archaeology. Cambridge University Press.
Plog, S & M Hegmon 1993
The sample size-richness relation: the relevance of research questions, sampling strategies and behavioral
variation, American Antiquity 58(3):489-96.
Plog, S, F Plog & W Wait 1978
Decision Making in Modern Surveys, in MB Schiffer (ed), Advances in Archaeological Method & Theory
1:384-421.
Rossignol, J & L Wandsnider (eds) 1992
Space, Time, and Archaeological Landscapes. New York: Plenum Press.
Shennan, S 1985
4 -
20
P OST -D E P O S I T I O N A L A N D R E S E A R C H B IASES
Experiments in the Collection and Analysis of Archaeological Survey Data: The East Hampshire Survey.
Sheffield: Department of Archaeology and Prehistory.
Shennan, S 1988
Quantifying Archaeology. Edinburgh: Edinburgh University Press.
Siegel, S & NJ Castellan 1988
Nonparametric statistics for the behavioural sciences. New York: McGraw-Hill.
Small, A et al. 1998
Field Survey in the Basentello Valley on the Basilicata-Puglia Border, in Echos du Monde Classique /
Classical Views XLII, n.s. 17: 337-371.
Spence, C 1990
Archaeological site manual, 2nd Edition. London: Museum of London.
Suenson-Taylor, K, D Sully & C Orton 1999
Data in conservation: the missing link in the process. Studies in Conservation 44:184-94.
Terrenato, N & AJ Ammerman 1995
Visibility and Site Recovery in the Cecina Valley Survey, Italy. Journal of Field Archaeology 23
(1996):91-110.
Terrenato, N 2000
The visibility of sites and the interpretation of field survey results: towards the analysis of incomplete
distributions, in R Francovich & H Patterson (eds), Extracting Meaning from Ploughsoil Assemblages (The
archaeology of Mediterranean landscapes 5): 60-71. Oxford: Oxbow Books.
Terrenato, N 1996
Field Survey Methods in Central Italy (Etruria and Umbria), Archaeological Dialogues 3 (1996),2:216-230.
Velde, P van de 1996
Off-site probleemstellingen, nonsite technieken: De Riu Mannu survey op Sardinië, Tijdschrift voor
Mediterrane Archeologie 17 (1996): 22-29.
Velde, P van de 2001
An Extensive Alternative to Intensive Survey: Point Sampling in the Riu Mannu Survey Project, Sardinia,
JMA 14(1):24-52.
Verhoeven, AAA 1991
Visibility factors affecting artifact recovery in the Agro Pontino survey, in Voorrips, A, SH Loving, and H
Kamermans (eds), The Agro Pontino Survey Project (Studies in Prae- and Protohistorie 6): 87-98.
Amsterdam: Instituut voor Pre- en Protohistorische Archeologie.
Vermeulen, F & M De Dapper (eds) 2000
Geoarchaeology of the Landscapes of Classical Antiquity (Bulletin van de Antieke Beschaving
Supplementa 5). Leiden: Stichting Babesch.
Yntema, D 1993
In Search of an Ancient Countryside. The Amsterdam Free University Field Survey at Oria Province of
Brindisi, South Italy (1981-1983) (Scrinium 6). Amsterdam: Thesis publishers.
4 -
21
C HAPTER 5
A REVIEW OF WIDE-AREA
PREDICTIVE MODELING USING
GIS∗
1
INTRODUCTION
"Predictive modelling is a technique to predict, at a minimum, the location of archaeological
sites or materials in a region, based either on the observed pattern in a sample or on
assumptions about human behaviour"
- Kohler & Parker 1986:400
The above quote, from a paper published in one of the later edited volumes on New Archaeology theory
and method (Schiffer 1986), encapsulates the principles of what, only a few years later, was to become a
cottage industry of predictive models when GIS became more widely available to archaeologists (Scollar
1999:7). Since then, several hundreds of articles and tens of edited volumes have appeared, most of them
concerned with the predictive modelling of the location of as yet undiscovered archaeological remains,
either in the context of cultural resource management (CRM) or of academic research.
Archaeological predictive modelling can be conceptualised as a specialised form of what planners call
location-allocation analysis, in which the object is to allocate ‘suitable’locations to specific types of human
activities (and, by extension, to their archaeological remains), and in which the criteria for suitability are
derived by location analysis – the generation of behavioural rules from a set of observations about how
people actually behave or have behaved in the past. The study of ‘spatial’archaeology began in the 1970s,
but had different origins in the USA (economic geography) and Britain (the diffusionist school). Although
by the early 1980s archaeological theorists had largely turned away from the rule-based approaches
advocated by the New Archaeologists, this type of reasoning and analysis still received a boost by the end of
the 1980s when its implementation was much facilitated by affordable computers and GIS software (see the
overviews by Kvamme 1990, 1999).
As this new area of archaeological research unfolded, many became uncomfortable with the lack of
theoretical depth and methodological rigour of most of the published work. In 1993 I presented two review
papers on the role of GIS in locational modelling, one concentrating on then current approaches in Dutch
archaeology (Van Leusen 1996), the other on its potential for archaeological resource management (Van
Leusen 1995). More recently, I and others co-authored an updated review of predictive modelling in the
Netherlands (Verhagen et al. 2000), concluding that basic concerns about the quality of published and
ongoing work had not been adequately addressed in the meantime. Looking ahead, the current chapter is
intended to assess the potential of predictive modelling for wide-area (regional and supra-regional)
∗
This chapter is partly based on my earlier review of Dutch approaches to archaeological predictive modeling (Van Leusen
1996b). Many of the issues discussed here were developed and clarified in meetings of the 'bath-house' group, resulting in a
preliminary version of this chapter, co-authored with Philip Verhagen and Milco Wansleeben, being presented at the 4th
international conference 'Archäologie und Computer' in Vienna (1999, published as Verhagen et al. 2000), and the successful
submission of a project proposal for an in-depth study of predictive modeling in Dutch archaeological resource management
(Kamermans 2001). The current chapter, however, substantially reflects my own personal research and opinions with regard to
predictive modeling.
5 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
archaeological research, and to argue the importance of adopting formal modelling procedures. In this
introductory section I shall first discuss the range of aims and approaches to predictive modelling for which
GIS has been used since the early 1990’s, followed by an evaluation of the underlying theory and concepts.
In section 2 the scope and limitations of predictive models are discussed with reference to data quality and
methodological issues. A concluding section looks at the future use of predictive models in both CRM and
academic research.
1.1
AIMS AND APPROAC HES OF PREDICTIVE MO DELLING
Archaeological location models have been made with two types of aim in mind. In academic contexts, the
aim has generally been to generate models that explain the observed distribution of archaeological
remains, whereas in CRM contexts the aim has been rather to generate models that estimate the
probability of an archaeological site being present anywhere within the study region. Whilst in theory
these two aims might have been approached in different ways, as we shall see in practise there is little
difference between the approaches adopted by cultural resource managers and academic archaeologists.
Many conventional accounts of predictive modelling in archaeology attempt to draw a geographical
distinction in which North American approaches are set against those prevalent in Europe. The North
American attitude toward the use of GIS for predictive modelling is said to be ‘pragmatic’: GIS is a tool
that can be used to apply traditional archaeological analytical methods to very large (previously too
complex) data sets, especially in the context of CRM where it can be used for prediction as well as
modelling the state of preservation and vulnerability of archaeological remains, and to provide
management options (Wescott & Brandon 2000, chapters 3-5). In other words, society needs to manage
and protect its cultural resources, and predictive modelling is a relatively cheap and effective way of doing
this.
The British, and to some extent European, approach is ‘idealist’: we must attempt to understand past
behaviour before we can successfully attempt to predict it. These divergent approaches to the issue of
archaeological prediction have been seen to exist since the early use of GIS in the late 1980s and early 1990s,
and to be exemplified by the studies presented in two recent edited volumes (Lock 2000, Wescott &
Brandon 2000). However, on closer reading we find that the papers published in the latter volume were
originally read at the 1996 Society of American Archaeologists meeting, in reaction to the 1990-5 phase of
early and uncritical enthusiasm about GIS, and a direct comparison with the papers in the former volume
(presented at a 1999 symposium in Ravello) would be unfair.
In addition to the ‘minimum’aim of modelling the location of archaeological remains, predictive models
could conceivably also be used to predict the type and quality of those remains, their current state of
conservation and likely rate of deterioration, and from these deduce their cultural and scientific interest.
Work in this direction has so far been limited to quality studies of known monuments (Darvill & Fulton
1998) and theoretical work (Deeben et al. 1999).
MANAGEMENT VS. RESEARCH BASED MODELS
Predictive modelling was initially developed in the USA in the late 1970s and early 1980s, evolving from
governmental land management projects in which archaeological remains became regarded as ‘finite, nonrenewable resources’, and gave rise to considerable academic debate (Carr 1985; Savage 1990). Until the
start of the 1990s the emphasis of this debate was on the statistical methods used to evaluate the
correlation between archaeological parameters and the physical landscape (e.g., Parker 1985, Kvamme
1985). Within Europe, the Dutch practice of predictive modelling has been most clearly influenced by the
American tradition, probably because archaeological predictive modelling was first introduced in the
Netherlands relatively early on by Kvamme (Ankum & Groenewoudt 1990, Brandt et al. 1992), and has
since been used widely for CRM purposes at regional and national scales (Verhagen 1995; Deeben et al.
1997, 2001; Deeben & Wiemer 1999).
5 -
2
W IDE -A R E A P R E D I C T I V E M O D E L I N G
European academic interest in predictive models using GIS grew out of its long-standing concern with
locational models in general, and has been largely directed at an understanding of the modelling process
itself. The primary result of this has been a series of papers critical of the inductive, CRM oriented
approach common in Dutch predictive modelling (van Leusen 1995, 1996; Kamermans & Rensink 1998;
Kamermans & Wansleeben 1999); at the same time alternative methods and techniques were explored as
well (Wansleeben & Verhart 1992, 1997, 1998; Kamermans 2000; Verhagen & Berger 2001). More
recently, European researchers have begun to concentrate on the incorporation of social variables into
their predictive models (Wheatley 1996; Stancic & Kvamme 1999; cf. papers in Lock 2000).
The contrast between academic and CRM-driven predictive models is likely to continue to play an
important role for the foreseeable future. Reflecting a European trend, all three Dutch universities with a
European archaeology department (Leiden, Groningen and Amsterdam) have in recent years founded
excavation firms1. The privatisation and commercialisation of the archaeological field has unmistakably
increased the influence of tight schedules and customers waiting for the end product, on the actual work.
Unlike academically employed archaeologists, commercial firms have only limited possibilities to
investigate new lines of research, to contribute to the scientific interpretation of their finds, and to
improve research methodologies. Good archaeological research in a commercial context is equivalent to
efficient research: only a limited number of tried and tested methods will be applied. The development of
new methods is restricted to situations where direct benefits are expected for the company. These
benefits can be either a more efficient research strategy, or a new product that will attract the attention of
potential customers. Predictive modelling in a CRM context has been employed in both ways: it can be
used by consultancy firms to guide surveys more efficiently, and it can be used by planners to integrate
archaeology in urban and rural planning at an early stage. In addition to prediction, GIS can be used for
modelling the state of preservation and vulnerability of archaeological remains, and to provide
management options.
Since cultural resource management continues to be a driving force behind the development of predictive
modelling methods using GIS, academic researchers now face the choice of ignoring this development
altogether, or of attempting to establish a research programme that will result in the improvement of
current management-oriented predictive models. For example, Lang (2000:216) noted that GIS “are
becoming increasingly common tools for the national and local inventory records (… ), and are essential
elements of the 30 or so Urban Archaeological Databases developed in England (… )”; hence, new
research into spatial analysis in archaeology should be especially concerned with archaeological resource
management. Within Europe, archaeological risk assessment – in which predictive models play a central
role - will become a standard procedure in planning after the implementation of the Valletta treaty
(Verhagen 2000:234).
THE TYPICAL AND THE EXCEPTIONAL: TRENDS AND RESIDUALS
Whereas predictive modellers typically attempt to discover and model patterns and trends in the
characteristics of a set of archaeological site locations, these models can also be used to detect the
converse – the exceptional or, in modelling terms, the residuals remaining after removal of the trend. The
potential value of such an approach for both management and academic purposes, while recognised by
some (eg. Altschul 1990:231), has remained largely unexplored. As we shall see below, their primary
interest may be in removing broad environmental trends and focusing attention on less well understood
aspects of the data.
1.2
THEORY AND CONCE PTS
Kvamme (1999: 171) recently reviewed the analytical capabilities of GIS, describing predictive models of
archaeological location as ‘[models that] go a step further by multidimensional merging of what is known
1
respectively Archol, ARC and AAC.
5 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
through locational analysis. In one sense such models are descriptive statements that summarise the
multivariate environmental and spatial pattern of archaeological sites; in another they form predictive
statements because a good model can indicate likely locations of as yet undiscovered sites’(my emphasisMvL). In what follows I shall discuss the various elements of this description.
THEORETICAL BASIS: THE LOGIC OF PREDICTION
A review of the literature reveals that much of the energy of the first generation of archaeological
predictive modellers has gone into developing and debating an understanding of the theoretical basis for
predictive modelling itself. This debate has generally been cast in the form of a series of dichotomies, the
first two of which have already been mentioned:
North America
ßà
Europe
CRM
ßà
Academic
Inductive
ßà
Deductive
Ecological
ßà
Cognitive
Parallel to the American/European divide and the CRM/academic contrast, debate has also raged on the
appropriate logic of prediction, and has become polarised around the issues of the use of inductive vs.
deductive logic, and of ecological determinism vs. social/cognitive models. I will briefly discuss the
content and significance of these debates before proposing some alternative concepts on which to base
archaeological prediction.
Inductive vs. deductive approach
Part of the early appeal of GIS for archaeological predictive modellers was its ability to handle and
visualise large and complex data sets consisting of both archaeological site records and large numbers of
mapped environmental variables. Querying these allowed researchers to derive the ‘properties’ of
archaeological sites with ease, and to extrapolate likely locations of unknown archaeological sites on the
basis of these properties in the form of maps. This has been termed the ‘inductive’approach because it
derives rules from observations rather than from theory. Inductive models have been popular in academic
archaeology as well as in CRM, but more attention was paid to methodological issues than to actual
‘working’predictions of site densities. By the early 1990’s papers began to appear criticising the predictive
models made for CRM as being rather crude, lacking a theoretical foundation and therefore failing to take
into account the cultural and environmental mechanisms that produced the statistical correlations that
were found (Wansleeben & Verhart 1992, 1997; van Leusen 1993; 1995; 1996). At the time, Ebert (2000)
argued strongly against the then current practise of purely inductive predictive modelling, and for the
need to include archaeological explanation within a systems theoretical context. In inductive models gain,
he thought, might never get higher than about 70% because of inherent limitations to the approach
(Ebert 2000:133). Yet cultural resource managers in the Netherlands as elsewhere have continued to
produce inductive predictions, taking into account some of the methodological critique (Verhagen 1995,
Wescott & Brandon 2000).
The alternative, ‘deductive’ approach attempts to construct predictive models on the basis of our
understanding of past human behaviour – especially economic behaviour. For a particular archaeological
period and region an assumption of "self supporting agriculture without manuring, but with long fallow
periods" might be made. On the basis of environmental variables that are relevant to this assumption, the
site distribution is predicted and the known archaeological sites are only then used to evaluate the
prediction. One early study which demonstrates the potential of this approach is Chadwick’s (1978, 1979)
model of the Late Helladic (Mycenaean) settlement system, entirely on the basis of premises about
environmental preferences and the prior Middle Helladic population distribution. Later, Kamermans
introduced land evaluation into Dutch archaeology as a fully deductive way of predictive modelling
5 -
4
W IDE -A R E A P R E D I C T I V E M O D E L I N G
(Kamermans 1993, 2000; Kamermans et al. 1985; Kamermans & Rensink 1998). This work has to date
not resulted in a formal methodology for deductive modelling that can easily be applied in the context of
archaeological resources management, although some attempts in that direction have been made by Dalla
Bona (1993, 1994, 2000) and Dalla Bona and Larcombe (1996). The trend to reject inductive models for
deductive ones continues to gather pace in the USA as well as in Europe. In their concluding paper to the
‘predictive modelling kit’volume, Church et al. (2000) advocate adapting approaches from landscape
ecology. And this is indeed the direction taken by subsequent modellers, witness Krist’s recent thesis on
paleo-indian subsistence and settlement in northern Michigan (Krist 2001), and papers presented at a
recent conference on the future of predictive modelling in Argonne (IL, USA; esp. Whitley 2000, 2001).
However, it should be noted that several influential authors (among whom Kvamme) are resisting this
trend and continue to point out the advantages of the inductive approach.
It should be noted that the inductive-deductive dichotomy does not run parallel to the environmentalcognitive dichotomy. More-over, the division between inductive and deductive approaches to predictive
modelling itself is in practice not very distinct2. On the one hand, supposedly ‘inductive’ models
incorporate many assumptions about past human behaviour – why else would one attempt to correlate
the location of sites with, say, terrain slope? Critique by many post-processualists and some processualists
that induction lacks a theoretical basis is therefore misguided (cf. Kvamme 1999:173). On the other hand,
at least part of the archaeological ‘expertise’that goes into deductive models is based on informal induction
- Why do we think that the Linear Band Ceramic people in the southern Netherlands preferred loess
soils? Because that is where we have found most of their settlements. For example, Dalla Bona (2000:90)
claims no actual sites were used to generate his predictive model of boreal forests in Ontario (Canada)
and therefore it is a deductive model. However, this is not quite true because the geographic rules
established for prehistoric activities have been formed partly under the influence of known sites. Recent
predictive models by RAAP and ROB have therefore been termed ‘hybrid’, since inductive statistics are
only used to obtain a first impression of site location characteristics, and general knowledge about the
locational behaviour of human societies in the past is then added to the model.
Ecological determinism vs. Post-modernist approaches
A second dichotomy which has unduly polarised the debate regarding the theoretical foundations of
predictive modelling is to do with the perceived theoretical poverty of what has sometimes been termed
‘ecological determinism’(for an extensive treatment see Gaffney & Van Leusen 1995), usually contrasted
with the theory-laden humanistic approaches advocated by various strands of post-modernist
archaeologists. As a dispassionate evaluation of the practical differences in approach between the two sides
in this debate shows, the only significant difference is in the use of ‘cognitive’variables (see also the brief
discussion by Kvamme (1999:182)). Since both sides in the debate have stuck to deterministic modelling,
the middle ground in a theoretical sense may be said to be accurately represented by Renfrew and
Zubrow's (1994) cognitive processualism.
Alternative distinctions based on the Model Aims
Our understanding of the logic of archaeological predictive modelling is therefore not helped by the
above distinctions. I would therefore like to propose here two alternative sets of distinctions based on the
model aims rather than its methods or theoretical stance. A first useful alternative classification of
predictive models, into correlative and explanatory classes, bases itself on the ultimate aim of the modelling
attempt. If the ultimate aim of a model is to understand aspects of past settlement and land use behaviour,
then prediction is only the means by which that understanding can be tested, and the model may be
termed explanatory. If, on the other hand, the ultimate aim is to conserve the archaeological heritage, then
Kvamme (pers. comm.) notes that the terms 'inductive' and 'deductive' were originally used to describe only the method by
which the so-called layer weights in predictive models were determined; derivation by multivariate statistics was 'inductive', while
derivation by expert judgement or intuition was 'deductive'. Later, these terms were applied to predictive models as a whole,
leading to the current terminological confusion. He further notes that pure induction, though rare, has been used in prediction of
archaeological site locations by regression analysis of satellite remote sensing data.
2
5 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
the task of prediction is to estimate, as accurately as possible, the probability of the presence of
archaeological remains in all parts of the study region, and the model may be called correlative. This
distinction was in fact made early on but has since been forgotten (cf. Sebastian & Judge 1988:4).
Strangely, very little attention has been paid to a second alternative distinction, concerning the choice
between two fundamentally different approaches to prediction: possibilism and probabilism. Almost
without exception, archaeological predictive models have been possibilistic: they only indicate how suitable
an area is for a specific activity. In a possibilistic model gain (see below) can never be very high –
providing an explanation for Ebert’s (2000:133) observation that in practise it never seems to get higher
than about 70%. Despite its restricted scope it has been confused with the probabilistic approach which
expresses how likely an area is to have been used for a specific activity.
MODEL ASSUMPTIONS AND EVALUATION
Archaeological models are simplified representations of processes or phenomena occurring in reality
(depending on whether they take an explanatory or a correlative approach). Their value as ‘predictors’is
constrained by their aims and assumptions, and by the means available for testing. The validity of the
assumptions depends on the aims and vice versa. Testing can provide an independent method for
evaluating the quality of a model.
A fundamental but debatable assumption of all current ‘inductive’(and, for that matter, most deductive)
models is that the known archaeological remains are a representative sample of all extant archaeological
remains. If a precise and accurate description of the known sample can be made, so the argument goes, then
we will automatically have a precise and accurate prediction of the parent population of sites. As I have
argued elsewhere in this thesis (chapter 3.1), this assumption has serious consequences for both our
management and our understanding of archaeological resources. In management, predictive models are
employed in order to locate and protect archaeological sites even in areas where no direct proof of their
existence is available. As archaeological evidence only becomes apparent through the destruction of sites surface finds and crop marks implying, for instance, that agricultural practices have damaged a site - we
should expect the best preserved and therefore most valuable archaeology to be in areas from which none
or few ‘sample’sites are known.
While, in theory, predictive models could attempt to predict one or more of location, quantity, quality,
and nature of archaeological values (cf. Kuna 2000:181-2), in practise they have been almost exclusively
concerned with predicting the location of archaeological settlement remains. The precision with which such
predictions have been made varies widely, depending on the modelling approach taken. In the simplest
case, illustrated here in figure 1, predictions are binary (a site is either present in a particular map area, or
it is absent). In this approach there are only four possible ‘states’of the model. In general, the correct
prediction of site presence (state 1) and site absence (state 4) should be maximised, with the
corresponding states of incorrect prediction of site presence (3) and absence (2) minimised, but
archaeological resource management (ARM) concerns mean that a greater importance may well be
attached to lowering the incorrect prediction that no site will be present (state 2). Predictive models therefore
do not aim to obtain the statistical maximum.
P(site)
P(no site)
O(site)
1
2
O(no site)
3
4
Figure 1: predicted vs. observed presence /absence of sites
5 -
6
W IDE -A R E A P R E D I C T I V E M O D E L I N G
In addition to this simple approach resulting in nominal predictions, other approaches have been used to
yield predictions at ordinal/interval (Boolean multivariate) and ratio (probabilistic multivariate) scales of
measurement. Such approaches to prediction are possibilistic or weakly probabilistic in nature, that is,
they only provide relative degrees of probability of site presence – usually of the ‘low, medium, high’type.
Such weak predictions can only be tested through an appropriate sampling scheme, which if properly
executed will result in a particular probability that the prediction is correct, at some desired level of
confidence.
As already discussed above, predictive modelling rests on the assumption that human activities in the
landscape are patterned in various ways and scales. Correlative approaches further assume that available
archaeological data are representative of the ensemble of discovered and undiscovered archaeological
remains (henceforward referred to as the 'soil archive') in general (a sampling-theoretical assumption), and
that better (more, and more detailed) data and statistical techniques will result in better predictions.
Similarly, explanatory approaches assume that past societies have particular structures and economies, and
that taphonomic and post-depositional processes have transformed their remains into the current
archaeological soil archive. Whilst it must be emphasised here that under both types of approach testing is
crucial if any of these assumptions is to be falsified (proven incorrect), one important practical ‘advantage’
of the correlative approach is that it can be conceptualised as the modelling of the discovery of
archaeological remains rather than of their presence. In other words, the prediction is not concerned with
what may be in the ground, but only with what will be discovered in the ground if past mechanisms
causing the discovery of archaeological remains continue to operate in the future - a defensible stance
from a CRM point of view. However, in both approaches the issue of data quality remains crucial (see
section 2.2).
In the calculation of measures of statistical correlation between the location of archaeological remains and
properties of the physical or social landscape, modellers implicitly rely on assumptions inherent in the
statistical tools applied. Foremost of these is the assumption of normality, that is, the assumption that the
values taken on by a variable, when plotted in a histogram, are distributed according to a normal or
Gaussian curve. It can easily be shown that this assumption is incorrect for many of the variables typically
used in predictive models (eg, the distance of sites from the nearest water source). In general, it cannot be
assumed that any relevant population distribution is normal, and therefore non-parametric statistics such as
logistic regression, discriminant analysis etc. are to be preferred unless the data can be normalized.
A second inherent, but mistaken, assumption is that statistical tools developed for non-spatial
applications (think of the drawing of red and white balls from a vase, familiar from high school statistics)
can be applied to geographical data as well. It has long been observed that “… conventional statistical
tests usually require independence among observations, something that is generally untrue of spatially
distributed information, and these procedures are usually aspatial in nature and design” (Cliff & Ord
1981). The most common methods of measuring the statistical relationship between a pair of variables Pearson’s product-moment coefficient and rank order index and Kendall’s tau - do not consider the
association that may exist between nearby locations (spatial autocorrelation) and cannot therefore be used
for the analysis of spatial data in this form. However, some other non-spatial analytical techniques, such
as Spearman’s rank-size rule and principal components analysis (PCA)3, continue to be important in
archaeological analysis. Thus, Kuna (2000:37-41) uses factor analysis to determine diachronic change in
settlement patterns from survey data in Bohemia (Czech Republic). Kvamme (1993:92) provided a clear
demonstration of the significance of the ‘first law of geography’in a GIS context by reshuffling his
initially uncorrelated sample data to produce two obviously correlated spatial variables, which these tests
would still claim to be uncorrelated. Statistically, a high degree of spatial autocorrelation means a lower
effective sample size, thus lowering the significance of any correlations as measured by non-spatial
statistics.
3 To perform principal component transformation on GRASS data layers, r.covar is used to get the covariance (or correlation)
matrix for a set of data layers, m.eigensystem to extract the related eigen vectors, and r.mapcalc to form the desired components.
Then, using the W vector, new maps are created of the principal components in the input data, in decreasing order of variance.
5 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
Most research on archaeological predictive modelling has concentrated on correlating a set of known
archaeological sites to a multivariate environmental data set, and modelling the presumed distribution of
such sites by extrapolation. Various rival methods have been developed, of which the main two are the
logistic regression (logit) analysis perfected by Warren (e.g. Warren 1990, Warren & Asch 2000) and the
rule-based methods exemplified by the work of most European researchers. Criticism of these methods
has been targeted at both the theoretical underpinnings (as in the ‘environmental determinism’debate, see
Gaffney and Van Leusen 1995) and the methodology (Van Leusen 1996). Some researchers are turning to
qualitative evidence derived from oral history and ethnographic studies in an attempt to construct
predictive models that are partly based on cognitive factors (Pilon et al. 1997). Others, taking a leap of
faith, interpret correlation as causation. A recent example of this can be found in Stancic & Veljanovski
(2000), who interpret the statistically significant nearness of some Roman villas on the island of Brac
(Dalmatia) to a geological unit known as ‘Brac marble’as an indication that these villas were somehow
associated with marble exploitation.
Model Quality
Since many alternative models could conceivably be made for any specific area and period, it is important
to be able to rate the models relative to each other. How is one model 'better' than another? What is a
‘good’predictive model? Several answers to this question have been suggested, depending on whether
one’s focus is on results, methodology, or explanation.
• Specificity
If the aim of a predictive model is to circumscribe the ‘allowable’geographical space for a specific set of
archaeological remains, then a good model might be one that circumscribes this space very narrowly – it
should be specific. A non-specific prediction is a useless prediction. But of course the predictions made by
a good model should also be accurate, because a very specific but also inaccurate prediction is worse than
useless. The quantitative quality of a model should therefore be measured along both of these axes, and
this is in fact done by Kvamme’s gain measurement (%sites - %area), popular mainly in the USA (cf.
papers in Wescott & Brandon 2000). Because the proportion of sites included in the model is important
in itself, and normalisation of the gain parameter is desirable, Wansleeben and Verhart (1992:103-7)
advocate a refined measurement Kj=√(%sites * gain / %area without sites)4. However, it should be born
in mind that such calculation of model gain are typically based on existing site records, and strictly
speaking they therefore do not measure the quality of prediction at all, but rather of retrodiction. While for
CRM purposes being able to predict the absence of archaeological remains might be extremely valuable,
current approaches have no method for handling nonsite data - that is, the proven absence of
archaeological sites in an area - exists. Such data are needed for a) exercising control over statistically
derived predictive models and b) optimising any predictive model (see Kvamme 1983 for the role
nonsites play in the calculation of probabilistic models).
• Falsifiability
From a procedural point of view a ‘good’predictive model is one that follows a defined set of rules
(protocol), is testable, and responds in predictable ways to new data. These characteristics ensure that models
can be evaluated, without which no progress can be made in their scientific understanding. Many current
models incorporate 'black box' stages of expert assessment and adjustment, and therefore fail on the
criterion of protocol. Examples of good models in this procedural sense are presented by Warren
(Warren & Asch 2000:6) and Dalla Bona (2000:77).
A model is a simplified version of reality; a useful model must suggest a hypothesis that allows the model
builder to do an experiment or test. If a model generates no testable hypotheses, then it is useless. If the
hypotheses generated by a useful model are not tested properly, then the model may be incorrectly
4
In cases where nothing is known about non-sites they revert to the simpler formula √(%sites * gain).
5 -
8
W IDE -A R E A P R E D I C T I V E M O D E L I N G
believed or disbelieved. The logical concomitant of a prediction is a test, resulting in a measurement of
model quality, and in some cases even in a verification or a falsification of predictions made by the model.
Whereas proper testing requires observations to be made 'blind' (i.e. independently of the model being
tested), in practise such tests have rarely if ever been carried out. Instead, much weaker forms of testing
have been used, usually by seeing how well the model fits existing evidence, or by looking for
confirmatory evidence (e.g. by surveying areas of high predicted probability).
• Expert consensus
A ‘good’model may also be defined as one that is recognisably congruent with patterns and processes
occurring in the past. Since such a judgement can only be made by appropriate experts, this definition
carries the seeds of circular reasoning - a model is 'good' if its results conform to the expectations of the
experts. This criterion should therefore be discounted on procedural grounds.
From this brief discussion it is clear that the choice of model assessment criteria is fundamental. The
three potential criteria cannot be easily brought into accord with each other, and much therefore depends
on the practical use to which the model will be put. For example, how specific and accurate archaeological
predictive models should be is a question to be answered by planning scientists; current models such as
the Dutch national Indicative Map of Archaeological Values (IKAW) get away with a high level of
generalisation and predictions of unknown accuracy. On the other hand, the principle of public
accountability requires that any models being employed in a formal sense to apply legal restrictions must
be procedurally transparent.
The contrast between the inductive and deductive approaches, discussed earlier, is also expressed here
with respect to the role of expertise in the modelling process, and focuses on the question of whether
models excluding expert knowledge can ever attain a high quality, and whether expert judgement can ever
be formalised in a procedural sense.
Testing
As mentioned above, one way to assess the quality of a model is to test it. Generally, any useful model
must suggest at least one hypothesis that allows the model builder to do an experiment. Since we are
concerned here with predictive models, the logical test to perform would be to see if predictions of site
presence/absence or probability are born out in field research. Ideally, such a test should result in the
adjustment of confidence limits associated with the model being tested (an outright verification or
falsification of model predictions is much less likely). Whereas proper testing requires observations to be
made 'blind' (i.e. independently of the model being tested), in practise such tests have rarely if ever been
carried out. Instead, much weaker forms of testing have been used, usually by seeing how well the model
fits existing evidence, or by looking for confirmatory evidence (e.g. by surveying areas of high predicted
probability). In the Netherlands, an opportunity for a rather stronger form of testing is presenting itself in
the large-scale infrastructural and sand/gravel extraction works being conducted. First generation national
and regional predictive models developed at the ROB are now being ‘field tested’through archaeological
watching briefs at these works.
Both locational and predictive modelling relies heavily on the use of statistics in order to determine
whether the characteristics of a particular set of locations (namely, those where archaeological sites of
interest were found) is sufficiently ‘unusual’to imply something of interest to the archaeologist. The
‘unusualness’of the sample characteristics may be tested against those of a random sample of the same
size (requiring two-sample tests), against all locations in the study area (‘the population’, requiring onesample tests), or against a large number of random samples of the same size (Monte Carlo approach, see
Kvamme 1996).
5 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
2
2.1
METHODOLOGY
IMPACT ASSESSMEN T
It is not the presence per se, but the value of the archaeology which should result in the imposition of
planning restrictions and the listing of monuments; any working predictive model should therefore
incorporate a formal evaluative stage in which such a value is assigned. This is as much a political as a
scientific decision and is therefore to some extent outside the scope of this review. Although researchers
at the Dutch State Service for the Archaeological Heritage (ROB) have been active since the mid-1990s in
studying the methods and data available for impact assessment (Groenewoudt 1994, Groenewoudt et al.
1994, Groenewoudt & Bloemers 1997) and resource evaluation (Deeben et al. 1999), these developments
lag some years behind those in England. There, at a national level, the Monuments Protection Programme
(MPP) monument assessments indicate how important a particular monument is and how much it needs
conservation (Darvill et al. 1987, Startin 1992, English Heritage 1996), mainly with a view of providing
statutory protective designations to the most important monuments. Evaluation systems developed for
this purpose unfortunately remain unpublished in order to prevent their being used as an automated
judgement mechanism (English Heritage 1996:2-3). The MPP methodology to arrive at a national
evaluation of a particular monument type consists of four steps:
• classification and characterisation; relevant information is collected and a full monument type description
is written
• data collection; a thesaurus of monuments is created following consultation with experts, and sites of
potential national importance are identified
• assessment; site-by-site evaluation resulting in overall quantification and ranking
• evaluation; conservation and management options are set out so that policy can be formulated
The MPP evaluation system is based exclusively on known archaeological sites and landscapes, an
approach also taken by CRM groups in the Netherlands and fundamental flaws in which I have discussed
earlier (Van Leusen 1995, 1996). Formal models of threats to recorded archaeological remains in England
were to be based on the national census of the condition and survival of archaeological monuments
conducted by the Monuments At Risk Survey (MARS) project (Darvill & Wainwright n.d., Darvill &
Fulton 1998, Anon. 1998), but so far none have been published. Furthest along in the implementation of
actual threat models seems to be the Ontario Ministry of Natural Resources, where archaeologists have
been building blocks of an ARM system for the past five years (Dalla Bona & Larcombe 1996, Dalla
Bona 2000). This includes the detailed modelling of type and amount of damage (impact assessment)
expected from the activities of the logging industry, which has a direct impact on the planning restrictions
imposed (Gibson 1997).
Nor are data regarding the survival of sites available in anything like the required amount. Many of the
sites on record for more than a decade have disappeared by now. The two major factors here will be
geology (erosion/deposition) and land use (drainage, building, deep ploughing; cf. Hinchcliffe & SchadlaHall 1980 and various books and papers on taphonomic processes by M Schiffer, e.g. Gould & Schiffer
1981).
The role of GIS within a wider archaeological information system has already become crucial in CRM in
many parts of the western world. In the Netherlands, the second version of the IKAW and its local
offspring are currently being used to help make planning decisions from the national to the municipal
level (Deeben et al. 2001). A third generation of the IKAW, currently under development at the ROB
(internal memorandum, 2001), aims to improve its potential as an advisory and decision-making tool, as
5 -
10
W IDE -A R E A P R E D I C T I V E M O D E L I N G
well as to gain more insight into the relations between the potentially extant soil archive, our current
record of it, and the levels of threat and protection afforded by the current landscape. To this end, the
IKAW must be enhanced with a data layer covering all land- and coastal surfaces, and assessing their
paleogeographic, hydro- and pedological potential for preservation of cultural remains; a similar layer will
be needed to estimate the probability of future degradation through land use (especially the piecemeal
degradation through agricultural land use, forestry, and nature development), urban outlays and
infrastructural works; and finally, procedures for legal protection of archaeological resources are to be
reviewed in a manner similar to MARS and MPP, so that in future the value of a resource can be assessed
using a scoring system based on the underlying IKAW data layers for rarity, preservation, group value,
etc. (Deeben et al. 1999).
2.2
DATA QUALITY
Although I have in the previous pages already indicated that issues of data quality play an important role
in limiting what can be done with predictive models, a more detailed discussion of issues relating to the
quality of the archaeological and environmental input data follows.
THE ARCHAEOLOGICAL RECORD
The ‘official’archaeological record contains only a small subset of both currently known and historically
attested archaeological observations, a fact recognised by most if not all students of archaeological
records. Lang (2000:225) neatly encapsulated the problem when writing about the gap between ‘current
knowledge’and ‘deposited knowledge’. Exactly how much of a gap there is, remains to be determined by
appropriate studies, but the author's experience suggests that as little as one quarter of currently available
knowledge (as measured in numbers of find spots) may have been deposited in the central archaeological
archive of the Netherlands. The same impression is conveyed by Verhagen (2000:232) who suggests that
“the amount of data that has never been published in an accessible form is probably staggeringly large”.
We must therefore ask ourselves by what mechanisms archaeological observations in the past became (or
did not become) part of the official record, and study the potential biases caused by this process (this
thesis, chapter 4). We must further be aware that predictions based on the limited and biased subset of
observations that has become part of deposited knowledge run the risk of being substantially incorrect – an
insupportable situation from the management point of view if not from the academic one. Any regional
predictive model should therefore be preceded by an assessment of the relation of deposited to current
knowledge.
Furthermore, the archaeological record is a historically accreted one, with varying amounts of quality
control applied during the entry process and typically without the metadata required to assess the quality
of the data (cf. Garcia Sanjuan & Wheatley 1999). When the contents of the Dutch ‘paper’archives were
being transferred into a relational database format in the early 1990’s (Roorda & Wiemer 1992), many
records could not be transferred because the quality of the information (especially its geographical,
chronological, and functional resolution) was too low. These latter records have therefore been, to all
intents and purposes, lost to archaeological prediction. Undoubtedly, a large part of the knowledge that is
not part of the formal archaeological record is still part of current individual or institutional knowledge of
professional and amateur archaeologists; but it is also precarious knowledge in that, unrecorded, it will
disappear with the death of its bearer. The use of ‘expert judgement’in predictive archaeological models
either through the Bayesian mechanism of ‘prior probabilities’or for ‘tuning’the results is therefore
unstable for two reasons: firstly the expert judgement cannot be scrutinised because it is unpublished, and
secondly it all depends on which expert’s opinion is being taken.
Lang’s (2000) idea that CRM databases could function as test beds for research hypotheses, pattern
detection, etc, rests on a belief that more data will do the trick; given the many problems with the quality
and representativity of such data this may be doubted. A better understanding of the biases in deposited
5 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
as well as current knowledge is urgently needed (see chapter 4).
PHYSICAL AND COGNITIVE LANDSCAPE PARAMETERS
Problems associated with the quality and appropriateness of parameters of the current physical landscape
have been the subject of abundant and detailed discussion, and need not be repeated here. Among the
less thoroughly reviewed issues, especially relevant to predictive modelling in active geological areas, are
the use of historical and palaeo-geographical reconstructions, and land evaluation based on methods
developed for the world Food and Agricultural Organisation (FAO 1976). In most parts of Europe land
evaluation, which present research suggests may become more widely used as a basis for ‘deductive’
landscape ecological modelling, will require detailed and extensive geopedological fieldwork and historical
research if reasonably accurate landscape reconstructions are to be generated. The latter are again crucial
in the evaluation of the presence, quality, value and sensitivity to threats of archaeological resources for
management policies.
Whilst for some regions and periods, historical sources and documents may provide some evidence of
past cognitive landscapes, this will not be the case for most pre-modern landscapes. The use of cognitive
landscape parameters in predictive models therefore rests in the main on ethnographic analogy, the
simple transfer of modern landscape interpretations, or unsupported ‘narrative’constrained only by the
characteristics of the extant archaeological record. Foremost among the cognitive parameters investigated
are landscape visibility and accessibility (and their converses), which are discussed elsewhere in this thesis
(chapter 6). Recently, the context for their use has been mainly derived from landscape architecture and
has been applied to relatively well-preserved archaeological landscapes such as the ritual landscape of the
southern British chalk downs (see especially the papers in Lock (ed) 2000). Authors such as Lock himself
have stressed the hypothetical and heuristic nature of such reconstructions, and it is not yet to what
degree rule-based approaches to cognitive landscape reconstruction will be able to improve predictive
settlement models.
Both physical and cognitive factors tend to be used in the modelling of static patterns of archaeological
settlement or ritual land use, and only implicitly of the processes that result in these patterns. This is
largely due to the poor temporal resolution of the available archaeological data (see next section) and
tends to hide the fact that the most obvious constraint to the use of the landscape at any particular
moment is its current use – any activity resulting in recognisable archaeological remains is normally
performed within the context of a fully developed, and continually changing, physico-cognitive landscape.
It is the ‘how’and ‘why’of this change, the dynamics, that we ultimately wish to understand. Rather than
the quality of the physical and cognitive map layers themselves, it may therefore be that the quality of the
sociological-behavioural rules governing the actions and reactions of a society will become paramount in
future predictive models, which will then become based on simulations rather than on locational analysis.
One simple example of this could be the application of a ‘splitting threshold’for communities to model
the process by which agricultural colonisation of a region takes place.
SCALE
At all stages of a predictive model, it is necessary to specify aspects of scale. Scale refers both to the spatiotemporal extent of the model and to the resolution of the data used by it. In addition to the relatively
well-known importance of specifying and properly handling cartographic (spatial) scale (e.g., Sydoriak Allen
2000), chronological (temporal) and analytical (functional) scales assume roles of central importance in
archaeological predictive modelling5. Since the past processes being targeted for modelling also occur at a
variety of spatio-temporal scales, the quality of a model cannot be said to depend on high-quality input
data only (the ‘more is better’approach); rather, the scale of the data must be appropriate to the scale of
the problem.
Note that this is not a reference to so-called ‘spatio-temporal’GIS, the name for experimental systems that can handle fourdimensional data (esp. Harris & Lock 1996).
5
5 -
12
W IDE -A R E A P R E D I C T I V E M O D E L I N G
Space, time, and function can be thought of as different axes along which the available data can be
differentiated; and the smaller the scale, the less differentiation is possible. In the realm of cartographic
scale this means that mapped variables are always averaged across an area of space, eliminating variations
that may have archaeological significance. Along the temporal axis a smaller scale means that
archaeological and palaeo-geographical dynamics become averaged over periods of hundreds of years or
'bunched' into a small number of relatively well-recognised periods. Along the functional axis it means
that many different types of human actions and their remains are lumped together to obtain a generic
‘settlement’model. Archaeological predictive modellers have been forced into using small scales because
of the limitations of the available data (see previous section), resulting in models of low gain (Ebert 2000).
Aggregate archaeological units
One instructive and central scale problem in predictive modelling relates to the use of the ‘site’(a pointlike feature on all but the largest-scale maps) as the basic unit of record and analysis. The statistical nature
of much predictive modelling requires that careful thought be given to the archaeological ‘dependent
variable’ being analysed. In essence, the problem consists of deciding which are the appropriate
archaeological units to analyse, and under what circumstances should multiple observations (‘sites’) really
be taken to represent one such unit6. One practical reason for aggregating ‘sites’into larger units might be
the fact that the resolution of most environmental maps is too low anyway to be able to say anything
reliably about a single site (cf. Sydoriak Allen 2000:103).
The Dutch State Service for Archaeological Investigations has attempted to tackle this issue by defining a
new area unit, the Archaeological Resource Area (ARA), which typically includes a settlement and its
‘infields’up to a distance of 200m (Deeben et al. 1997), and in the context of Mediterranean archaeology
it might entail the aggregation of features such as building remains, ceramic scatters, and terracing into a
new unit ‘farm / rural villa’(see also my discussion of regional database design, this thesis, chapter 13).
The aggregation of a number of point-like observations into an areal unit of a variable size takes us into
uncharted methodological waters. For example, the increase in the number of raster cells taken up by the
ARA relative to the area of the ‘sites’it was based on, affects the calculation of statistical measures of
correlation because it produces sets of highly clustered ‘observations’(Van Leusen 1996:190). Other
unanticipated effects on the outcomes of our models are likely to occur as seemingly ‘technical’variables
such as our choice of analysis region, scale, and resolution impinge on the archaeological problem being
analysed. Future predictive models, not only in the area of CRM but in academic usage as well, should
contain safeguards against such fatal mistakes.
2.3
DATA QUANTITY
Although the need to create separate models for each significant chronological and functional subset of
sites was already apparent by the early 1980s (Kohler and Parker 1986), subsequent CRM-oriented models
often disregarded this aspect, and frequent reminders have therefore continued to appear in the literature
(cf. Verhagen 2000:229-232 for a recent example). The reason for this lies in the inavailability of a
sufficient number of observations. Methods for establishing the existence and strength of statistical
correlations invariably require a set minimum number of observations to be made before a given
confidence level can be reached. In many cases where the observations must be selected from a limited
set of archaeological ‘site’records of sufficient quality, this quantitative minimum can not be reached
unless a very low confidence level is accepted.
Although not the subject of discussion here, it may be noted that an object-oriented approach to the construction of regional
archaeological data sets seems the most appropriate in this respect.
6
5 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
2.4
EXTENSIONS
Given the current state of predictive modelling methodology in the Netherlands, it may be claimed that
the approach is not yet sufficiently matured to begin yielding high quality predictions. While data quality
can only be improved with great effort and much time, better methods can be developed and adopted
relatively quickly. The following four extensions are therefore proposed as likely avenues for the further
development of an improved predictive modelling methodology.
Bayesian inference
Van Dalen (1999) experimented with the use of Bayesian inference techniques in models applied to
archaeological data from the Rieti basin (Italy) survey. A formal Bayesian approach has the advantage of
transparency – allowing methodological separation of expert judgement (= prior belief) and observations,
and therefore represents a step towards the formalisation of the 'seat of the pants' archaeological
predictive models typically used in Dutch CRM (e.g., Scholte Lubberink et al. 1994). Verhagen (2001)
recently discussed the potential of this method as well.
Fuzzy Logic
Given the uncertainties inherent in mapped environmental data and, most worryingly, in archaeological
records, predictive models would benefit from a methodology that can deal with uncertainty. Enabling
fuzzy GIS and database types and operations may be one way in which uncertain data can be represented
and analysed. Although fuzzy GIS operations are not entirely unknown in archaeology (e.g., Nackaerts et
al. 1999), Crescioli et al. (2000) only recently introduced the use of fuzzy logic in the database part of a
GIS. Using the public domain PostGreSQL-GRASS combination, they added fuzzy data types and
functions in order to store, query, and display fuzzy age, gender, and chronology attributes for graves and
skeletons of the Pontecagnano cemetery. In as much as this enables operators to store the many uncertain
properties of both archaeological objects and cartographic representations of real-world features, this
development has the potential of clearing the way for a considerably improved practise of predictive
modelling.
Landscape Reconstruction
The potential of land evaluation as a formal method for modelling environmental potential and constraint
has so far been explored in a limited number of case studies only, but can be applied to any early
agriculturalist society for which the physical landscape can be reconstructed to a sufficient degree. Models
based on land evaluation have the further advantage that they are generic (they can be applied to any area
with a similar environment without reference to its archaeology) and falsifiable (they can be tested both
against existing archaeological records and by a straightforward programme of fieldwork); they therefore
offer hope of a more constructive and objective approach to the study of past landscapes than has
hitherto been possible. Dalla Bona’s models of prehistoric non-agricultural occupation of the wooded
Canadian landscape hold out the hope that the logic, if not the method, of land evaluation can be
extended to cover pastoralist/arboricultural/hunting lifestyles as well (see also Kamermans 1993 for
similar work in central Italy). Land evaluation is an important component of the RPC project (Van
Joolen, forthcoming).
However, since land evaluation will often require a large investment in palaeo-geographic reconstruction
(coring programmes, palynological reconstructions), other means of reconstructing past landscapes
deserve more attention as well. Early historic landscapes might, for instance, be reconstructed using
additional sources of historical information deriving from place-name etymology, historic literary and
cartographic sources, etc.
5 -
14
W IDE -A R E A P R E D I C T I V E M O D E L I N G
Spatial Statistics
“Archaeology is an eclectic discipline; where it calls on scientific and statistical knowledge, few individuals
combine all the necessary archaeological, scientific and statistical skills at a high level. Ideally,
collaboration should take place between archaeologists and statisticians in those areas where a statistical
input is of potential interest. Unfortunately, not all archaeologists regard statisticians as useful creatures
and there are, in any case, not enough interested statisticians to go around.”
- Baxter (1994: 219)
This quotation is particularly relevant to the use of GIS by archaeologists, because the visual nature of the
software makes it easy to remain unaware of, or disregard, the essentially quantitative nature of operations
and the biased nature of the available data samples.
Statistical decorrelation
Earlier (Van Leusen 1996:190) I identified the failure of practitioners to deal with the regular occurrence
of strong correlations between the variables typically used in predictive modelling as one of its principal
methodological shortcomings. Yet methods do exist to remedy the statistical problem, albeit usually at the
price of producing a set of decorrelated variables which cannot easily be understood in real-world terms7.
Principal components analysis (PCA) can be used to construct a set of less correlated components from
an original set of potentially strongly correlated cases (Q-mode) or variables (R-mode), followed by Kmeans clustering of components in order to find out if interpretable clusters exist.
Spatial Autocorrelation and Geostatistics
As argued in section 1.2.2 above, statistical tests of significance must be used in a manner appropriate to
the type of data being analysed. In particular, tests assuming independence between observations (cases)
and/or normality of test distributions should be replaced by tests which take into account the degree of
spatial autocorrelation displayed by each variable involved, and which either make no assumptions about
the shape of their distributions at all (eg, Monte Carlo type tests), or make assumptions which can be
shown to be realistic (eg, many distributions may resemble that of the Poisson curve). In spatial
autocorrelation, nearby observations tend to be similar because geographic variables do not change
quickly over short distances. If a set of archaeological observations is spatially clustered (as may result, for
example, from an intensive survey of a small area) their geographical characteristics are likely to be similar,
and could therefore lead to the derivation of overly strong locational 'preferences'. Dealing with this issue,
Kvamme (1993) presented Moran’s I test, which attempts to measure the correlation of a single variable
with itself over space - the distance between any pair of observations being measured, for example, as
Euclidean distance (but any measurement of ‘distance’is acceptable). Moran’s I statistic can be calculated
on the basis of the covariance of the variables under consideration. Using the apparent sample size N and
Moran’s I, a new corrected (lower) sample size may be calculated to which non-spatial ‘critical-point’
tables such as chi2-tables can be applied. In other words, the original cluster of observations is reduced to
a lower number of observations from which more realistic 'preferences' may be inferred.
Geostatistics are a body of theory and methods designed for the analysis of spatially correlated,
geographical variables. Despite the reservations expressed by Barceló and Pallarés in their discussion of
the theory and method of social space (1998:65), that geostatistical methods do not perfectly fit
archaeological purposes because social action and, with it, social space is discrete rather than continuous,
I believe that the construction of geopedological units on the basis of point measurements (corings) and
areal observations (geomorphological units) is sufficiently similar to the construction of meaningful
archaeological entities (eg, site catchment areas and urban manuring zones) on the basis of excavations
This need not be an objection if the model is not intended to further understanding, but rather to maximise predictive power.
Factor analysis, which is an attempt to explain the correlations between observable variables in terms of underlying factors which
themselves are not observable, may be a more appropriate technique if meaningful explanations are desired.
7
5 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
and surveys to warrant a further exploration of the potential of geostatistical models for variables such as
artefact density. Methods such as constrained co-Kriging would even allow the use of correlated variables
such as slope and distance to water, and the modeling of discontinuous variables. The underlying
assumptions (eg, regarding the normal distribution of the variables) should of course be born in mind
when applying geostatistical methods to archaeological data.
3
CONCLUSIONS
Wide-area predictive modelling using GIS is poised to play a very important role in CRM at the national
level in most of Europe because of the imminent implementation of the Valletta Treaty, but at the same
time it has remained an important tool for academic research as well. For both types of users the ability to
generate formal, rule-based, and testable hypotheses in the form of predictive maps is fundamental, and
these require a better understanding of the underlying theory, data and methods. In this article, I have
identified and discussed several issues which are relevant to future wide-area predictive models. This leads
me to the following conclusions:
• research into, and discussion of, predictive modelling has been hampered by a lack of definition of
core notions, e.g. what exactly is a predictive model supposed to predict? How do we decide what is a
‘good’model? Many ‘predictive’models in fact do no more than describe the input sample of
archaeological site data, and none have formal quality criteria that were actually tested. Such
definitions and criteria should be a requisite part of any predictive model;
• there are still major quality problems with current predictive models. They do not yet have sufficient
spatial, functional, and temporal resolution to provide predictions to rival those of experts, they do
not allow for the formal inclusion of archaeological theory and expertise, and they do not formally
incorporate stages of source criticism (bias correction) and quality testing. The surest (and perhaps
even fastest) way to improving predictive modelling of archaeological site distributions is to conduct
properly designed field tests. Expert knowledge must be given a formal place in the process of
predictive modelling, possibly through the use of Bayesian inference.
• correlative predictive modelling rests on statistical procedures for determining presence and type of
patterning in the input data. This has two important consequences. Firstly, improper use of statistical
procedures strikes at the heart of the models; secondly, the ensuing predictions take the form of
probabilistic statements which can easily be misunderstood by end users of predictive maps. It is
therefore imperative that predictive models incorporate safeguards against incorrect use of statistical
inference, and that a clear distinction be made between the predictive model itself, and any derivative
maps that indicate the value and/or need for protection of the archaeological record;
• Most known archaeological patterns are the result of archaeological research bias, whether this is by
the influence of vegetation, the difficulty of detecting buried sites, or specific interests of
archaeologists in certain kinds of sites and artefacts. A phase of source criticism of both
archaeological and environmental input data should therefore be mandatory, and the modelling
methodology should be sensitive to the characteristics of the available data set. Taphonomical maps
that assess the nature and extent of the distortions of the known material heritage should be an
integral part of any predictive model. If this is not done, ‘low potential’zones run the risk of being
regarded as ‘zones of no interest’, whereas they may in fact be zones of insufficient data within the
archaeological record. Predictive maps run the risk of turning into self-fulfilling prophecies if these
zones, because of their ‘low potential’,are not included in subsequent surveys.
• predictive models, and especially the maps they give rise to, already play a significant and useful role
in the cultural resource management (CRM) process, not just because they provide a structured and
5 -
16
W IDE -A R E A P R E D I C T I V E M O D E L I N G
formal archaeological participation in this process for the first time but, at a technical level, because
they have helped shift attention from site-based to zone-based conservation. However, such models
have barely touched the question of how to model quality, rarity, nature, and indeed ‘value’of
archaeological remains (Deeben et al. 1999). Incorporating these factors will bring predictive models
closer to true expert systems and must be regarded as the next major stage in the development of
geographical models for archaeological resource management.
REFERENCES
Altschul, JH 1990
Red flag models: the use of modelling in management contexts, in Allen, KMS, SW Green & EBW Zubrow
(eds), Interpreting Space: Geographical Information Systems and Archaeology: 226-38.
Ankum, LA & BJ Groenewoudt 1990
De situering van archeologische vindplaatsen. RAAP-rapport 42. Amsterdam: Stichting RAAP.
Anonymous 1998
MARS: The Monuments at Risk Survey of England, 1995. Summary Report. Bournemouth/London:
Bournemouth University and English Heritage.
Barceló, JA & M Pallarés 1998
Beyond GIS: The archaeology of social spaces, in Archeologia a Calcolatori 9:47-80.
Baxter, MJ 1994
Exploratory Multivariate Analysis in Archaeology. Edinburgh: Edinburgh UP.
Brandt, R, BJ Groenewoudt & KL Kvamme 1992
An experiment in archaeological site location: modelling in the Netherlands using GIS techniques. World
Archaeology 2: 268-282.
Carr, C 1985
Introductory remarks on Regional Analysis. In: C. Carr (ed), For Concordance in Archaeological Analysis.
Bridging Data Structure, Quantitative Technique, and Theory. Westport Publishers, Kansas City, pp. 114127.
Chadwick, AJ 1978
A computer simulation of Mycenaean settlement, in Hodder, I (ed), Simulation Studies in Archaeology, 4757. Cambridge: Cambridge UP.
Chadwick, AJ 1979
Settlement simulation, in Renfrew, C & D Cooke (eds), Transformations: Mathematical Approaches to
Culture Change, 237-255. New York: Academic Press.
Church, T, RJ Brandon & GR Burgett 2000
GIS Applications in Archaeology: Method in Search of Theory, in Wescott, KL & RJ Brandon (eds)
2000:135-155.
Cliff, AD & JK Ord 1981
Spatial processes : models & applications. London: Pion.
Crescioli, M, A D’Andrea & F Niccolucci 2000
A GIS-based analysis of the Etruscan cemetery of Pontecagnano using fuzzy logic, in Lock, G (ed)
2000:157-79.
Dalla Bona, L 1993
A preliminary predictive model of prehistoric activity location for the western Lake Nipigon watershed,
Archaeological Computing Newsletter 37:11-19.
Dalla Bona, L 1994
Archaeological Predictive Modelling Project, Ontario Ministry of Natural Resources. Centre for
Archaeological Resource Prediction, Lakehead University, Thunder Bay.
Dalla Bona, L 2000
Protecting Cultural Resources through Forest Management Planning in Ontario Using Archaeological
Predictive Modelling, in Wescott, KL & RJ Brandon (eds) 2000:73-99.
Dalla Bona, L & L Larcombe 1996
Modelling Prehistoric Land use in northern Ontario, in HD Maschner ed., New Methods, Old Problems.
Geographic Information Systems in Modern Archaeological Research (Southern Illinois University Centre
for Archaeological Investigations Occasional Paper 23): 252-271.
Darvill, T & K Fulton 1998
MARS: the Monuments at Risk Survey of England. Bournemouth/London: School of Conservation Sciences,
Bournemouth University / English Heritage.
5 -
17
V A N L EUSEN : P A T T E R N T O P ROCESS
Darvill, TC, AD Saunders & DW Startin 1987
A question of national importance: approaches to the evaluation of ancient monuments for the Monuments
Protection Programme in England, Antiquity 61:393-408.
Darvill, TC & G Wainwright n.d.
The MARS Project: An Introduction, at http://csweb.bournemouth.ac.uk/consci/text_mars/intro.htm.
Deeben, J, BJ Groenewoudt, DP Hallewas & WJH Willems 1999
Proposals for a practical system of significance evaluation in archaeological heritage management,
European Journal of Archaeology 2(2):177-99.
Deeben, J, DP Hallewas, J Kolen & R Wiemer 1997
Beyond the crystal ball: predictive modelling as a tool in archaeological heritage management and
occupation history. In: Willems, W, H Kars & D Hallewas (eds), Archaeological Heritage Management in
the Netherlands. Fifty Years State Service for Archaeological Investigations. ROB, Amersfoort, pp. 76-118.
Deeben, J, DP Hallewas, J, & ThJ Maarlevelt, N.D. (2001)
Predictive modelling in archaeological heritage management of the Netherlands: the indicative map of
archaeological values (2nd generation). Unpublished manuscript. Amersfoort: ROB.
Deeben, J & R Wiemer 1999
Het onbekende voorspeld: de ontwikkeling van een indicatieve kaart van archeologische waarden. In:
Willems, W (ed), Nieuwe ontwikkelingen in de Archeologische Monumentenzorg. Nederlandse
Archeologische Rapporten 20:29-42. Rijksdienst voor het Oudheidkundig Bodemonderzoek, Amersfoort.
Ebert, J 2000
The State of the Art in “Inductive” Predictive Modelling: Seven Big Mistakes (and Lots of Smaller Ones),
in Wescott, KL & RJ Brandon (eds) 2000:129-134.
English Heritage 1996
The Monuments Protection Programme 1986-96 in retrospect. English Heritage leaflet.
FAO 1976
A framework for land evaluation, ILRI Publication 22. Wageningen.
Gaffney, VL & PM van Leusen 1995
GIS and environmental determinism, pp. 367-82 in Lock, G & Z Stancic (eds), GIS and Archaeology: a
European Perspective. London: Francis & Taylor.
García Sanjuán, L & DW Wheatley 1999
The state of the Arc: differential rates of adoption of GIS for European heritage management, European
Journal of Archaeology 2(2):201-28.
Gibson, T 1997
Forestry Impacts. Paper presented at the Predictive Modelling Thinktank, Sault Ste-Marie (Canada),
February 1997.
Gould, RA & MB Schiffer (eds) 1981
Modern Material Culture: the Archaeology of Us. New York: Academic Press.
Groenewoudt, BJ 1994
Prospectie, waardering en selectie van archeologische vindplaatsen. Een beleidsgerichte verkenning van
middelen en mogelijkheden. Nederlandse Archeologische Rapporten 17. Amersfoort: Rijksdienst voor het
Oudheidkundig Bodemonderzoek.
Groenewoudt, BJ & JHF Bloemers 1997
Dealing with significance: Concepts, Strategies and Priorities for Archaeological Heritage Management in
the Netherlands, in: WJH Willems et al. (eds), Archaeological Heritage Management in the Netherlands.
Fifty Years State service for Archaeological Investigations, pp 119-72. Assen/Amersfoort: Van Gorcum.
Groenewoudt, BJ, DP Hallewas & PAM Zoetbrood 1994
De degradatie van de archeologische betekenis van de Nederlandse bodem. (Interne Rapporten ROB 8).
Amersfoort: Rijksdienst voor het Oudheidkundig Bodemonderzoek.
Harris, TM & GR Lock 1996
Multi-dimensional GIS: exploratory approaches to spatial and temporal relationships within archaeological
stratigraphy, Analecta Praehistorica Leidensia 28:307-316.
Hinchcliffe, J & RT Schadla-Hall (eds) 1980
The Past under the Plough: papers presented at the Seminar on Plough Damage and Archaeology held at
Salisbury February 1977. London: Department of the Environment.
IKAW 2 2000
Indicatieve Kaart van Archeologische Waarden (IKAW), 2e generatie. Amersfoort: ROB.
Judge, WJ & L Sebastian 1988
Quantifying the Present and Predicting the Past: Theory, method, and application of archaeological
5 -
18
W IDE -A R E A P R E D I C T I V E M O D E L I N G
predictive modelling. Denver: US Department of the Interior.
Kamermans, H 1993.
Archeologie en landevaluatie in de Agro Pontino (Lazio, Italië). Academisch proefschrift. Amsterdam.
Kamermans, H 2000
Land evaluation as predictive modelling: a deductive approach. In: Lock, G (ed), Beyond the Map.
Archaeology and Spatial Technologies. NATO Sciences Series. IOS Press. Amsterdam, pp. 124-146.
Kamermans, H 2001
Strategic research into, and development of best practise for, predictive modelling on behalf of Dutch
Cultural Resource Management. Unpublished research proposal.
Kamermans, H, SH Loving & A Voorrips 1985
Changing patterns of prehistoric land use in the Agro Pontino, in Malone, C & S Stoddart (eds), Papers in
Italian Archaeology IV. Part i: The Human Landscape. BAR International Series 243:53-68.
Kamermans, H & E Rensink 1998
GIS in Palaeolithic Archaeology. A case study from the southern Netherlands. In Dingwall, L et al. (eds),
Archaeology in the Age of the Internet. Computer Applications and Quantitative Methods in Archaeology.
BAR International Series 750: 81 and CD-ROM.
Kamermans, H & M Wansleeben 1999
Predictive modelling in Dutch archaeology, joining forces. In: JA Barceló, I Briz & A Vila (eds), New
Techniques for Old Times - CAA98. Computer Applications and Quantitative Methods in Archaeology.
BAR International Series 757: 225-230.
Kohler, TA & SC Parker 1986
Predictive models for archaeological resource location. In: Schiffer, MB (ed), Advances in Archaeological
Method and Theory 9, Academic Press, New York, pp. 397-452.
Krist, FJ 2001
A Predictive Model of Paleo-Indian Subsistence and Settlement. PhD dissertation, Michigan State
University.
Kuna, M 2000
Comments on archaeological prediction, in Lock, G (ed) 2000: 180-185.
Kvamme, KL 1983
A Manual for Predictive Site Location Models: Examples from the Grand Junction District, Colorado, Report
to the U.S. Bureau of Land Management, Grand Junction District, Colorado.
Kvamme, KL 1985
Determining empirical relationships between the natural environment and prehistoric site location: a huntergatherer example. In: Carr, C (ed), For Concordance in Archaeological Analysis. Bridging Data Structure,
Quantitative Technique, and Theory. Westport Publishers, Kansas City, pp. 208-238.
Kvamme, K 1990
The fundamental principles and practice of predictive archaeological modelling. In: Voorrips, A (ed),
Mathematics and Information Science in Archaeology, Volume 3. Holos Verlag, Bonn, pp. 257-295.
Kvamme, KL 1993
Spatial Statistics and GIS: an integrated approach, pp 91-104 in: Andresen, J, T Madsen & I Scollar (eds),
Computer Applications and Quantitative Methods in Archaeology 1992. Aarhus University Press, Aarhus.
Kvamme, KL 1996
Randomization methods for statistical inference in raster GIS context, in Bietti et al. (eds), The Colloquia of
the XIII International Congress of Prehistoric and Protohistoric Sciences, Vol 1: Theoretical and
Methodological Problems: 107-114. Forli: ABACO.
Kvamme, KL 1999
Recent Directions and Developments in Geographical Information Systems. Journal of Archaeological
Research 7(2):153-201.
Lang, N 2000
Beyond the map: harmonising research and Cultural resource Management, in Lock, G (ed) 2000: 214-227.
Larsen, CU (ed) 1992
Sites & Monuments: national archaeological records. Copenhagen: National Museum of Denmark.
Lock, G (ed) 2000
Beyond the Map. Archaeology and Spatial Technologies. NATO Science Series A: Life Sciences, Vol 321.
Amsterdam etc: IOS Press.
Nackaerts, K, G Govers & L Loots 1999
The Use of Monte Carlo Techniques for the Estimation of Visibility. In Dingwall, L et al. (eds),
Archaeology in the Age of the Internet. Computer Applications and Quantitative Methods in Archaeology.
5 -
19
V A N L EUSEN : P A T T E R N T O P ROCESS
BAR International Series 750: 63-5.
Parker, S 1985
Predictive modelling of site settlement systems using multivariate logistics. In: Carr, C (ed), For
Concordance in Archaeological Analysis. Bridging Data Structure, Quantitative Technique, and Theory.
Westport Publishers, Kansas City, pp. 173-207.
Pilon, J-L, L Nolin & K Swayze 1997
Traditional Knowledge and Archaeological Understanding: An example from the lower Mackenzie Valley,
paper presented at the Predictive Modelling Thinktank, Sault Ste-Marie (Canada), February 1997.
Renfrew, C & E Zubrow 1994
The ancient mind: elements of cognitive archaeology (New directions in archaeology). Cambridge:
Cambridge UP.
Roorda, I & R Wiemer 1992
The ARCHIS Project: Towards a New National Archaeological Record in the Netherlands, in Larsen, CU
(ed) 2000: 117-122.
Savage, SH 1990
GIS in archaeological research. In: Allen, KMS, SW Green & EBW Zubrow (eds), Interpreting Space: GIS
and archaeology. London: Taylor & Francis, pp. 22-32.
Scollar, I 1999
Twenty Five Years of Computer Applications to Archaeology, in Dingwall et al. (eds) 1999:5-10.
Schiffer, MB (ed) 1986
Advances in Archaeological Method and Theory 9. New York: Academic Press.
Scholte Lubberink, HBG, JWHP. Verhagen & H van Londen 1994
Archeologisch onderzoek ten behoeve van de trajectstudie / m.e.r. Rijksweg 4: Kruithuisweg (Delft) –
Kethelplein (Schiedam). RAAP-rapport 94. Stichting RAAP, Amsterdam.
Sebastian, L & WJ Judge 1988
Predicting the Past: correlation, explanation, and the use of archaeological models. In Judge, WJ & L
Sebastian (eds), Quantifying the Present and Predicting the Past: Theory, Method, and Application of
Archaeological Predictive Modelling. Denver: Bureau of Land Management, 1-18.
Stancic, Z & KL Kvamme 1999
Settlement Pattern Modelling through Boolean Overlays of Social and Environmental Variables. In Barceló,
JA, I Briz & A Vila (eds), New Techniques for Old Times - CAA98. Computer Applications and
Quantitative Methods in Archaeology. BAR International Series 757: 231-237.
Stancic, Z & T Veljanovski 2000
Understanding Roman settlement patterns through multivariate statistics and predictive modelling, in Lock,
G (ed) 2000:147-156.
Startin, DW 1992
The Monuments Protection Programme: archaeological records. In Larsen, CU (ed) 1992: 201-206.
Sydoriak Allen, K M 2000
Considerations of Scale in Modelling Settlement Patterns Using GIS: An Iroquois Example, in Wescott, KL
& RJ Brandon (eds) 2000:101-112.
Van Dalen, J 1999
Probability Modelling: a Bayesian and a Geometric Example. In Gillings, M et al. (eds), Geographical
Information Systems and Landscape Archaeology (The Archaeology of Mediterranean Landscapes 3), pp
117-124. Oxford: Oxbow Books.
Van Joolen, E forthcoming
Reconstruction of ancient landscapes, agriculture and vegetation history in Central and South Italy during
the first millennium BC. Ph.D. thesis, University of Groningen.
Van Leusen, PM 1993
Cartographic Modelling in a cell-based GIS, pp 103-122 in Andresen, J, T Madsen & I Scollar (eds),
Computer Applications and Quantitative Methods in Archaeology 1992. Aarhus University Press, Aarhus.
Van Leusen, PM 1995
GIS and Archaeological Resource Management: A European Agenda. In: Lock, G & Z Stancic (eds),
Archaeology and Geographical Information Systems. Taylor & Francis, London, pp. 27-41.
Van Leusen, PM 1996
Locational Modelling in Dutch Archaeology. In: Maschner, HDG (ed), New Methods, Old Problems:
Geographic Information Systems in Modern Archaeological Research. Occasional Paper no. 23. Centre for
Archaeological Investigations, Southern Illinois University, pp. 177-197.
Verhagen, Ph 1995
5 -
20
W IDE -A R E A P R E D I C T I V E M O D E L I N G
De archeologische potentiekaart in Nederland: een methodologie voor het voorspellen van archeologische
waarden op basis van archeologische en landschappelijke gegevens. Westerheem 44(5): 177-187.
Verhagen, Ph 2000
Archaeology, GIS and Cultural Resource Management, in Lock, G (ed) 2000: 229-235.
Verhagen, Ph 2001
Quantifying the Qualified: The Use of Multi-Criteria Methods and Bayesian Statistics for the Development
of Archaeological Predictive Models. Paper presented at the conference “GIS and Archaeological
Predictive Modelling: Larger-Scale Approaches to Establish a Baseline For Site Location Models” Argonne
(IL, USA) 21-24 March 2001.
Verhagen, Ph & JF Berger 2001
The Hidden Reserve: Predictive Modelling of Buried Archaeological Sites in the Tricastin-Valdaine Region
(Middle Rhone Valley, France). In: Stancic, Z & T Veljanovski (eds), Computing Archaeology for
Understanding the Past. CAA2000. Computer Applications and Quantitative Methods in Archaeology. BAR
International Series 931: 219-231.
Verhagen, Ph, M Wansleeben & M van Leusen 2000
Predictive Modelling in the Netherlands. The prediction of archaeological values in Cultural Resource
Management and academic research. In: Harl, O (ed), Archäologie und Computer 1999.
Forschungsgeselschaft Wiener Stadtarchäeologie 4: 66-82.
Wansleeben, M & LBM Verhart 1992
The Meuse Valley Project: GIS and site location statistics. Analecta Praehistorica Leidensia 25: 99-108.
Wansleeben, M & LBM Verhart 1997
Geographical Information Systems. Methodical progress and theoretical decline? Archaeological Dialogues
4.1: 53-70.
Wansleeben, M & LBM Verhart 1998
Graphical analysis of regional archaeological data. The use of site typology to explore the Dutch
Neolithization
process.
Internet
Archaeology
4,
http://interarch.york.ac.uk/journal/issue4/wansleeben/index.html
Warren, RE 1990
Predictive Modelling in Archaeology: A Primer, pp. 90-111 in Allen, KMS, SW Green & EBW Zubrow
(eds), Interpreting Space: GIS and Archaeology, London: Taylor & Francis.
Warren, RE & D Asch 2000
A Predictive Model of Archaeological Site Location in the Eastern Prairie Peninsula, in Wescott, KL & RJ
Brandon (eds) 2000:5-32.
Wescott, KL & RJ Brandon (eds) 2000
Practical applications of GIS for archaeologists. A predictive modelling kit. London: Taylor & Francis.
Wheatley, D 1996
Between the lines: the role of GIS-based predictive modelling in the interpretation of extensive survey data.
In: Kamermans, H & K Fennema (eds). Interfacing the Past. Computer applications and quantitative
methods in Archaeology CAA95. Analecta Praehistorica Leidensia 28: 275-292.
Whitley, TG 2000
Dynamical Systems Modelling in Archaeology: A GIS Approach to Site Selection Processes in the Greater
Yellowstone Region. Unpublished Dissertation, Department of Anthropology, University of Pittsburgh, PA.
Whitley, TG 2001
Using GIS to Model Potential Site Areas at the Charleston Naval Weapons Station, South Carolina: An
Alternative Approach to Inferential Predictive Modelling. Paper presented at the conference “GIS and
Archaeological Predictive Modelling: Larger-Scale Approaches to Establish a Baseline For Site Location
Models”. Argonne (IL, USA), 21-24 March 2001.
5 -
21
C HAPTER 6
LINE-OF-SIGHT AND COST
S U R FAC E A N A L Y S I S U S I N G G I S ∗
1
INTRODUCTION
This review article assesses the majority of accessible archaeological studies based on two GIS techniques
(viewshed analysis and cost surface analysis) over the decade 1990-2000. The two techniques are taken
together because of certain similarities in methodology and underlying theoretical principles, which
express an emphasis on the human experience of being and moving in the landscape; not surprisingly,
these techniques have been at the centre of processual – postprocessual debate almost from the
beginning.
1.1
AIMS
This paper reviews published work in two related areas of GIS application in archaeology – line-of-sight
analysis (LOSA) and cost surface analysis (CSA). Line-of-sight analysis uses the ability of most GIS to
calculate the intervisibility of two given points on a given digital elevation model; cost surface analysis
uses cost accumulation algorithms to calculate the cumulative cost of travelling over a digital cost
landscape. These two techniques have received much attention in recent years, first because they were
seen to be implementations of well-established ‘processualist’analytical procedures, and latterly because
of their supposed potential to escape from naïve quantitative processualism into ‘enriched’qualitative
post-processualist (post-structuralist) types of analysis. Because it provides the context for much current
work, this debate is summarised here in section 1.2. However, as I hope to show in this article, the
theoretical affiliations and rhetoric of the various researchers appear to have little influence on the
practicalities of implementing GIS-based LOSA and CSA; rather, it is the type of question that is being
asked that determines methodological possibilities and constraints.
The larger part of this article (sections 2 and 3) is therefore concerned with reviewing procedural aspects of
CSA and LOSA studies.
The field of CSA and LOSA has reached a stage where stocktaking has become useful (see, for example,
Van Leusen 1999, Witcher 1999, Wheatley & Gillings 2000, in press). A subsidiary aim of this review has
been to provide the reader with a starting point for locating case studies and methodological studies
relevant to their own research – hence the extended bibliography section.
∗
An earlier version of this article, dealing with developments up to about 1998, was published in the proceedings of the 1998
Annual Meeting of CAA (Barceló et al. 1999), but this has been substantially revised and updated for the current version. In
conjunction with this article, two other chapters in my thesis (15 and 16) present case studies conducted over the years in order
to investigate aspects of ‘dominance’,territory, and accessibility arising from archaeological thinking about the role of Late Iron
Age hillforts and markets in the Wroxeter hinterland (UK), of early Roman colonies on the Lepine Margin (Pontine Region,
central Italy), and of Middle Bronze Age to Early Iron Age settlements in the Sibaritide (southern Italy).
6 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
Much of the work discussed in this review has only comparatively recently become accessible in the form
of published proceedings (Dingwall et al. 1999, Barceló et al. 1999, Gillings et al. 1999, Lock 2000,
Stancic & Veljanovski 2001) which, however, tend to give voice to a specifically Anglophone community.
Although I have tried to alleviate this bias by also including examples drawn from work done elsewhere,
I cannot claim to have succeeded in this. I hope that my emphasis on the methodology rather than the
content of the examples has palliated the effects of this failing. Furthermore, I have taken care to include
the work and views of non-archaeologists (geographers and landscape planners in particular), as well as
materials that are only available on CD-ROM (Dingwall et al. 1999, Johnson & North 1997).
1.2
THEORETICAL CONT EXT
Following an initial period in which an increasing number of archaeologists experimented with the use of
GIS to grapple with a variety of questions, many critics argued that the naive use of GIS has led to a
revival of environmental determinism (an issue discussed more fully in Gaffney & Van Leusen 1995), and
have advocated a post-processualist approach to using GIS. The root of the problem was seen to lie in
the geographic approaches from which GIS were built, in which space was treated as an abstract
geographical concept (‘Cartesian space’). As Llobera (1996) puts it, there is no observer, no perspective,
and no history in this kind of space. The alternative, post-processual concept of space has, by contrast,
been ‘humanised’:space derives its meaning and properties from the presence of observers.
Others think however that despite appearances, GIS can be used in various ways for the modelling of
cognitive landscapes (Wheatley 1993, Taylor & Johnston 1995). Attempts to address the perceived
rigidity of current GIS applications include the incorporation of concepts of uncertainty (Gillings 1996,
1998, Loots et al. 1999, Nackaerts et al. 1999), the ideal organisation of space and society (Zubrow 1994),
time and change, and of affordances (Llobera 1996). This is signalled as an important current
development in the GIS Guide to Good Practice (Gillings & Wise 1998), but it is not yet clear what, if any,
improvements these approaches will bring. This issue is discussed in more depth in section 3.3.
What makes us think GIS can be used in reconstructing past landscapes? The landscape, both in the past
and in the present, is structured by the fact that resources are distributed unequally over it. This applies to
both natural and social resources - drinking water and infrastructure are only available in some places;
good farming and stock rearing land is not available everywhere or is already occupied by others; centres
of political power, civic administration, and ritual significance are few and far between. People’s choices
both structure this ‘resource landscape’and are structured by it, and we therefore expect archaeological
remains to exhibit structuring of this type. Viewshed and cost surface analysis are two ways to reveal
such structuring.
The latter has aroused widespread scepticism about the usefulness of GIS in archaeological research as,
for instance, at the 1995 meeting of the UISPP (Bietti et al. 1996; Johnson & North 1997; reviewed by
Bampton 1997). A balanced outsider view of the issue can be found in Taylor and Johnston (1995), who
placed then current uses of GIS in the context of the ‘quantitative revolution’and the ‘New Geography’
that took place in the 50s and 60s. These authors provide a useful and provoking discussion of the
dangers of much current data-led GIS use but also stress the potential – mainly in pattern analysis (see
also Gaffney & Van Leusen 1995).
Whilst American researchers have tended to continue in this processualist tradition, much recent
European work, largely driven by a British post-structuralist school, stresses humanistic (as opposed to
abstract, Cartesian) forms of spatial reasoning. The latter rely heavily on LOSA and, to a lesser extent, on
CSA for modelling past perceptions of the natural and human environment. Hermeneutics (the art of
interpretation) has been put forward as offering a theoretical basis for viewshed analysis in particular
(Lock et al. 1999:61). Thus, philosophical positions appear to be irreconcilable at the moment. On the
other hand, it is not clear that these positions result in substantially different approaches to LOSA and
CSA. Post-processual contributions may be replete with references to Bender’s (1993) edited volume on
landscape perspectives and Tilley’s Interpretative Archaeology (1993) and Phenomenology of Landscape
6 -
2
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
(1994); talk may be of ‘perception’and ‘meaning’rather than of ‘viewsheds’and ‘patterns’, of ‘the
hermeneutic spiral’rather than of ‘exploratory data analysis’; but there the differences end. As Tschan et
al. (2000:33) remark, many recent theoretical ‘advances’are entirely devoid of any current methodologies
or even of the potential of such.
Rather than classifying LOSA- and CSA- based studies on the basis of –isms, a more fruitful approach
would be to look at the subject matter (type and amount of archaeological data, geographical and
temporal scale) and research aims. Accordingly, two types can be distinguished:
• At a relatively local scale, students want to explore cognitive space around single monuments or
synchronic/diachronic systems of monuments. This type of study often has a dynamic component,
either in space (movement) or in time (creation of monumental landscapes). The objective is to
explore what is unique about the situation.
• At regional scales, students are intent on using social/cognitive variables in order to build models
that allow us to detect and explain observed patterns site locations and attributes. Here, the objective
is to explore similarities between situations, and models are typically static and quantitative in nature.
The former, qualitative, approach seems most useful in situations where high quality archaeological data
are available, whereas the latter represents more of a continuation of the quantitative deterministic
approach - albeit with the use of an enlarged set of variables to play with. In this regard it is noteworthy
that qualitative GIS models are nearly all based on well-studied landscapes replete with monuments and
symbolic meanings.
The aim of the current review of methods is to allow us to turn our attention to the more fruitful task of
answering archaeological questions. But there is also a second reason for reviewing current archaeological
applications of viewshed and cost surface analyses - archaeological arguments that are ultimately, if only
partly, based on their outcome become invalid if they have been improperly applied or if the results have
been wrongly interpreted.
1.3
CSA AND LOSA: TW IN TOOLS FOR COGNITI VE LANDSCAPE ANALYSI S
The reader might wonder why the two separate techniques are discussed together. Although they are
superficially different, viewshed and cost surface analyses are intimately related techniques because they
both define aspects of the social space surrounding an observer. Take, for example, territorial markers.
These must be highly visible (though not indiscriminately so, as we shall see below) and also be located at
the edge of some kind of ‘territory’. The former is modelled through LOSA, the latter through CSA; the
two cannot be separated.
A more abstract way of looking at the relation between the two techniques would be to note that both are
based on the notions of focus (in the sense of a point-like location which might be the current or intended
location of the protagonist, or might have a significant level of visibility or accessibility) and direction (as in
megalithic alignments pointing at midsummer sunrise points, or in the travel networks discussed below).
As was recently pointed out by Wheatley and Gillings (2000:4), post-structuralist theory provides another
argument for linking the two techniques, in that perception is meaningful only to mobile observers.
Such interrelations are expressed, consciously or unconsciously, in studies that combine both types of
analysis in order to construct an archaeological argument. See, for example, the work of Gaffney et al.
(1996a, 1996b:37-8), where cost surface derived catchments are compared with viewsheds of Iron Age
hillforts on the Dalmatian islands; the suggestion in Belcher et al. (1999:98-100) that Archaic tombs
around the urban settlement of Nepi in Tuscany are situated in areas difficult to reach while easy to see;
Madry and Rakos’(1996:123) use of the visibility of Roman roads from hillforts in Burgundy as an
6 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
input variable for cost surface analysis; and Llobera’s (2000:72-5) use of viewsheds to define ‘attractive’
and ‘repellent’factors influencing movement near certain types of monuments.
2
2.1
COST SURFACE ANALY S I S
PRINCIPLES AND A PPLICATIONS
‘Cost surface analysis’is here used as the generic name for a series of GIS techniques based on the
ability to assign a cost to each cell in a raster map, and to accumulate these costs by travelling over the
map. Early examples were published by archaeologists working with the Arkansas Archaeological Survey
and the US National Parks Service (Limp 1989, 1990). Cost surface analysis is rooted in traditional site
catchment analysis, introduced to archaeology by Vita-Finzi and Higgs (1970), who wanted to study the
economic basis of prehistoric life by looking at optimal foraging models of resources available within a
catchment area or territory associated with a settlement1.
The first step in site catchment analysis is to derive a territory (catchment) belonging to a given focus
(site) by applying some geographical rule. In its simplest form this would be a distance rule, resulting in
circular catchments with a typical radius of 5 or 10 km. The second step is to analyse the properties of
the catchment area, usually to see what economic benefits (e.g., agricultural yields) would accrue to the
focus. The radius would be chosen by experimenting with actual travel times, or by reference to
ethnographic data (Chisholm 1968). Deriving a circular catchment area within a GIS is a trivial operation,
and reporting and tabulating the presence of variables within each catchment can be automated. Recent
examples can be found in Saile (1997) and Lock and Harris (1996:234-8), who used such buffers to
model areas of in- and outfield agriculture around Iron Age Danebury hillfort. A more sophisticated
form of such analysis is the construction of distance buffers around the focus, allowing a statistical
analysis to see if some archaeological correlate gravitates significantly toward or away from it. One wellknown example of this is Hodder and Orton’s (1976) calculation of the distance distribution between
coins and Roman roads in southern Britain2; a very similar application with an interesting twist in the tail
was presented recently by Rajala et al. (1999), who interpret the correlation between site locations and
distance to Roman roads as indicating discovery biases.
Closely related to catchments are tessellations, which have a more specific and theory-laden meaning.
Whereas catchments are generally used to describe economic characteristics of the archaeological
landscape, tessellations of archaeological landscapes are used to postulate a social (political,
administrative, religious) structuring – for example, in Renfrew’s (1986) peer polities. The most widely
known traditional method for doing this is the calculation of Thiessen (Voronoi, Dirichlet) polygons.
These are based on a simple nearest-neighbour method for partitioning a featureless space, equivalent to
a gravity model operating in ‘Flatland’, and result in a complete tessellation of space. Both traditional
catchments and tessellations rely on the simplifying assumption that the landscape is a flat, twodimensional space, and resistance to movement across it is isotropic (the same in all directions). They
also result in choroplethe maps rather than mapping the continuous fall-off of variables such as
accessibility and control. In a real landscape the size and shape of a catchment area or territory would be
much more variable, depending on the nature of the terrain, the topography, and a host of other factors.
In a real landscape, the economic use of, and social control over, an area becomes less with distance
rather than suddenly switching from yes to no (1 to 0) as the boundary of the catchment or territory is
crossed.
Cost surface analysis provides a way out of this by allowing the simple ‘flat’geographical space to be
supplanted by a set of complex cost surfaces incorporating many relevant properties of the terrain. It also
allows for the distance- and gravity based rules for defining the catchment or territory boundaries to be
replaced by a time- or energy expenditure based rule for accumulating costs. As the resulting cumulative
cost surface is a continuous raster map, any number of values may subsequently be used to provide
6 -
4
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
‘cut-off points’or boundaries to the catchment or territory. Alternatively, cost accumulation starting at
multiple points may be allowed to run on until all available space has been used, in which case a
tessellation of space similar to Voronoi tessellation has resulted. For example, Verhagen et al. (1999)
calculate cumulative travel time in order to construct ‘accessibility catchments’which are then used as an
input variable in a predictive settlement model. Stancic (1994) Stancic et al. (1995) compare and contrast
all three methods (Thiessen polygons, circular catchments, and cost-derived catchments) using
protohistoric settlement data from the Dolenjska region in Slovenia. As we shall see in the next section,
the possibilities offered by this technique have led to some confusion as to the best way of calculating
costs.
2.2
ALGORITHMIC CONF USION
Employing a simple radius to define a catchment area is equivalent to travelling over a flat cost surface;
accumulation in this special case is constant in all directions and the maximum horizontal distance
therefore defines the boundaries of the catchment area. If accumulation from a number of starting points
is allowed to continue until the entire cost surface is covered, a Voronoi tessellation results. Just as the
traditional method can be modified to employ travel time as a limiting factor, so cost surfaces can be
modified to reflect the difficulty of travelling over various types of terrain. Accumulating such costs will
result in irregularly shaped catchments for any particular total energy expenditure. This principle can be
extended so that any combination of factors can be used to define costs, and any combination of criteria
can be used to derive a cumulative cost surface from those costs. Exactly how all this should be
implemented is a question that seems to have been answered differently by each individual author. In the
published research there is a wide variety in the parameters used to calculate cost/energy surfaces and in
the algorithms used to perform cost accumulation - a sign of the immaturity of the field.
A further refinement of the technique, originally discussed by Renfrew and Level (1979) as ‘XTENT
modelling’, would result from assigning differential weights to the sites or foci of the catchments, so that
accumulation proceeds with different degrees of ease over any particular cost surface. Ruggles and
Church (1996) first applied it in a GIS context in a weighted Thiessen tessellation of their Mexican study
area, but no implementations using CSA have been published as yet.
Most studies have relied exclusively - and continue to do so - on slope as the factor determining cost
(e.g., Gaffney and Stancic 1991, Gaffney et al. 1993, Massagrande 1996, 1999, Bell & Lock 2000). This
may work in areas where topography has an overriding effect on human behaviour; see, for instance,
Huckerby’s (1999) study of how well four rival foraging theories fit with the costs of accessing
mammalian resources in Queensland, Australia. But more realistic calculations, based on physiological
measurements of energy expenditure on different types of terrain, are now feasible and have recently
been employed in several studies (see below). A recent comprehensive review of the literature on all
forms of walking is contained in a volume edited by Rose and Gamble (1994).
Some authors have attempted to derive costs inductively, from archaeological observations relating to
actual territorial boundaries or actual distances travelled per time slice. One example of this is the work
of Glass et al. (1999) and Anderson and Gillam (2000), who derive costs from observed dates of first
occupation of North and South American sites, and assume that the ‘delays’between colonisation of
successive areas are caused by the cost of travelling from one to the next. Apparently no universal set of
absolute real-world travel costs is to be found in the literature, but this need not be a problem so long as
a universal set of relative costs can be found.
Travel cost surfaces can be isotropic (the same in all directions) or anisotropic3. Because of the effect of
both slope and terrain, the cost of traversing a particular location may differ depending on which
direction it is being crossed in. Crossing cells representing a river is an obvious example of terrain
anisotropy - travelling down-river in a boat incurs different costs from travelling up-river, and different
costs again when crossing the river. Surprisingly, until very recently raster GIS did not provide the
functionality to introduce anisotropy; a closer merging with the functionality generally present in
6 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
vector GIS seems needed. Examples of models based exclusively on isotropic cost surfaces can be found
in Savage (1990); Rajala (1998) mapped territories in the Ager Faliscus using an isotropic cost surface
derived from slope and based on empirical walking effort data; finally, Bell and Lock (2000:88-9) derive
the relative isotropic slope-related cost from the ratio between its tangent and the tangent of 1°, assigning
the latter a cost of 1. Thus, the cost of ascending or descending any slope ? is determined by the formula
tan(?) / tan(1°)
which produces a non-linear relation between slope and cost (also visible in the downslopes of figure 1),
becoming significant on slopes steeper than about 25°.
However, most authors agree that travel cost has both an isotropic and an anisotropic component; the
former exemplified by costs relating to the type of terrain (soil, vegetation, and wetness), the latter by
costs relating to slope and streams. Verhagen et al. (1999) calculate the accessibility of settlements in the
Vera Basin, Spain, on the basis of slope according to a formula provided by Gorenflo and Gale (1990).
They specify the effect of slope on travelling speed by foot as:
v = 6 e -3.5 | s + 0.05 |
where v = walking speed in km/h, s = slope of terrain in degrees, and e = the base for natural
logarithms. This function is symmetric but slightly offset from a slope of zero so the estimated velocity
will be greatest when walking down a slight incline. Bell (Bell et al. 2002) also employs an anisotropic
cost surface based on slope to generate a cumulative path network between Samnite sites in central Italy.
It should be noted that anisotropic functions only work if the direction of travel is taken into account –
otherwise they revert to isotropy. For instance, the variable representing slope in the Gorenflo and Gale
formula (above) has to be signed in order for the slight ‘preference’ for downslopes to emerge.
Introducing anisotropy in slope costs is not trivial, and recent approaches, which are based on the
capability of GIS to generate aspect (direction of steepest slope) maps, are forced into making simplifying
assumptions about the direction of travel. Bell and Lock (2000:90) introduce anisotropy in slope related
costs by interposing an aspect checking step – cutting costs by 50% for ‘angled’ascents and benefits by
50% for ‘angled’descents4. Likewise, the isotropic formula
Effort = (percent slope) / 10
was modified by Hayden (pers. comm.) into an anisotropic formula by calculating full cost upslope, no
cost cross-slope, and half-cost downslope. Hayden then added an isotropic cost layer for different terrain
types and terrain roughness (calculated as the change in slope). However, the best published work in this
regard is by Krist (2001a,b), who used the general orientation of historic native American trails in
Michigan to determine which aspects represent up, down, and sideways. On this basis he calculated an
adjusted slope Sa using the formula
Sa = S * cos(At – A)
where S and A are the original slope (in percent rise) and aspect values, and At is the direction to the
starting point. The multiplier cos(At – A) ranges from –1 to 1, generating negative values for downslopes
while attenuating the effect of sideslopes. To convert the adjusted slope values into cost surfaces
representing energy expenditure in kcal, they and a constant for average human walking speed of
100m/minute were entered into McDonald’s (1961) three physiological equations for topographic energy
expenditure5. The resultant cost surface could then be combined with a second, terrain, cost surface
representing additional barriers thrown up by lakes and wetlands.
Because anisotropy was introduced into the cost calculations, Krist then had to repeat them for both
directions of travel along each trail segment. A very similar approach was taken by De Silva and Pizziolo
(2001), who adapted an anisotropic function deriving from backpacking (Ericson and Goldstein
6 -
6
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
1980) to calculate a maximum friction surface for round trip movement in the Neolithic of the Biferno
valley (Italy), modifying it to:
effective friction = stated friction |cos^k ? a|
where ^k is a user defined coefficient (2 for movement on foot), and ? a is the difference in degrees
between the walking direction and the direction of maximum friction.
Figure 1 – Relation between terrain slope and energy cost (after Llobera 2000: fig 2).
Marble (1996) suggests that, since the function relating physiological expenditure to slope is
approximately symmetrical, we can safely ignore the whole problem of anisotropy. However, as the graph
in Figure 1 shows, the axis of approximate symmetry is at –10% (or -6°) of slope6. Consequently,
different optimal paths may obtain between two points depending on the direction of travel; hence, the
sign of the slope becomes a significant factor in the calculation of friction surfaces and least cost paths.
Krist’s approach (above) for calculating a direction adjusted slope map may be combined with the idea of
an axis of cost symmetry at -6° of slope, and the formula developed by Pandolf et al. (1977):
M = 1.5W + 2.0 (W + L) (L / W)2 + N (W + L) (1.5V2 + 0.35V * abs(G + 6))
This formula calculates the actual physiological expenditure M (metabolic rate in Watts) involved in
moving over natural terrain, and incorporates total weight moved (body W plus load L), velocity V, a
terrain factor N describing ease of movement, and percent slope G. I have replaced the final term G in
Pandolf’s formula by the absolute value of G plus 6 to represent the cost symmetry depicted in Figure 1.
Although it is not clear that Pandolf’s simple formula represents physiological expenditure as accurately
as the more complex functions advocated by McDonald (1961), it at least has the advantage of
incorporating the additional factors of load (L) and terrain (N). The terrain factor N is represented by a
separate cost surface constructed on the basis of terrain features known to influence movement - marshy
areas, roads, and streams of various widths. Marble (1996:5) supplies coefficients for a number of
different terrain types, relative to a standard hard but unmetalled surface.
2.3
DISCUSSION
Many of the techniques discussed above, by which costs are calculated, yield relative rather than absolute
costs; accessibility indices derived from these are therefore also relative. How should this affect our
interpretations? I have not come across any cases where absolute (physiological) cost played a direct
6 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
role; however, relative costs sometimes do have to be ‘calibrated’afterward. For example, in order to
derive 1 hour catchments for hillfort sites on the island of Hvar, Gaffney and Stancic (1991) were forced
to establish experimentally which cumulative costs were equivalent to 1 hour’s walk, by logging them at
the 1 hour cut-off point in several walks starting at one of these sites.
Llobera (2000:75) wondered how to combine the effect of landscape features (by which he meant
archaeological monuments) and ‘topographic’(physiological) cost into a cumulative cost calculation. On
the surface, this may seem to be a ‘technical’question, and Llobera himself hints at potential approaches
(2000:81-2), but a more fundamental issue surfaces as well. Whereas measurement and experimentation
can establish physiological costs, it is unlikely that ‘social’costs can be established with any degree of
objectivity. More importantly, it is possible to imagine an endless variety of ‘social’cost factors whose
effects may overlap, interact, and vary over space and time. The incorporation of social (cognitive) costs
into CSA therefore implies that the objective of establishing cost surfaces and paths whose values have
intrinsic meaning (for example, expressing travel time or metabolic energy) has been abandoned. It is not
at all clear that the values in the new, ‘enriched’cost surfaces may even be regarded as relative (ratio
scale) measurements, as seems to be suggested in several of the case studies presented in the volume
edited by Lock (2000).
Other than trying to agree among ourselves on the actual cost of travelling, are there any other immediate
tasks before us? I can see two. The first concerns one of the improvements I suggested in 1992 (Van
Leusen 1993), namely the differential weighting of the sites or foci used for cost surface calculations. In a
thought experiment, Llobera (2000:74) provided an example of this when he used a monument’s ‘rank’
(the derivation of which was not specified by him) as a multiplier to increase or reduce costs nearby.
The other task concerns improvement of least cost path analysis, one of the more promising areas of
development in cost surface analysis. Least cost paths between any pair of points can be generated from
cost surfaces in two steps: a cumulative cost surface is generated from the end-point of the pair, which is
then ‘drained’from the starting point to find the lowest-cost route between the two. Single least cost path
calculation has been used by archaeologists on a few occasions, for instance to derive optimum routes
between pairs of hillforts in Burgundy (Madry and Rakos 1996:113-117) or between Anasazi
communities in New Mexico (Katner 1996). Bi-directional least cost paths and corridors (i.e., least cost
zones wider than one pixel) were implemented by Krist (2001a,b) in recognition of the fact that many
approximate least-cost solutions could have been used in the past.
Compiling multiple least cost paths into a ‘least cost network’ was suggested early on by Tomlin
(1990:170-176 and 212-223) in an application searching for an optimum logging road network and was
first archaeologically implemented by Gaffney (pers. comm.) in order to model approaches to
Stonehenge; together with Gaffney I calculated similar least cost networks in the late Iron Age and
Roman landscape around the town of Wroxeter (Shropshire, UK; this thesis, chapter 16). The first
published European implementation is by De Silva and Pizziolo (2001:284-5), who calculate least cost
pathways between major Neolithic settlements of the Biferno valley (Italy), and note that secondary
settlements are located along these paths. Bell and his colleagues (2002) demonstrate a similar least cost
network simulating routes connecting Samnite settlements in central Italy, but refine the implementation
by basing it on the calculation of all possible reciprocal pathways.
However, these early examples are still based on relatively simplistic assumptions and coarse data. Route
networks are maintained by a range of user groups for a range of purposes: some routes – especially local
ones between individual settlements – are used relatively often by a small number of people, whilst other
routes are used much less often but by a much larger group of people. The latter tend to form a dendritic
network originating at the habitual locations of the inhabitants of a region, and converge on a small set of
shared resources such as market and cult places. Whilst current approaches concentrate on such dendritic
resource-centred networks, little or no attention has yet been paid to the importance of day-to-day social
networks by which neighbouring families and villages form and maintain a community. In future models
of archaeological landscapes it would make sense to combine resource and social networks.
6 -
8
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
At a technical level, the accumulation and drainage algorithms used for creating least-cost paths are also
far from perfect:
• Even a basic choice such as the selection of the grid resolution for analysis can have a major
influence on the outcome of a cost accumulation algorithm, implying that no confidence should be
placed in the precise line of a ‘least cost’path; hence least-cost corridors could well present a more
realistic approach.
• Cost accumulation is usually performed using either a 4-neighbour or an 8-neighbour filter; even if
the latter is used, the ‘Knight’s Jump’accumulation results in slightly incorrect accumulated costs for
most cells. Allowing more directions of movement can further reduce this so-called ‘elongation error’
of geometric distortion (Harris 2000:121).
• Drainage algorithms, in looking for the lowest neighbouring cell value, cannot reproduce the actual
overall least cost path, and in fact are quite likely to deviate significantly from it. Harris (2000:121)
mentions several alternative algorithms for calculating optimal routes, and these will have to be
evaluated for archaeological use.
Finally, we should carefully examine our model assumptions. Travel rarely if ever happens in a virgin
landscape – the landscape has a history of use, which means it is riddled with animal tracks and human
infrastructure as established and maintained by the forebears of the current inhabitants. These, in turn,
would have had intimate knowledge of this landscape. One could almost assume that, wherever one
could wish to go, some sort of path would have already existed! On a less grand note, rather than
climbing or descending very steep slopes, people will resort to hairpin bends in order to keep to a
comfortable degree of slope7. Usually there will be animal tracks to allow this. Thus, surmounting a steep
slope (ridge) only requires travelling a greater horizontal distance at a lesser vertical angle.
Most fundamentally and worryingly, a real traveller uses his knowledge of the terrain, the expected length
of the trip, the weather forecast, the final and intermediate goals, etc., to decide on the route - a decision
that weighs the global costs of alternative routes (cf. Bell and Lock 2000:92). All of the GIS least cost
implementations discussed here, in contrast, only make local decisions as to which neighbouring cell has
the highest or lowest value - they incorporate no global knowledge of the landscape at all. This defect can
perhaps be turned to good stead if GIS-generated least cost corridors are compared to historic routes:
deviations from the ‘optimum’route should then indicate the presence of intermediate goals which can
be further investigated. In general, more research into such comparisons is indicated, so that we are able
to assess precisely how far from reality our GIS-generated models still are.
3
3.1
LINE - OF - SIGHT ANAL Y S I S
8
PRINCIPLES AND A PPLICATIONS
The significance of visibility in the study of archaeological monuments was documented as long ago as
the early 18th century, when Stukeley first remarked on the ‘false horizon’setting of barrows. Since then,
visibility has been an acknowledged factor in the location and construction of archaeological monuments
such as hillforts, henges, watch towers as well as barrows, and both intervisibility and viewsheds were
formalised, though not yet digital, techniques by the 1970s. Until appropriate digital tools became
available in the late 1980s, the laboriousness of having to derive viewshed properties by field
observations and manual cartography meant that it never received more than incidental attention. One of
the tools that GIS has offered to archaeologists is viewshed analysis. It not only enables researchers to
quickly generate and test hypotheses about the (non-) visibility of salient sites and landscape features, but
also breathes new life into the study of landscape perception or cognitive archaeology.
6 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
Single viewshed analysis is now a well-trodden area in archaeological landscape modelling. The basic
technique operates on a digital elevation or terrain model (DEM, DTM9) to determine which areas are
visible from a given three-dimensional location. Single viewsheds indicate whether any two points are
intervisible and which area is visible from a particular point; they may also include information about the
angle of view. Applications in archaeology range from visual impact analysis for cultural resource
management - minimising the visual impact of modern development upon an archaeological landscape
(Katsaridis and Tsigouragos 1993, Knoerl and Chittenden 1990, Kvamme 1992) - to reconstructions of
Celtic road systems (Madry and Rakos 1996) and explorations of how prehistoric ritual landscapes might
have been perceived by contemporary populations (Fisher et al 1997, Ruggles and Medyckyj-Scott 1996,
Wheatley 1995, 1996). In further GIS analysis, the basic viewshed can be used to derive properties of the
visible areas, relating to such activities as hunting (van Leusen 1993, Krist and Brown 1994), security
(Loots et al. 1999, Madry and Rakos 1996), and the confirmation of cultural identity (discussed below).
Ruggles et al. (1993, 1996) and Fisher et al. (1997) employed viewshed analysis in the study of bronze age
monuments on the island of Mull, western Scotland, extending the idea of visibility to include prominent
horizon features and astronomical events. Prehistoric stone rows add the idea of directionality to viewshed
analysis, possibly aligning with landscape features to ‘pinpoint’relevant astronomical locations such as
points where the moon rises and sets.
For specific purposes, the concept of viewshed calculation has been refined in order to study intervisibility
(whether two or more monuments are intervisible and might therefore be part of the same ‘system’; e.g.,
Gaffney & Stancic 1991, Bradley et al. 1993, Ozawa et al. 1995, Bell 1999, Haas and Craemer 1993,
Moscatelli 1998) and visual alignment (whether two points align in order to visually emphasise or frame a
third point; Ruggles et al 1993). Single viewsheds have also been merged to yield multiple viewsheds
(Jacobson et al 1994) and added to yield cumulative viewsheds (Gaffney et al. 1996b; Wheatley 1995,
1996), both of which will be discussed in more detail below. Finally, the concept of viewshed analysis
logically extends to the complement of visibility, the study of non-visible areas and monuments. Whereas
one particular viewshed will show which areas are hidden from view from a particular vantage point,
multiple viewsheds will highlight areas hidden from view from a class of monuments, with the potential
of having a regional (ritual?) significance. Cumulative viewsheds refine this idea by giving a measure of
how hidden particular locations are, enabling us to rank these locations by degree of seclusion. While
viewshed exclusion – the deliberate placing of monument or activity so as not to be visible from specific
other locations - has figured in archaeological studies, notably Tilley’s Phenomenology of Landscape (1994), it
has only been the subject of one GIS publication in recent years (Lock and Harris 1996:224).
A rather different set of applications arises when viewsheds are used in ‘cookie cutters’fashion to derive
properties of other data layers in the GIS. Such an approach looks beyond quantitative visibility
properties per se and asks which objects and terrain features, present within the viewshed, might provide a
reason for there being a viewshed in the first place. I suggested the example of deriving physiographical
properties of the viewsheds of Mesolithic sites in the southern Netherlands, some of which might be
camps relating to big game hunting (Van Leusen 1993:118-121). Recently, Wheatley and Gillings (2000:
14-23) have applied the cookie cutter technique in the framework of the so-called ‘Higuchi’viewshed
properties used in landscape planning. They demonstrated the derivation of the distance-related property
of clarity (using distance classes based on the visual appearance of standard objects (trees)) and the
property of directionality based on a calculated aspect layer and ‘directionality’.
3.2
METHODOLOGICAL I SSUES
In ascending order of complexity three areas of methodological concern can be distinguished with
current viewshed applications – those of realism, of edge effects, and of significance.
Firstly, the issue of realism – is the modelled viewshed sufficiently congruent with the ‘real’viewshed to
allow archaeological interpretation? In one sense, this is a fairly straightforward technical issue, and
6 -
10
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
a fair amount of GIS literature already comments on the pertinent issues of data quality (especially of the
DEM that underlies all viewshed analysis – see the important study by Wood (1996), who also supplies
further references), operational assumptions such as the viewing parameters and the use of palaeoenvironmental reconstructions (cf. Tschan et al. 2000), and the algorithm employed in the calculation of
viewsheds (Fisher 1991, 1992, 1993, 1994; see also Loots et al. 1999; Nackaerts et al. 1999 on the
calculation of ‘fuzzy’viewsheds). The issue is complicated, however, by theoretical considerations such as
the relative merits of employing an ‘objective’ Cartesian view of geographical space, or of using
subjective notions that involve viewer and viewed in a more complex interaction.
Secondly, the issue of edge effects. Since viewsheds are generally large relative to the study region
(especially if their radius is unconstrained), they tend to ‘fall off the edge’of the region. Conversely,
viewsheds of sites lying outside the region would fall partly within the region – but those sites are not
part of the analysis so their viewsheds are never calculated! If not properly corrected for, this effect will
lead to incorrect multiple and cumulative viewshed calculations, hence to incorrect archaeological
interpretations. For example, in a 20 by 20 km study region, calculating viewsheds with a 7 km radius
would leave only a 6 by 6 km area in the centre of the study region where the visibility index values are
correct; in all areas within 7 km of the edge of the region the cumulative viewshed index (CVI) rises
inversely proportional to the distance from that edge (see chapter 16 for a relevant case study).
The edge effect can manifest itself in unexpected ways. For example, Madry and Rakos’(1996) study of
the Celtic road network in the Arroux valley in Burgundy suggests that there is a viewshed relation
between these roads and the nearby hillforts, and that the intention was to keep the transportation
network under constant visual control from these defensive sites. A cumulative hillfort viewshed is
calculated and the roads are found to lie largely within the high visibility values. Statistical support for
this is obtained by comparing the visibility index of the roads with those of the total study region.
However, as no account was taken of the edge effect, the visibility values for the region are incorrect and
the conclusion that the roads have a significantly high visibility is unsupported (though it may well be
true).
Thirdly, the issue of statistical significance – are the visibility characteristics associated with
archaeological remains significantly different from background values? Wheatley (1995) discusses one
correction that should be standard in all cumulative viewshed operations - the ‘view to itself’effect. In
the examples discussed by him, this effect entails that the number of barrows observed to occur in a
particular viewshed is always one higher than it should be, leading to misinterpretation of statistical
results. Even more insidious is the ‘viewshed radius effect’; I conducted some simulations (this thesis,
chapter 16) that show that the size of the viewshed radius has a profound effect on the distribution of
visibility index values across the terrain. For any set of points (including archaeological objects), choosing
a small viewshed radius will result in a ‘preference’for the lower elevations (valley bottoms) occurring in
the study area, whereas choosing a large radius will result in a ‘preference’for the higher elevations
(peaks and ridges). A good example of this effect at work can be found in Lock and Harris (1996: 224,
fig 13.5), who note that viewsheds of Neolithic long barrows in the Danebury region are apparently
selected so as to ‘alert people crossing the surrounding ridgetops’. My work indicates that this ‘rim effect’
might or might not be entirely due to the choice of viewshed radius.
It is no longer sufficient just to report on the properties of the viewsheds generated for groups of
archaeological monuments - archaeological relevance depends on such viewsheds being sufficiently
different from the background visibility properties of the study area. For example, viewsheds taken from
high points in the landscape will tend to include relatively many other high points - ridges, peaks and
such. A sample of viewpoints drawn from such locations (hillforts, barrows) will therefore preferentially
‘see itself’. For example, Wheatley (1995, Plate 1) employs cumulative viewshed analysis to study the
spatial relationship between barrows in the Stonehenge and Avebury areas. His analysis clearly
demonstrates the correlation between viewsheds and landscape morphology, with ridges and peaks being
preferentially seen. Wheatley rightly cautions (ibid., 180) against equating such statistical correlation with
causation, but does conclude that being able to see other barrows is likely to have been a determining
6 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
factor for barrow placement in the Stonehenge area.
It is all too easy to employ viewshed analysis simply to support one’s preconceived ideas about the
cultural and cognitive significance of archaeological monuments, especially if there is little or no
methodological control on these quantitative models. Gaffney et al. (1996a:148ff), in their discussion of
the viewsheds of monuments in the Kilmartin area of Scotland, fail to convince for this very reason. If,
as these authors themselves state, the rock art and standing stones in this area are not visible from more
than 100 meters and 3km away respectively, then what is the use of calculating 15km viewsheds?
3.3
VISIBILITY, PERC EPTION, AND THE COGN ITIVE LANDSCAPE
It is becoming increasingly clear that archaeologists working with GIS want to be able to escape from the
‘objective’geographical space enforced upon them by the design of the software. They want to be able to
represent social space - the subjective experience of past people, their perception of their physical and
social environment, and their cognitive representation of their world. This is part and parcel of the
general cognitive-processual trend in recent theoretical work, a good overview of which can be found in
Renfrew and Zubrow’s The Ancient Mind (1994). Perception and cognition of the landscape are two
different concepts, although our perception of the landscape is obviously steered and modified by our
cognition (or lack thereof) of its history and constituents.
Perception, as the simple act of being aware of the landscape, has already been to some extent the subject
of GIS-study among both geographers and archaeologists. Geographers intend to incorporate qualitative
spatial reasoning into formal GIS models (see, for instance, Frank 1996, for a discussion of how
reasoning with cardinal directions can be so formalised). Archaeologists have concentrated on less
complicated visibility issues involving significant ritual and political landscape features (e.g., Boaz and
Uleberg 1995, Nunez et al 1995, Gaffney et al. 1996a, Llobera 1996). Some thought but little action has
so far gone into the generation of perceptual variables such as ‘enclosedness’ vs. openness of the
landscape (Llobera, pers. comm.); the potential of such approaches is therefore not yet clear.
Cognitive archaeology in the context of landscape archaeology is the archaeology that concerns itself with the
cognitive aspects of past geographical and human landscapes, that is, the perception of significance.
According to Zubrow (1994), ‘one goal [of cognitive archaeology] is to show that people had preferences
independent of economic necessity, and some decisions are independent of utility’. He continues ‘as
archaeologists, one of our ultimate goals is to extract the cultural ideals from the complicated reality in
the complex patterns of prehistoric material remains’.
If we abandon our viewpoint as an external, even extra-spatial, observer of the archaeological landscape
as represented by GIS-generated maps, we may instead adopt another role - that of the participant in a
cognitive landscape. The link between visibility and cognition has been well made by Gaffney et al.
(1996b):
“A viewshed represents the area in which a location or monument may communicate visual
information. Viewsheds may overlap, producing zones in which an observer might be aware of the
presence of many such locations, all of which may carry information. The increased density of
such information can in some circumstances be interpreted as a measure of the importance of a
particular area. It provides a spatial index of perception, mapping the cognitive landscape within
which the monuments operated.”
Many authors have begun to explore the prehistoric cognitive landscape via visibility in recent years, even
though, as Fisher (2000:9) reminds us, ‘it is not at all clear how we might compute modern cognition in the
landscape with GIS’(my emphasis, MvL). For good reasons, such experimental GIS applications tend to
concentrate on well-preserved and well-studied ritual landscapes such as the Stonehenge environs, that
offer unusually complete data sets and a relatively high a priori degree of certainty that visibility was an
important consideration when the monuments in these areas were constructed. These explorations,
6 -
12
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
when visualised appropriately, have the potential of involving us much more closely with the past.
Woodward and Yorston (1996) bring the study of landscape perception closer to dynamic Virtual Reality
by interactively presenting changes in viewsheds as the viewer moves along the Stonehenge Avenue and
different groups of barrows come into view.
3.4
FURTHER WORK
Further work in improving viewshed analysis will need to deal with two issues. The first concerns the
technical application of viewsheds; the second, their theoretical justification.
Various technical improvements to viewshed analysis have already been proposed. For example, distance
decay functions and ‘fuzzy viewsheds’have been used in order to simulate the loss of visual resolution
with distance (Fisher 1991, 1992, 1993) and to move from deterministic to probabilistic viewshed models
(Fisher 1994, Nackaerts et al. 1999, Loots et al. 1999). Wheatley (1995: 181-2) includes a useful
discussion of error and uncertainty in viewshed maps. Others (Ruggles and Medyckyj-Scott 1996) have
applied a correction for earth curvature that is particularly relevant for the modelling of astronomical
observations10. Further improvements follow from the discussion of methodological problems above.
The preliminary visibility significance tests I conducted indicate that statistical control of viewshed
analysis needs to be much stronger before any archaeological interpretations can be built upon it. The
tests will have to be generalised so that reproducible results can be obtained from them, and a proper way
of incorporating background visibility data into viewshed analysis has been found. One way forward
might be by resorting to relative visibility measures – for example, Lock and Harris (1996:232 and fig
13.15) note that early Iron Age hillforts in the Danebury area are positioned to maximise visual
dominance over adjacent valleys and surrounding farmsteads at the cost of all-round defensive visibility. I
am less happy with the tack taken by Gillings (1999) and Woodward and Yorston (1996). Gillings looks
to Virtual Reality visualisations in order to explore the significance of archaeological viewsheds;
Woodward and Yorston have implemented an application similar in spirit, that uses Java software to
create interactive maps of the barrows in the Stonehenge area, where barrows visible from the current
position of the mouse cursor light up. Although this type of work certainly comes closer to the postprocessualist ideal of being participant in, rather than an observer of, the archaeological landscape, I am
worried by what must be an increasing temptation to throw technical rigour to the wind.
Justifying viewshed analysis on a theoretical level is equally as important as its technically competent
implementation. How important is it in fact to be able to see directly a particular site, monument, or
social activity, as opposed to knowing or being aware of its presence and location? Smoke and fires, the
latter especially after dusk, must surely rate among the most visible phenomena even from a great
distance, and it is not at all necessary to be able to see the source of the smoke and fire to be aware of what
is going on.
And of course there is no reason to limit ourselves to vision: things heard or smelled might be as
significant as things seen – the Neolithic barrows of the English chalklands, when first constructed,
would have been highly visible, but their visibility must have dropped as they gradually became
overgrown with moss, lichen, and grass. Could it be that the less permanent features of a barrow
cemetery were in fact the most visible (totem poles), audible (wind chimes), or smellable (decomposing
offerings?). The first hints of research turning in this direction are now detectable: classical farmsteads in
the countryside surrounding the ancient Greek city of Hyettos are thought to take ‘advantage of the
unique acoustic effects of the basin ‘auditorium’surrounding the acropolis and lower city, enabling them
to partake aurally as well as visually in the activities taking place’(Gillings 2000:115; Gillings refers to the
modelling of such perceptions as ‘sensuous’GIS).
Following this line of reasoning to its logical conclusion, one might even question whether the cognitive
landscape is not constituted as much by ‘unsensed’presences as it is of the more direct sensed kind.
Inspiration for such thinking will no doubt be found in the ethnographic literature, but there is also the
6 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
danger of over-interpretation – anything in the landscape could have had cognitive significance. That does
not mean to say that it had.
4
CONCLUSIONS
INCREASED REALISM
It is evident that, a decade after the first LOSA / CSA studies were conducted, an initial phase
characterised by naïve applications and constrained by the capabilities of generic GIS has drawn to an
end. It is currently being replaced by a phase in which specific procedures are being proposed in order to
implement ever more realistic models of human perception of the landscape.
Although opinions about how to improve studies based on viewshed and cost surface techniques are
divided, all are agreed that they can and should be improved. On the one hand, exploration of
refinements to the ‘environmental’approach current throughout the past decade continues - see, for
example, the experimentation with the inclusion of a variable representing seasonally changing vegetation
in the Polish Mesolithic in viewshed studies by Tschan et al. (2000). On the other hand, proponents of
the ‘enrichment’approach are particularly vocal regarding methodological improvement, which they see
coming mainly from the field of landscape analysis. Baldwin et al. (1996) provide a very useful review of
actual and potential approaches to modelling environmental cognition through line-of-sight analysis in
GIS from the perspective of landscape analysts. They divide personal experience of the landscape into
four categories - physiographical characteristics, the presence of specific physical features, cognitive
variables, and viewer interest, and conclude that deterministic analysis in GIS can become more accurate
by adopting a flexible approach to cognitive criteria. Early explorations of such an approach are Wheatley
and Gillings’(2000) investigations in the framework of Higuchi viewshed properties, and Llobera’s
(2000) implementation of ‘attractive’and ‘repellent’features in the landscape.
However, fundamental limitations to the use of GIS in such landscape studies have already come in
sight, causing Lock (2000:60-62) to predict that refining current approaches will continue to produce
inadequate models because they continue to represent meaning as attributes of the landscape rather than
as properties of the people in it. Such a fundamental shift in emphasis suggests that future models should
take the form of artificial societies built by programs representing individual humans, their intentions and
reactions to external stimuli (cf. Llobera 2000:81-2). This new phase is likely to be dominated by a much
more limited number of researchers who have access to the specialist software tools needed for this type
of landscape analysis.
TESTING & VALIDATION, OR EXPLORATION?
The apparent conflict between adherents of processualist and post-processualist approaches has been
shown to be beside the point from a pragmatic point of view. The practical differences between studies
presented so far by either side appear to be small, and a much more significant watershed is likely to
separate studies that fail to adduce proper supporting evidence to their interpretations, from those that
do. Whilst naïve processualist approaches have since long rightly been criticised for failing to address
pragmatic and procedural shortcomings (most recently in Wheatley & Gillings 2000: 5-14), having a poststructuralist outlook in itself does not help to establish procedural rigour. Both Fisher (1999:9-10) and
Llobera (2000:66) note the lack of validation or even methodology accompanying post-processual works.
A general point that emerges is the lack of supporting evidence given for claims of unusual cost or
viewshed properties for particular locations within a region. Fisher (1999:8) is not the only one who has
noted that most ‘contextual’studies do not attempt spatio-statistical analysis, and therefore lack proof for
their inferences. Yet tests can be carried out to demonstrate that the results obtained are unlikely to have
arisen by chance, as is shown by Fisher and others (1997), who employ Monte Carlo testing of their
6 -
14
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
hypotheses.
The general approach advocated to substantiate LOSA and CSA results obtained for a sample of
archaeologically meaningful locations is to compare them with one large or many similar-sized samples of
random chosen locations (cf. Wheatley 1995). This is typically done by generating a cumulative visibility
index (CVI; see, for example, Lake et al 1998:36-38, Bell & Lock 2000: 96-98) or a cumulative
accessibility index (CAI; for example, Llobera 2000: 70ff) for all or a representative subset of locations
within the study region. The result obtained for the sample of interest can then be formally compared to
the population (one-sample tests) or to a representative sample of it (two-sample tests).
Lake et al. (1998), investigating whether the viewshed sizes (areas) of Mesolithic sites on the island of
Islay off the Scottish west coast were significantly different from those of non-sites, used a two-sample
significance test of sites against a 5% random sample of locations and found that their hypothesis was
not supported by the evidence.
Another method by which LOSA- or CSA-based models may be supported is by comparison with
independent archaeological evidence. For example, networks of least-cost paths may be compared to
historically known networks such as the mule-paths that criss-crossed the Italian highlands until recently.
If this approach is taken, circular arguments are to be avoided: in some cases, models have been
‘tweaked’until they fit the evidence, after which the fit is taken as evidence for the correctness of the
model. For example, the optimal path calculated by Bell and Lock (2000) adheres to the known route of
the Ridgeway not only because of the constraints imposed by the distinctive landform of the region (as
noted by Harris 2000:119), but also because the authors adjusted several parameters in the cost
calculation in order to force the optimal path into resembling the route of the Ridgeway. Clearly, in such a
case the resemblance between the two paths cannot be taken to be supporting evidence for the
correctness of the calculation.
PROCEED WITH CAUTION
This paper critically examines the logic of assigning cognitive significance on the basis of multiple or
cumulative visibility and accessibility indices, and finds that insufficient attention has been paid to some
important methodological aspects of spatial analysis – most notably the need to calculate ‘background’or
‘potential’indices against which an actual outcome may be judged. Recent work points to least cost path
analysis as the most profitable avenue for further research in cost surface analysis. There are at least two
avenues for further work here; firstly, analysis of historic infrastructural networks which may serve to
‘calibrate’ data-independent models; and secondly, vector analysis of networks constructed through
raster-based cost surface analysis. Recent viewshed applications seem to concentrate on studying the
(inter-) visibility of ritual monuments, but as has been made clear here will need to apply a lot more
rigour to their technical execution.
Two specific approaches have been suggested: firstly, since there are a large number of potential sources
of error, it is deemed unwise to believe the outcome of any particular LOSA or CSA. As Wheatley and
Gillings (2000:5) suggest, we should instead study the trends emerging from an accumulation of such
single outcomes. Secondly, rather than attempting to interpret the viewshed or accessibility properties of
sites directly, we should study the differences between sites and between sites and ‘background’. These
approaches express a probabilistic view of modelling - one where models are used to indicate only how
probable it is that certain activities take place in certain places.
Rather than continuing a fruitless processual / post-processual debate, this paper shows current GIS
implementations of ‘cognitive’landscapes to be little more than a semantic change of clothes. The postprocessualist argument has mostly taken the form of a bashing of supposedly ‘data-led’, ‘environmentally
determinist’, and ‘naïve’ applications copying the worst of New Archaeology practices. However,
applications billing themselves as ‘cognitive archaeology’seem to boil down to the same combinations of
viewshed and cost surface analysis explored by others as well. It is also possible to argue that cost surface
and viewshed calculations are themselves deterministic methods. Llobera’s (1996) study of the
6 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
visibility of late prehistoric ditches in the Wessex chalklands, although couched in a theoretical context
rather different from that of systems theory and processualism, still attempts to derive cognitive aspects
of late Bronze Age society (the awareness of being inside a territory) in a deterministic manner - the
location of the ditches is fixed, the calculation of their visibility is based purely on properties of the
elevation model. So the difference with what has been termed environmental determinism is in the
environmental, not the determinism, and we might as well speak of cognitive determinism when describing
such work.
ACKNOWLEDGEMENTS
Much of the first version of this article was written while working as a Leverhulme Fellow with the
Wroxeter Hinterland Project at the University of Birmingham Field Archaeology Unit. I am most grateful
to the Leverhulme Trust for the grant that made the project possible, and to Dr Vince Gaffney for his
kind invitation to join the project and for providing stimulating discussion and much useful additional
information on the subject of GIS procedures. The Annotated GIS Bibliography by Petrie et al. (1995) was a
useful entry point into the less easily accessible literature, and I thank its authors for the service they have
thus provided. I would also like to thank the members of the GISARCH discussion list, and especially
Duane Marble, for information that has allowed me to improve and expand the paper.
REFERENCES
Aldenderfer, M & H Maschner 1996
Anthropology, Space, and Geographic Information Systems. New York, Oxford: Oxford University Press.
Anderson, DG & JC Gillam 2000
Paleoindian Colonisation of the Americas: Implications From an Examination of Physiography,
Demography, and Artifact Distribution, American Antiquity 65(1):43-66.
Baldwin, J, P Fisher, J Wood & M Langford 1996
Modelling environmental cognition of the view with GIS. Proceedings, 3rd International Conference on
Integrated GIS and Environmental Modelling.
Bampton, M 1997
Archaeology and GIS: the View from Outside, in Archeologia e Calcolatori 8: 9-26.
Barceló, JA, I Briz & A Vila (eds) 1999
New Techniques for Old Times. CAA 98. Computer Applications and Quantitative Methods in
Archaeology. Proceedings of the 26th Conference. BAR International Series, 757.
Belcher, M, A Harrison & S Stoddart 1999
Analyzing Rome’s Hinterland, in Gillings, M et al. (eds), Geographical Information Systems and
Landscape Archaeology. The Archaeology of Mediterranean Landscapes 3: 95-101.
Bell, T 1999
Reconstructing Archaeology from the Landscape: GIS, CAD and The Roman Signal Station at Whitby. In
Dingwall, L et al. (eds), Archaeology in the Age of the Internet. Proceedings of the CAA97 conference.
BAR International Series 750.
Bell, T & G Lock 2000
Topographic and cultural influences on walking the Ridgeway in later prehistoric times, in Lock, G (ed)
2000, Beyond the Map: Archaeology and spatial technologies: 85-100. Amsterdam, etc: IOS Press.
Bell, T, A Wilson & A Wickham 2002
Tracking the Samnites: landscape and communications routes in the Sango valley, Italy, American
Journal of Archaeology 106(2).
Bietti, A, A Cazzella, I Johnson & A Voorrips (eds) 1996
Theoretical and Methodological Problems. Proceedings of the XIIIth International Congress of Prehistoric
and Protohistoric Sciences, Vol. 1. Forlí: ABACO.
Boaz, JS & E Uleberg 1995
The potential of GIS-based studies of Iron Age cultural landscapes in Eastern Norway, in Lock, G & Z
Stancic (eds), GIS and Archaeology: a European Perspective: 249-260. London: Taylor & Francis.
Bradley, R, J Harding & M Matthews 1993
The siting of prehistoric rock art in Galloway, south-west Scotland, Proceedings of the Prehistoric Society
59:269-283.
Bradley, R 1994
6 -
16
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
Prehistoric land divisions on Salisbury Plain: the work of the Wessex Archaeological Trust. London:
English Heritage.
Chapman, H 2000
Understanding wetland archaeological landscapes: GIS, environmental analysis and landscape
reconstruction; pathways and narratives, in Lock, G (ed) 2000, Beyond the Map: Archaeology and spatial
technologies: 49-59. Amsterdam, etc: IOS Press.
Chisholm, M 1968
Rural Settlement Land Use. London: Hutchinson.
De Silva, M & G Pizziolo 2001
Setting up a ‘Human Calibrated’ Anisotropic Cost Surface for Archaeological Landscape Investigation, in
Stancic, Z & T Veljanovski (eds) 2001:279-286.
Dingwall, L, SC Exon, VL Gaffney, S Laflin & PM van Leusen 1999
Archaeology in the Age of the Internet. Proceedings of the CAA97 conference. BAR International Series
750. Oxford: Archaeopress.
Ericson, JE & R Goldstein 1980
Work Space: A New Approach to the Analysis of Energy Expenditure Within Site Catchments. In:
Findlow, FJ & JE Ericson (eds), Catchment Analysis: Essays on Prehistoric Resource Space: 21-30.
UCLA: Anthropology UUCLA Vol. 10.
Fisher, PF 1991
First experiments in viewshed uncertainty: the accuracy of the viewshed area, in Photogrammetric
Engineering & Remote Sensing 57(10):1321-1327.
Fisher, PF 1992
First experiments in viewshed uncertainty: simulating fuzzy viewsheds, in Photogrammetric Engineering
& Remote Sensing 58(3):345-352.
Fisher, PF 1993
Algorithm and implementation uncertainty in viewshed analysis, International Journal of GIS 7(4):331347.
Fisher, PF 1994
Probable and fuzzy concepts of the uncertain viewshed, in Worboys, M (ed), Innovations in GIS 1:161175. London: Taylor & Francis.
Fisher, PF 1996a
Extending the applicability of viewsheds in landscape planning, Photogrammetric Engineering & Remote
Sensing 62 (11): 1297-1302.
Fisher, PF 1999
Geographical Information Systems: Today or Tomorrow?, in Gillings, M et al. (eds), Geographical
Information Systems and Landscape Archaeology. The Archaeology of Mediterranean Landscapes 3:5-11.
Fisher, PF, C Farrelly, A Maddocks & C Ruggles 1997
Spatial analysis of visible areas from the bronze age cairns of Mull, Journal of Archaeological Science
24:581-592.
Frank, AU 1996
Qualitative spatial reasoning: cardinal directions as an example, International Journal of Geographical
Information Systems 10-3: 269-290.
Gaffney, VL & PM van Leusen 1995
Postscript - GIS, Environmental Determinism and Archaeology, in Lock, G & Z Stancic (eds),
Archaeology and Geographical Information Systems: 367-382. London: Taylor & Francis.
Gaffney, V, K Ostir, T Podobnikar & Z Stancic 1996b
Spatial analyses, field survey, territories and mental maps on the Island of Brac, Archeologia e
Calcolatori 7: 27-41.
Gaffney, VL & Z Stancic 1991
GIS Approaches to Regional Analysis: A case study of the island of Hvar. Ljubljana: University of
Ljubljana.
Gaffney, VL, Z Stancic, J Farley et al. 1993
Geographical Information Systems, Territorial Analysis and Prehistoric Agriculture on the Island of Hvar,
Dalmatia, in Fabis,M, I Kuzma & K Markova (eds), Actes du XIIe Congres International des Sciences
Pre- et Protohistoriques, Bratislava, 1-7 Septembre 1991: 407-415. Bratislava: Institut Archeologique de
l’Academie Slovaque des Sciences a Nitra.
6 -
17
V A N L EUSEN : P A T T E R N T O P ROCESS
Gaffney, VL, Z Stancic & H Watson 1996a
Moving from Catchments to Cognition: Tentative Steps Toward a Larger Archaeological Context for GIS,
in Aldenderfer, M & H Maschner (eds), Anthropology, Space, and Geographic Information Systems: 13254. New York, Oxford: Oxford University Press.
Gillings, M 1996
Not Drowning but Waving? – Re-Humanising GIS, the Tisza Floodplain Revisited, in A Bietti et al. (eds),
Theoretical and Methodological Problems. Proceedings of the XIIIth International Congress of Prehistoric
and Protohistoric Sciences, Vol. 1: 69-84. Forlí: ABACO.
Gillings, M 1998
Embracing uncertainty and challenging dualism in the GIS-based study of a palaeo-flood plain, European
Journal of Archaeology 1(1):117-144.
Gillings, M 1999
Engaging Place: a Framework for the Integration and Realisation of Virtual-Reality Approaches in
Archaeology. In Dingwall, L et al. (eds), Archaeology in the Age of the Internet. Proceedings of the
CAA97 conference. BAR International Series 750.
Gillings, M & DW Wheatley in press
Seeing is not believing: unresolved issues in archaeological visibility analysis, in Slapsak, B (ed), On the
good use of GIS in Ancient Landscape Studies. Ljubljana.
Gillings, M & A Wise 1998
GIS Guide to Good Practice. http://ads.ahds.ac.uk/project/goodguides/gis/. York: Archaeology Data
Service.
Goodchild, H 1999
An Appraisal of the Urban Dependency of Wroxeter Roman City. Unpublished MA Dissertation,
University of Birmingham.
Gorenflo, LJ & N Gale 1990
Mapping regional settlement in information space, Journal of Anthropological Archaeology 9:240-274.
Haas, J & W Creamer 1993
Stress and Warfare among the Kayenta Anasazi of the Thirteenth Century AD, Fieldiana Anthropology,
New Series 21. Chicago: Field Museum of Natural History.
Harris, T 2000
Moving GIS: exploring movement within prehistoric cultural landscapes using GIS, in Lock, G (ed) 2000,
Beyond the Map: Archaeology and spatial technologies: 116-123. Amsterdam, etc: IOS Press.
Helbing, D, J Keltsch & P Molnár 1997
Modelling the Evolution of Human Trail Systems, Nature 388:47-50.
Hillier, B 1996
Space is the Machine. Cambridge: Cambridge University Press.
Hillier, B & J Hanson 1984
The Social Logic of Space. Cambridge: Cambridge University Press.
Hodder, I & C Orton 1976
Spatial Analysis in Archaeology. Cambridge: Cambridge University Press.
Huckerby, CL 1999
GIS and Prehistoric Mammal Acquisition Patterns. In Dingwall, L et al. (eds), Archaeology in the Age of
the Internet. Proceedings of the CAA97 conference. BAR International Series 750.
Jacobson, E, J Meachan & D Cutting 1994
Patterns on the Steppe: Applying GIS to the Archaeology of the Altay Mountains, Geo Info Systems 4(3):
32-45.
Johnson, I & M North 1997
Archaeological Applications of GIS. Proceedings of Colloquium II, UISPP XIIIth Congress, Forlí, Italy,
September 1996. CD-ROM, Sydney University Archaeological Methods Series 5.
Kamon, E 1970
Negative and Positive Work in Climbing a Laddermill, Journal of Applied Physiology 29:1-5.
Kantner, J 1996
An Evaluation of Chaco Anasazi Roadways, paper presented at the 61st SAA Annual Meeting, New
Orleans, Louisiana.
Katsaridis, P & Y Tsigourakos 1993
The Use of GIS in Land Use Planning for the Protection of the Delfi Hinterland (Greece), in Proceedings
6 -
18
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
of the 13th Annual ESRI Conference, Palm Springs, California: 321-327. Redlands: ESRI.
Knoerl, JJ & B Chittenden 1990
Boundary Analysis of the Dune Shacks of Peaked Hill Bars Historic District, Cape Cod, Massachusetts, in
Chittenden, B (ed), Cultural Resources Information Management Series: 1-15. Washington: US National
Park Service.
Krist, FJ 2001a
A Predictive Model of Paleo-Indian Subsistence and Settlement, dissertation submitted to the Department
of Anthropology, Michigan State University.
Krist, FJ 2001b
Anisotropic Cost Modeling as a Tool to Better Understand Hunter/Gatherer Mobility Within the Great
Lakes Region, paper presented at the 6th SAA meeting, New Orleans 2001.
Krist, FJ & DG Brown 1994
GIS Modelling of Paleo-Indian Period Caribou Migrations and Viewsheds in Northeastern Lower
Michigan, Photogrammetric Engineering and Remote Sensing 65 (9): 1129-1137.
Kvamme, KL 1992
Geographic information systems and archaeology, in Lock, G & J Moffett (eds), Computer Applications
and Quantitative Methods in Archaeology 1991. BAR International Series 577: 77-84.
Kvamme, KL 1999
Recent Directions and Developments in Geographical Information Systems. Journal of Archaeological
Research 7(2):153-201.
Lake, MW, PE Woodman & SJ Mithen 1998
Tailoring GIS Software for Archaeological Applications: An Example Concerning Viewshed Analysis,
Journal of Archaeological Science 25:27-38.
Limp, WF 1989
The Use of Multispectral Digital Imagery in Archaeological Investigations, Arkansas Archaeological
Survey Series 34. Fayetteville: AAS.
Limp, WF 1990
Intersite Analysis: Aboriginal use of the Rush locality, in Sabo, GI (ed), Archaeological Investigations at
3MR80-Area D in the Rush Development Area, Buffalo National River, Arkansas, Professional Paper 38,
vol. 1: 295-345. Santa Fe: US National Park Service.
Llobera, M 1996
Exploring the topography of mind: GIS, social space and archaeology, Antiquity 70: 612-22.
Llobera, M 1999
Llobera, M 2000
Understanding movement: a pilot model towards the sociology of movement, in Lock, G (ed) 2000,
Beyond the Map: Archaeology and spatial technologies: 65-84. Amsterdam, etc: IOS Press.
Lock, G (ed) 2000
Beyond the Map: Archaeology and spatial technologies. Amsterdam, etc: IOS Press.
Lock, G, T Bell & J Lloyd 1999
Towards a Methodology for Modelling Surface Survey Data: The Sangro Valley Project, in Gillings, M et
al. (eds), Geographical Information Systems and Landscape Archaeology. The Archaeology of
Mediterranean Landscapes 3: 55-63.
Lock, G & Z Stancic 1995
Archaeology and Geographical Information Systems. London: Taylor & Francis.
Lock, GR & TM Harris 1996
Danebury Revisited: An English Iron Age Hillfort in a Digital Landscape, in: M Aldenderfer & H
Maschner (eds), Anthropology, Space, and Geographic Information Systems: 214-40. New York, Oxford:
Oxford University Press.
Lockyear, K, TJT Sly & V Mihailescu-Birliba (eds) 2000
CAA96. Computer Applications and Quantitative Methods in Archaeology. BAR International Series 845.
Oxford: Archaeopress.
Loots, L, K Nackaerts, & M Waelkens 1999
Fuzzy viewshed analysis of Hellenistic city defence systems at Sagalassos, Turkey. In Dingwall, L et al.
(eds), Archaeology in the Age of the Internet. Proceedings of the CAA97 conference. BAR International
Series 750.
Madry, SLH & CL Crumley 1990
6 -
19
V A N L EUSEN : P A T T E R N T O P ROCESS
An application of remote sensing and GIS in a regional archaeological settlement pattern analysis, in
Allen, KMS et al. (eds), Interpreting Space: GIS and Archaeology: 364-380. London: Taylor & Francis.
Madry, SLH & L Rakos 1996
Line-of-sight and cost-surface techniques for regional research in the Arroux River valley, in HD
Maschner ed., New Methods, Old Problems. Geographic Information Systems in Modern Archaeological
Research (Southern Illinois University Center for Archaeological Investigations Occasional Paper 23):
104-126.
Marble, DF 1996
The Human Effort Involved in Movement over Natural Terrain: A Working Bibliography, Unpublished
report, Department of Geography, Ohio State University.
Margaria, R 1938
Sulla fisiologia e especialmente sul consumo energetico della marcia e della corsa a varia velocitá ed
inclinazione del terreno, in Atti e Memorie Accademia Nazionale delle Lincei: 299-368.
Massagrande, F 1996
The Romans in southwestern Spain: total conquest or partial assimilation? Can GIS answer?, in H
Kamermans & K Fennema eds., Interfacing the Past: CAA95 (Analecta Praehistorica Leidensia 28): 325330.
Massagrande, F 1999
A GIS Study on the Spatial Development of Coastal Catalunya. In Dingwall, L et al. (eds), Archaeology in
the Age of the Internet. Proceedings of the CAA97 conference. BAR International Series 750.
McDonald, I 1961
Statistical Studies of Recorded Energy Expenditure of Man, Nutritional Abstracts and Reviews 31(3):739762.
Minetti, AE 1995
Optimum Gradient of Mountain Paths, Journal of Applied Physiology 79(5):1698-1703.
Minetti, AE, LP Ardigò & F Saibene 1993
Mechanical Determinants of Gradient Walking, Journal of Physiology 471:725-735.
Moscatelli, U 1998
GIS Applications in the Lower Sangro Valley. Paper presented at the CAA98 conference, Barcelona.
Nackaerts, K, G Govers & L Loots 1999
The use of Monte Carlo techniques for the estimation of visibility. In Dingwall, L et al. (eds), Archaeology
in the Age of the Internet. Proceedings of the CAA97 conference. BAR International Series 750.
Nunez, M, A Vikkula & T Kirkinen 1995
Perceiving time and space in an isostatically rising region, in Lock, G & Z Stancic (eds), GIS and
Archaeology: a European Perspective: 141-152. Taylor & Francis.
Ozawa, K, T Kato & H Tsude 1995
Detection of beacon networks between ancient hillforts using a digital terrain model based GIS, in
Huggett, J & N Ryan (eds), Computer Applications and Quantitative Methods in Archaeology 1994. BAR
International Series 600: 157-162. Oxford: Archaeopress.
Pandolf, KB, B Givoni & RF Goldman 1977
Predicting energy expenditure with loads while standing or walking very slowly, Journal of Applied
Physiology 43:577-581.
Petrie, L, I Johnson, B Cullen & K Kvamme 1995
GIS in Archaeology, an annotated bibliography. Sydney University Archaeological Methods Series 1.
Rajala, UM, AR Harrison & SKF Stoddart 1999
The Enhancement of the South Etruria Survey: GIS in the Study of the Research History of the Southern
Faliscan Area, in Dingwall, L et al. (eds), Archaeology in the Age of the Internet. Proceedings of the
CAA97 conference. BAR International Series 750.
Rajala, U 1998
Using GIS to study territories and rural settlement in south-east Etruria. Paper presented at the CAA98
conference, Barcelona.
Renfrew, C 1986
Introduction: peer polity interaction and socio-political change. In Renfrew, C & JF Cherry (eds), Peer
polity interaction and socio-political change: 1-18. Cambridge: Cambridge University Press.
Renfrew, C & EV Level 1979
Exploring dominance: Predicting polities from centers. In Renfrew, C & KL Cooke (eds),
6 -
20
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
Transformations: Mathematical Approaches to Culture Change. New York: Academic Press, pp. 145-167.
Renfrew, C & EBW Zubrow (eds) 1994
The Ancient Mind: elements of cognitive archaeology. Cambridge UP.
Roper, DC 1978
The Method and Theory of Site Catchment Analysis: A Review, in Schiffer, MB (ed) 1978, Advances in
Archaeological Method and Theory 2: 119-140.
Rose, J & JG Gamble (eds) 1994
Human Walking. Baltimore: Willliams & Wilkins.
Ruggles, AJ & RL Church 1996
Spatial allocation in archaeology: An opportunity for reevaluation. In Maschner, HDG (ed), New Methods,
Old Problems: Geographic Information Systems in Modern Archaeological Research (Center for
Archaeological Investigations Occasional Paper No. 23). Carbondale: Southern Illinois University, pp.
147-176.
Ruggles, CLN & D Medyckyj-Scott 1996
Site Location, Landscape visibility, and Symbolic Astronomy: A Scottish Case Study, in HD Maschner
ed., New Methods, Old Problems. Geographic Information Systems in Modern Archaeological Research
(Southern Illinois University Center for Archaeological Investigations Occasional Paper 23): 127-146.
Ruggles, CLN, DJ Medyckyj-Scott & A Gruffydd 1993
Multiple Viewshed Analysis Using GIS and its Archaeological Application: A case study in northern
Mull, in Andresen, J, T Madsen & I Scollar (eds), Computing the Past: Computer Applications and
Quantitative Methods in Archaeology 1992: 125-131. Aarhus: Aarhus University Press.
Saile, T 1997
Landscape Archaeology in Central Germany: Site Catchment Analysis Using GIS, in Johnson, I & M
North (eds), Archaeological Applications of GIS. Proceedings of Colloquium II, UISPP XIIIth Congress,
Forlí, Italy, September 1996. CD-ROM, Sydney University Archaeological Methods Series 5.
Savage, SH 1990
Modeling the Late Archaic Social Landscape, in Allen, KMS, SW Green, & EBW Zubrow (eds),
Interpreting Space: GIS and Archaeology. New York: Taylor & Francis.
Stancic, Z 1994
Locational analysis and settlement studies with GIS, in Johnson, I (ed), Methods in the Mountains:
Proceedings of the UISPP Commission IV Meeting, Mount Victoria, Australia (Archaeological Methods
Series 2): 73-80. Sydney: Sydney University.
Stancic, Z, J Dular, V Gaffney & S Tecco-Hvala 1995
A GIS-based analysis of later prehistoric settlement patterns in Dolenjska, Slovenia, in Wilcock, J & K
Lockyear (eds), Computer Applications and Quantitative Methods in Archaeology, 1993 (BAR
International Series 598): 161-164. Oxford: Archaeopress.
Stancic, Z & KL Kvamme 1999
Settlement Pattern Modelling through Boolean Overlays of Social and Environmental Variables, in
Barceló, JA et al. (eds), New Techniques for Old Times. CAA 98: Proceedings of the 26th Conference.
BAR International Series 757: 231-23
Stancic, Z & T Veljanovski (eds) 2001
Computing Archaeology for Understanding the Past. Proceedings of the CAA2000 conference. BAR
International Series 931. Oxford: Archaeopress.
Taylor, PJ & RJ Johnston 1995
GIS and Geography, in Pickles, J (ed), Ground Truth. The social implications of Geographic Information
Systems: 51-63. New York, London: The Guilford Press.
Tilley, C 1994
A Phenomenology of Landscape: Places, paths and Monuments. Oxford: Berg.
Tomlin, CD 1990
Geographic information systems and cartographic modeling. Englewood Cliffs (NJ): Prentice Hall.
Tschan, AP, W Raczkowski & M Latalowa 2000
Perception and viewsheds: are they mutually inclusive?, in Lock, G (ed) 2000, Beyond the Map:
Archaeology and spatial technologies: 28-48. Amsterdam, etc: IOS Press.
Van Leusen, PM forthcoming
Exploring Digital Archaeological Landscapes: a discussion of regional map-forming patterns and
processes. PhD thesis, University of Groningen.
Van Leusen, PM 1993
6 -
21
V A N L EUSEN : P A T T E R N T O P ROCESS
Cartographic modeling in a cell-based GIS, in Andresen, J, T Madsen & I Scollar (eds), Predicting the
Past. Computer Applications and Quantitative Methods in Archaeology 1992: 105-124, Aarhus: Aarhus
University Press.
Verhagen, Ph, S Gili, R Micó, & R Risch 1999
Modelling Prehistoric Land Use Distribution in the Rio Aguas Valley (SE Spain), in Dingwall, L et al.
(eds), Archaeology in the Age of the Internet. Proceedings of the CAA97 conference. BAR International
Series 750.
Vita-Finzi, C & ES Higgs 1970
Prehistoric Economy in the Mount Carmel Area of Palestine. Site Catchment Analysis, Proceedings of the
Prehistoric Society 36: 1-37.
Wheatley, D 1993
Going over old ground: GIS, archaeological theory and the act of perception, in Andresen, J, T Madsen &
I Scollar (eds), Predicting the Past. Computer Applications and Quantitative Methods in Archaeology
1992: 133-138. Aarhus: Aarhus University Press.
Wheatley, D 1995
Cumulative viewshed analysis: a GIS-based method for investigating intervisibility, and its archaeological
application, in Lock, G & Z Stancic (eds), GIS and Archaeology: a European Perspective: 171-186.
Wheatley, D 1996
The Use of GIS to Understand Regional Variation in earlier Neolithic Wessex, in Maschner, HD (ed), New
Methods, Old Problems. Geographic Information Systems in Modern Archaeological Research (Southern
Illinois University Center for Archaeological Investigations Occasional Paper 23): 75-103.
Wheatley, D 2000
Spatial technology and archaeological theory revisited, in Lockyear, K et al. (eds), CAA96. Computer
Applications and Quantitative Methods in Archaeology. BAR International Series 845: 123-131.
Wheatley, D & M Gillings 2000
Vision, Perception, and GIS: developing enriched approaches to the study of archaeological visibility, in
Lock, G (ed) 2000, Beyond the Map: Archaeology and spatial technologies: 1-27. Amsterdam, etc: IOS
Press.
Witcher, RE 1999
GIS and Landscapes of Perception, in Gillings, M et al. (eds), Geographical Information Systems and
Landscape Archaeology. The Archaeology of Mediterranean Landscapes 3: 13-22.
Wood, JD 1996
The geomorphological characterisation of Digital Elevation Models. PhD Thesis, University of Leicester.
Woodward, A & R Yorston 1996
Barrows, Ritual, Landscape and Land-use in the Early Bronze Age of Central Southern England, Report
to the Nuffield Foundation.
Yorston, RM 1999
Presenting Archaeological Information with Java Applets, in Dingwall, L et al. (eds), Archaeology in the
Age of the Internet. Proceedings of the CAA97 conference. BAR International Series 750: 255-8.
Zubrow, EBW 1994
Knowledge representation and archaeology: a cognitive example using GIS, in Renfrew, C & E Zubrow
(eds), The ancient mind. Elements of cognitive archaeology. Cambridge UP.
For an accessible review of the underlying theory and method of site catchment analysis, see Roper 1978. For a brief
discussion of spatial allocation, territoriality, and cost surfaces in a GIS context, see Kvamme 1999:174-176.
1
2
For a GIS re-analysis of the same data, see Kvamme 1992.
Part of the following is reproduced from an e-mail sent by Mark Gillings to the GISARCH mailing list (Gillings, Fri. 10 Oct
1997).
3
4
Regrettably, insufficient details were published to allow a proper evaluation of this method.
For those who would like to experiment with alternative cost models, I reproduce these functions here. V is speed, G is the
adjusted slope value. For slopes from –40 to –20 degrees, McDonald (1961) calculates energy expenditure F as follows:
5
F1 = 0.000049V2 - 0.00415V - 0.13276G - 0.004692G2 - 0.00005213G3 - 0.0003257VG + 0.000002036V2G - 0.8588. For
slopes from –20 to +5 degrees, the function becomes:
6 -
22
L INE - OF -S I G H T A N D C O S T S U R F A C E A NALYSIS
F2 = 0.00202V + 0.000021V2 + 0.0256G + 0.00154G2 + 0.000044VG – 0.00000314V2G + 0.3515. For slopes from +5 to
maximum, slope S must be entered in degrees instead of percent, and the function becomes:
F3 = V * (0.00275 + 0.049sin(S)) * cos(S) + V2 * (0.00002 – 0.00033sin(S)) * (cos(S))2 + 0.396 + 0.17sin(S).
After merging of the map layers generated by the functions F1-F3, and adjustment for pixel size, a ‘topographic’cost surface in
kcal results.
This physiological function was apparently derived by Llobera (2000: 71, Figure 2) from multiple sources: Margaria 1938,
Kamon 1970, Minetti et al. 1993, and Minetti 1995.
6
Thus, the cost of ascending a very steep slope to a specific height could be approximated by dividing that steep angle by some
acceptable angle of ascent – say, 15° – and using the outcome as a multiplier for the horizontal costs. For example, the cost of
climbing a 45° slope over a horizontal distance of 200m (i.e., to a height of 200m) is equivalent to 45 / 15 = 3 times the cost of
200m horizontal travelling.
7
Since the calculation of a line of sight between a pair of points is the ‘primitive’operation underlying all viewshed-related
analysis, I have thought it proper to use this term rather than the more widespread 'viewshed analysis'.
8
Although much of the literature treats these two terms as interchangeable, there is an important technical distinction between
them – a DTM contains information about terrain features such as ridgelines, breaks in slope, etc., while a DEM is a simple
rectangular lattice of elevation values.
9
The correction for earth curvature applied to DEM data is: d2 / 1.273 * 107, where d is the horizontal distance in meters to
any point in the study area. A point at 10 km distance would thus be set nearly 8 metres lower, influencing the viewshed
coverage.
10
6 -
23
C HAPTER 7
EDUCATING THE DIGITAL
FIELDWORK ASSISTANT∗
Martijn van Leusen & Nick Ryan
1
1.1
INTRODUCTION
IMPROVING THE EF FICIENCY AND ACCURAC Y OF FIELD WORK PROC EDURES
Modern field walking surveys have become increasingly labour-intensive both because a higher coverage
rate (often with total collection of artefacts) is now thought to be essential, and because more stringent
demands are now put on the precision and accuracy of field recording methods. The management and
analysis of modern survey data within GIS requires the accurate mapping of large numbers of small
collection units so that minor variations in the densities of all material categories can be detected. The
processing of large numbers of, often undiagnostic, finds puts a strain on the ceramic specialists. In some
recent surveys, the overall speed has slowed to as little as 1 hectare per person/day spent in the field;
there is therefore a need to improve efficiency by any means available.
Modern survey practise also requires the accurate mapping of collection units, and the detailed recording
of environmental variables which may influence the finds circumstances (‘visibility conditions’).
Collection units are usually mapped directly onto a large-scale topographic or cadastral map of the survey
area, using existing landmarks for orientation, and a series of forms is normally used to record
information about the collection unit and the finds made in it. Forms are also used to track finds through
the various pre-processing steps and the specialist classification stage into storage. The information is
later transferred from the forms into a DBMS. In current practise, a certain number of errors, omissions,
and illegal entries is unavoidable because maps may be incorrectly interpreted or outdated and hardcopy
forms cannot prevent erroneous or illegible entries; further errors may (and will) arise where the
procedure requires a transcription step. Procedures based on independent self-location and direct digital
data entry should prevent most types of errors from being made.
Finally, modern field surveys are multidisciplinary. Typically, geopedological research and a study of
historical maps and records will take place in conjunction with the field walking. In many cases it would
be helpful if the information collected by each of the participants were available to the others in the field.
For example, archaeologists might want to steer away from areas where the geopedologist has mapped
severe erosion or deposition, and might want to have 19th century maps available in order to correlate
finds distributions and landscape features to the pre-industrial landscape. In practise, this is hardly ever
∗
This chapter was recently published in: Burenhult, G & J Arvidsson (eds) 2002, Archaeological Informatics:
Pushing the Envelope – CAA2001 Computer Applications and Quantitative Methods in Archaeology (BAR S1016).
Oxford: Archaeopress.
7 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
possible because the additional hardcopy maps, if they are available in time, are too cumbersome to carry
about in the field.
1.2
DEVELOPMENT OF A DIGITAL FIELD -WORK ASSISTANT
A potential solution to this problem presented itself in the form of digital Fieldwork Assistants (dFA’s)
based on small handheld computers linked to GPS receivers. One such system that has been developed to
address the requirements of a range of field sciences is FieldNote. This originated in a project that set out
to examine the utility of “context-aware” systems as fieldwork tools in a range of disciplines, including
archaeology and ecology (Ryan et al. 1998). A context-aware system is one that actively monitors its
working environment, or context, and adjusts its behaviour according to changes in that context.
Contextual input may be supplied by the user, or derived automatically from internal or external sensors.
FieldNote includes several modules that work together to monitor context and trigger events under userdefined conditions. Essentially, they provide context-aware and communications services to support a
range of applications designed to meet specific fieldwork requirements. The most frequently used
modules are the LocationTracker and ContextClient. LocationTrackers exist in various specialised forms
designed to work with different proprietary GPS receiver protocols as well as the widely used, but limited,
NMEA protocol. The ContextClient handles network data exchange between handhelds and other
machines.
Note taking and mapping modules FieldNote and FieldMap build upon this infrastructure to directly
support the fieldworker’s recording and navigation needs. FieldNote uses the context services in two ways.
Firstly, when recording information, it automatically tags all records with additional context, including the
current location, derived from a GPS receiver, the date and time, and the user’s identity. Secondly,
whenever the context of a recorded note matches the user’s current context, that note will be ‘triggered’
and its contents displayed on-screen. Although many criteria might be used in matching recorded notes to
the user’s context, proximity of location is most commonly used. When the user approaches the location
of some previously recorded information, it will be brought to their attention.
The notes, together with background maps and other data, may be displayed using the FieldMap program.
This uses information from the context service to keep its display centred on the user’s current location
and to select which map layers (note symbols or cartography) are displayed.
This system had originally been developed for the now defunct Apple Newton, and a similar, but
conceptually simpler, set of programs, the Stick-e-note suite, had been developed and tested on early
Palm Pilot devices. These had been evaluated in several field trials with archaeologists from Southampton
University (Ryan et al. 1999a) and ecologists from Manchester Metropolitan University (Pascoe et al.
1998). These early trials confirmed the essential utility and potential of the system and, in 1999, work
began on a new version of the system based on the experience gained in these trials.
Previous versions had all been implemented using what was, at the time, the most appropriate language
for the target device: NewtonScript for the Newton and C/C++ for the Palm. To ease portability
between different hardware platforms, this new version was written in Java and has been successfully
implemented on a variety of handheld devices including those based on Windows CE and EPOC
operating systems, as well as laptop and desktop Windows, Macintosh and Unix machines1.
The note taking software had previously used either plain text or HTML formats for storage and
exchange between handheld and desktop. Although some experiments in using XML based formats for
1 In practice, we have not achieved the full platform independence potentially offered by Java. The WindowsCE version of
FieldNote used in this campaign was in fact written in waba (see www.wabasoft.com), a Java-like language with its own libraries
and independently developed virtual machine. Despite these differences, a high proportion of the code is common to this and the
pure Java versions.
7 -
2
EDUCATING THE DIGITAL FIELDWORK ASSISTANT
data exchange had been carried out, this aspect of the system was not extensively developed (Ryan et al.
1999b). The new version provided an opportunity to fully implement these areas with the aim of
simplifying the process of exchanging data between handheld field computers and either a remote server,
or a laptop or desktop machine at the survey team’s base.
One of the main strengths of FieldNote is that the user can carry their own and other researchers’
georeferenced data around in the field. Their current location is always available, either as raw coordinates
or as a moving cursor on a map display. This facility becomes particularly useful when there is a need to
confirm earlier work or to re-examine an area of interest located by another member of the team.
Potentially, this might be exploited to provide continuity between successive seasons of work in the same
area.
An opportunity to test the new version of FieldNote came when, in the spring of 2000, a survey team
composed of staff and students from two Dutch universities (University of Groningen and the Free
University of Amsterdam) prepared to conduct a three week survey campaign in October of that year,
near the village of Francavilla Marittima in the Sibaritide region, province of Calabria, in Italy.
Figure 1 - Major landscape units of the Sibaritide. The coastal and alluvial plain (D) is surrounded
by a series of marine and fluvial terraces (A) which merge into the lower slopes of the Pollino and
Sila ranges (B).
1.3
THE SIBA2000 CAM PAIGN
The Sibaritide is a small alluvial plain on the Ionian seacoast, named after the Archaic and Classical Greek
colonial town of Sybaris. It is surrounded by the limestone massifs of the Pollino and Sila ranges, and
7 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
its margins are formed by a series of terraces of marine and fluvial origin mostly composed of pebbly
sands and conglomerates, dissected by the wide pebble beds of several seasonal rivers. Early
archaeological research in the area concentrated on the rediscovery of the Greek colony of Sybaris and its
Pan-Hellenic and Roman successors (Thurioi and Copia), which are now covered by up to 6 meters of
alluvium. Later, archaeological interest expanded to include the larger indigenous settlements and
sanctuaries surrounding the plain, and the University of Groningen became involved in research at the
indigenous hilltop sanctuary of Timpone della Motta and its nearby necropolis (Kleibrink 1993, 2000).
Next to the small number of sites being excavated over the decades, the wider regional archaeological
record was mainly created by in a large-scale survey conducted by Lorenzo Quilici and his teams in the
late 1960s. These surveys, mapped at scale 1:10,000 but available only at the published scale of 1:200,000
(De Rossi et al 1969), were intended to provide the context for the ongoing excavations at Sybaris and its
successors. Largely targeted at the Hellenistic-Roman landscape, these data were more recently
complemented by a regional compilation of pre- and protohistoric sites supervised by Renato Peroni (eg,
Peroni & Trucco 1994).
The SIBA2000 fieldwork campaign forms part of the wider Regional Pathways to Complexity (RPC)
project running at the University of Groningen and the Free University of Amsterdam (for an overview,
see Attema et al. 1998). Its main objective was to assess the quality of the existing archaeological record in
preparation for an analysis of the regional settlement history in the light of processes of centralisation,
urbanisation and colonisation occurring from the Iron Age onwards. Specifically, it was unknown which
fields and areas had been visited by the Quilici teams, and whether the large-scale clustering of sites
visible in the Quilici maps might be related to differences in the accessibility and surface visibility of
agricultural fields in the 1960s. It was decided to approach this objective through a systematic survey of a
representative section of the transitional landscape unit (see figure 1) using both intensive and extensive
methods. This would allow us, on the one hand, to check the recorded size, location, and interpretation
of the sites mapped by the Quilici teams (plus one site mapped by Peroni) and, on the other, to collect the
distributional data needed to establish the presence and nature of any spatial patterning in the extant
archaeological record. A preliminary report on the campaign is provided in chapter 12 of this thesis (Van
Leusen & Attema, in press).
As in previous RPC project surveys, an important objective of the fieldwork was to develop appropriate
and efficient survey methodology for the local circumstances. The Sibaritide was known to be relatively
poorly mapped in a series of 1:10,000 scale map sheets produced in the 1950s, and parts of the survey
zone were expected to contain few if any topographical reference points by which collection units might
be located and sites relocated. Hence the decision to experiment with the FieldNote system developed by
Ryan and his team for efficient and accurate relocating and mapping of fields and features. The system
was expected to be particularly useful in a subsidiary project being carried out at the same time as the
main survey. This aimed to relocate and accurately map protohistoric settlements and caves in the
hinterland and to map subrecent transhumance routes from Francavilla Marittima into the mountains of
the Pollino massif (see figure 2). This project was carried out by a small team under the guidance of local
speleologist/archaeologist Nino Larocca.
Prior to the SIBA2000 campaign, the new version of the FieldNote system had received only limited
testing in familiar environments and by users who were experienced with its predecessors. The XMLbased mechanisms for exchange of data between the handheld devices and base system had yet to be
tested with large volumes of data. Experiments with the archaeological use of earlier versions had all
taken place in relatively well mapped areas and within clearly constrained areas, typically within, or very
close to, the boundaries of large urban sites. The SIBA2000 project provided an opportunity to evaluate
the new system on a significantly larger geographical scale under poorer map control, and with a group of
users who were not familiar with the system or its predecessors.
This campaign also provided an opportunity for the first significant testing of the system following the
decision by the US military to remove the deliberate degradation of GPS accuracy known as Selective
Availability (SA). This change took place on May 1st 2000 with the effect that even simple handheld GPS
receivers improved their apparent accuracy from around +/-80m to better than +/-10m. Previously
7 -
4
EDUCATING THE DIGITAL FIELDWORK ASSISTANT
we had needed either to use an additional receiver for live differential correction, or to record raw satellite
range data and to post-process all GPS measurements against data from fixed base stations in order to
remove the effects of SA. There was a need to evaluate whether the recently improved accuracy
obtainable with simple equipment was sufficient for a range of fieldwork needs.
Figure 2 - The SIBA2000 survey zone centres on the Late Iron Age to Hellenistic hilltop sanctuary
at the Timpone della Motta de Francavilla Marittima, under excavation by the Groningen Institute
of Archaeology since 1991. The locations of the survey areas, GPS tracks of highland survey, and
Quilici sites are indicated. Grid size: 1 km
7 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
2
THE FIELD TESTS
Field tests of the dFA during the SIBA2000 campaign were conducted in a range of conditions, from
extremely mobile recording of movements and observations during a survey of highland transhumance
routes, to the detailed recording of grids during intensive survey. The results of these tests are described
and evaluated in the following sections.
2.1
HIGHLAND SURVEY: THE RECORDING OF TR ANSHUMANCE ROUTES
Since the indigenous societies of southern Italy are assumed to have had largely pastoral lifestyles into the
early Iron Age, the presence of late prehistoric and early protohistoric remains was to be expected both in
the foothills and in the mountainous hinterland of the Sibaritide, and possibly correlating with subrecent
transhumance corridors. In preparation for the design of an appropriate research proposal, the team
decided to test the kit by walking along some of these routes (see figure 2), making digital notes of
archaeological material found along the way while at the same time recording accurate locations for a
number of known highland prehistoric settlements and cave sites. The equipment functioned as expected,
and a large number of observations, including georeferenced photographs, were recorded along routes
running from the foothills near Francavilla Marittima toward the highland villages of Alessandria Carreta
and San Lorenzo Bellizi. These included observations of Hellenistic farmsteads (up to 1000m asl),
junctions of transhumance routes, subrecent structures relating to pastoralism, natural springs, potential
locations for pollen cores, cave sites, and even individual sherds. The GPS trails of these surveys are
depicted in figure 2; figure 3 gives examples of notes taken at such observation points.
Figure 3 - Examples of archaeological FieldNotes. A) section of GPS trace along a transhumance
route, with fieldnote points; B) popup note for Quilici record no. 129; C) a georeferenced
photographic note.
2.2
(RE-) LOCATION A ND RECORDING OF SITE S
The capabilities of the dFA system for wide-area mapping tasks were tested during the SIBA2000
campaign by using it to relocate sites mapped in the 1960s by the Quilici teams, and by tracing agricultural
field boundaries and centroids and circumferences of archaeological sites to a specified accuracy (figure
4). In the absence of detailed up-to-date topographic maps for the area, the latter test was of direct
7 -
6
EDUCATING THE DIGITAL FIELDWORK ASSISTANT
practical utility to the survey team, as it turned out that the infrastructure of roads and tracks had changed
considerably over the years and changes in land use and ownership had resulted in the removal of
microrelief and old field boundaries. Thus, in many cases it was no longer possible to relocate sites
mapped more than 30 years earlier by reference to mapped landmarks only. In contrast, the use of the
dFA as a navigation instrument allowed existing sites (the positions and identities of which had been preloaded onto the kit) to be relocated in a straightforward manner. Both the map layer containing the
Quilici sites and the operator’s current position were marked on the display, and with the kit set to
respond to the operator’s current geographical position, nearby sites were brought to our attention by a
beep followed by a display of the site’s database record.
Figure 4 - Detail, showing disagreements between notional and actual positions of sites. Quilici site
locations as derived from De Rossi et al. 1969 are indicated with 25m and 50m radius;
approximate shape and location of RPC surface scatters is indicated by grey ellipses.
Only a few field and site boundaries were digitally recorded, but this was enough to show that the system
functioned well. However, it was noted that the current procedure for tracing field boundaries on foot is
inefficient, and alternative methods should be explored (see section 3). The digital recording of the
centroids of ceramic scatters using a single GPS reading with an attached note, which only involved
previously tested functionality of the dFA, again was a trivial exercise.
It should be noted parenthetically that the criterion used to draw site boundaries during the SIBA2000
survey was a simple finds density drop-off, and the reason for recording these boundaries was not
analytical but practical, enabling easy and reliable relocation at a later stage. It is recognised that the
concept of ‘site’has come under attack from many quarters in recent years, and that future surveys may
choose to ignore it altogether, preferring instead to record collection units at higher resolutions and
accuracies.
2.3
RECORDING OF TOP OGRAPHIC REFERENCE P OINTS AND COLLECTION UNITS
Doubts about the quality of the available topographic maps and about the possibility of accurately
mapping gridded collection units were the main reason for including measurements of topographic
reference points in the SIBA2000 fieldwork. Separate measurement grids, consisting of square units
approximately 50 by 50 m (0.25 hectare) in size, were set up in preparation to surveying (groups of)
agricultural fields, and cardinal points in each grid were located in reference to these topographic
landmarks. As the survey grids were established using a combination of sighting and pacing methods in
sometimes difficult terrain, we felt it would be a good idea to obtain additional GPS measurements of the
grids; accordingly, in some cases the locations of the canes used to mark the corners of collection
7 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
units were also recorded using the dFA. All grids, landmarks, and GPS positions were mapped on field
maps at a scale of 1:2000. Each of these field maps was later digitised along with its GPS points, and the
latter were used to calculate a simple 1st order georeferencing transformation2.
Although having a large number of GPS points, with attached notes, available during GIS processing did
allow us to resolve some mapping problems, it was found that the procedures described above were
insufficient for obtaining the desired mapping accuracy for many of the collection units. Where a large
number of corners of collection units had been measured by GPS, the collection units could be mapped
directly, without having recourse to transformations of the field maps. But for most field maps only a
small number (from 3 to 8) of GPS points had been taken, and it proved impossible to ‘rubber sheet’
some of these onto the relevant GPS points in a satisfactory manner – apparently because there were too
many internal distortions to the survey grids.
The georeferencing of the field maps brought to light other problems as well; in a few cases GPS points
were so poorly placed for georeferencing transformations that additional points had to be constructed
using plane geometry. In others, the inherent (standard) locational error of the GPS points confounded
our attempts at georeferencing. Clearly, we have to rethink our approach in this regard (see discussion in
sections 2.4, 2.5 and 3.1).
2.4
GPS ACCURACY
During the SIBA2000 survey, we experimented with three types of GPS measurements at different levels
of accuracy:
1. For most readings (eg, field boundaries) speed is more important than accuracy, so we took the
readings on the move or even set the equipment to continuous logging, which typically yields
positions with a standard error better than 7m.
2. For field reference points (topographical landmarks, cardinal survey grid points) our experience
suggests that the accuracy should be increased to 2 to 3m by taking several minutes for each reading.
3. For base reference points the accuracy can be increased even further to about 1 to 2m by leaving the
equipment to record a position for several hours.
Figure 5 shows a plot of GPS measured points collected over a period of nearly six hours at the
Francavilla Marittima museum on the morning of 26th October 2000. The antenna was fixed to the top of
a ranging rod which was then attached to the south-west corner of the railing at the front of the museum.
Measurements were collected at approximately 10 second intervals for a period of about three hours. The
antenna was then moved to the south-east corner of the railing and a further three hour sequence of
measurements was recorded. The observations at the western point were made at a time when the
geometry of the visible satellite constellation was near-optimal, and are probably indicative of the best
that can be achieved with a stand-alone single-frequency receiver. Those at the eastern point include a
period of poor geometry and the re-establishment of position following a failure of the GPS receiver
battery. They are probably more typical of average reception conditions. However, the introduction over
the next few years of additional satellites to provide Wide Area Augmentation Systems (WAAS), primarily
to support improved aircraft navigation, can be expected to bring typical performance into line with the
better results seen here from the western point.
Although a single receiver can be now used to obtain the level of spatial accuracy required for
archaeological surveys, it comes at the price of reduced measurement speed. It is also clear that “mission
planning” software, which predicts the number of visible satellites and the quality of their geometry, still
2
For these operations, we used PC Arc/Info; all other processing took place under GRASS 4.1.
7 -
8
EDUCATING THE DIGITAL FIELDWORK ASSISTANT
has an important role to play. However, the combination of extended point occupation times and the
need to avoid periods of poor geometry, which may amount to one or two hours of the working day,
hardly contribute to streamlining the fieldwork process. For our second and third types of measurements
there remains a case for using differential techniques to improve accuracy without sacrificing speed.
Figure 5 - GPS measurements of the two ends of the Timpone della Motta site museum balcony,
illustrating the accuracy obtainable with a single receiver (Trimble SK8). Ellipses indicate 1 and 2
SD of location. Interval: 10 seconds, duration: 6 hours.
The accuracy of GPS measurements made by a single receiver can be significantly improved using
differential methods (DGPS). In its simplest form, this involves using satellite-receiver distance
measurements taken by another receiver at a known fixed location. Details of the orbital position of a
satellite can be obtained as part of the satellite signal, so it is possible to calculate the difference between
the measured and “correct” satellite-receiver distances. These differences can then be used to correct
measurements made by a roving receiver. This simple technique effectively removes the measurement
errors due to atmospheric effects on the propagation of satellite signals. With two receivers separated by
only a few kilometres, accuracy can be improved to +/-1m or better. More complex techniques involving
signal carrier-phase measurements may be used to improve this to sub-metre levels. All this can be
achieved using inexpensive single-frequency GPS receivers; greater accuracy requires the use of purposebuilt and much more expensive dual-frequency ‘survey-quality’equipment.
Real-time differential correction services are available in many areas. These include freely available signals
from coastal beacon stations, intended for maritime use but often available at a considerable distance
inland. Their main benefit is as a source of reliable correction data for inexpensive GPS receivers
intended primarily for navigational use. Typically, the error in individual position fixes can be reduced to
around +/-3m, increasing with distance from the beacon station. Various commercial services are also
available, but these usually broadcast encrypted signals for which payment of license fees and a special
receiver and decoder are required.
7 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
If live correction is not required, much better results may be obtained by post-processing recorded
measurements against full measurement data from a second receiver. In many parts of the world, there
are stations operated as part of the International GPS Service (IGS, igscb.jpl.nasa.gov) which make
records available on a daily basis. For many areas, this is an invaluable source of highly accurate data.
Unfortunately, the nearest observatory to the SIBA2000 survey area is at Matera, some 100 kilometres to
the north. This distance is towards the upper limit of baseline distances for reliable corrections.
In view of the limited correction quality obtainable by post-processing against data from distant
observatories, a further option would be to operate our own base station at a fixed location throughout
the survey. This would require a dedicated computer and GPS receiver at the project base, recording
continuously whilst survey teams were in the field. Given a maximum range to the edges of the extensive
survey area of no more than 5 km and to the furthest point of the highland survey of about 15 km,
position accuracy relative to the base station of between one and two metres could be expected by this
method. In addition, long duration observations would help to significantly improve the absolute position
of the base station by comparison with data from an observatory such as Matera.
Clearly, for those applications where the accuracy requirements are higher than what can be achieved with
a single receiver, the last option would give the most satisfactory results. It is, of course, still a postprocessing option so would not offer any improvement in the real-time positioning in the field but, as
already mentioned, this does not appear to be a high priority because FieldMap provides sufficient
accuracy for the fieldworkers’location needs. Should it become necessary, corrections could be broadcast
from the base station and received at the rovers by using conventional wireless-modem transceivers.
2.5
SPATIAL ACCURACY AND THE PROBLEM OF IDENTITY
With the advent of accurate GPS location (and even before that with the increasing use of accurate field
equipment) a peculiar problem has begun to haunt archaeologists: conflicts between field measurements
and existing cartography. The position of topographical features as measured by GPS may not agree with
their position as mapped on the most accurate available maps3. In the case of the SIBA2000 survey, this
problem expressed itself in many conflicts between the GPS positions and the 1:10,000 scale topographic
map; since the latter already had a bad reputation we ‘resolved’the conflict by believing the GPS data to
be the more accurate. On the other hand, some disagreements between paced distances on our field maps
and GPS-measured distances could not be resolved at all because they were larger than could be
explained by the standard GPS error. Using differential GPS or taking redundant readings suggest
themselves as potential ways out of such conflicts. From this, and our problems in attempting to
georeference the field maps, we learnt that field mapping methods based on estimates of distances and
bearings are insufficiently precise in the kind of rolling terrain encountered in the Sibaritide foothills.
With respect to the sites mapped by the Quilici teams in the 1960s, the same problem was expressed in a
more archaeologically relevant set of decisions. How much disagreement should we allow between the
notional and measured locations of sites, before deciding that the two are not identical? Given the small
scale of the published map data from which we had to work, it will not come as a surprise to learn that
many Quilici sites were relocated up to 50m from their notional location. With larger disagreements it is
no longer clear whether we have relocated an existing Quilici site, or have found a new one. While efforts
continue to locate the Quilici’s original 1:10,000 scaled field maps we may still hope that some of the
remaining conflicts will be resolved.
This was noted, for example, when geophysical survey grids were very accurately positioned with the help of differential GPS at
the buried Roman town of Wroxeter (Van Leusen 1998).
3
7 -
10
EDUCATING THE DIGITAL FIELDWORK ASSISTANT
3
FURTHER WORK
Development of the dFA is ongoing; further work is needed to improve the functionality of both
hardware and software components, and the user interfaces. The system should be able to deal with a
range of fieldwork tasks under various field conditions, and should be to a large extent configurable by
the user. In the following section we discuss plans and potential for further work on the functionality of
the current dFA kit, on customising it with additional hardware and communication features for survey
work, and on improvements to user and networking interfaces.
3.1
ENHANCING CURREN T FUNCTIONALITY
The functionality of the system as tested during the SIBA2000 fieldwork should be improved by some
fairly straightforward changes to both hardware and software components, as outlined below.
Firstly, current systems are not well adapted to typical Mediterranean field conditions. Although
monochrome screens work well, colour displays are preferred for situations where multiple map layers are
displayed. Unfortunately, most colour displays are dependent on backlighting and are difficult to read in
direct sunlight. Manufacturers have been slow to realise that a mobile device might be used outdoors, but
a solution may be found in some newer PDA models with partially reflective screens intended for
outdoor use.
Data input is typically by an on-screen virtual keyboard or character or handwriting recognition software.
Although the recognition systems are much improved when compared with those available on early
handheld devices, these are still relatively slow processes, particularly for users who have not had
considerable practice. The main interfaces of the FieldMap and FieldNote programs have been designed to
minimise the amount of written input and, with large buttons and other controls, most interaction can be
performed using a fingertip, making the pen almost redundant. This facility needs to be extended to the
options screens, and we anticipate further development based on the Minimal Attention User Interface
(MAUI) devised in earlier work (Pascoe et al. 1999). This employed large input controls that could be
driven by the user’s thumb, thus enabling one-handed control of the software. Simple use of voice
actuation for selections from a constrained list of options is worth investigating as well. Recognition and
automated transcription of voice input is not yet possible on these devices but, with rapid increases in
computational power, may become available in the near future. In the meantime, we will investigate the
use of the built-in voice recording capabilities of the PDA for adding voice notes as an alternative to text
input.
Recording of geometric elements such as lines and polygons was a feature of earlier versions that was
omitted from the new version because of development time constraints. Although the necessary data is
recorded as part of the track log, restoring explicit user control of the geometry associated with a note is
now a high priority.
Software should also be developed to facilitate conducting various types of ‘gridded’ surveys. The
simplest option would be to record the locations of unit corners as they are being set out by the survey
team, but more helpful alternatives should also be made available. For example, the system could be set to
indicate the locations of unit corners at specified intervals and bearings from a given origin, or it could
simply track the ground covered by teams as they work and warn them at specified intervals that a new
unit is required. A major advantage of locating collection units in any of these ways, at least in rough and
poorly controlled terrain, is that the spatial error is non-cumulative. Overall error can never be larger than
that of an individual GPS point – about 5 to 10 m. A second advantage of the latter option is that the
time consuming stage of setting up survey grids can be largely skipped, leading to greater efficiency.
A third area in which the functionality of the system may be improved is in the ease with which data can
be exchanged between kits and with the project’s base computer systems. Properly defined XML-
7 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
coded data formats will ensure that information can be exchanged and downloaded, stored and edited
easily, and can be accessed by web servers and browsers (see section 3.3).
Lastly, a number of minor potential improvements to the functionality of the system were identified
during the SIBA2000 experiments:
• Addition of several simple utility functions, to allow, among other things, on-screen measurement of
distances and bearings using the pen
• The readability of the display may be further increased by the addition of line and area symbols for
monochrome display (eg, dotted lines for grid edges, thick line for paths; dithering for images such as
air photos), and by the use of transparency when displaying icons.
3.2
HARDWARE AND COM MUNICATIONS
If the paper recording trail is to be obviated altogether by the digital recording equipment proposed here,
then one further step should be taken – the use of a barcode reader to link bags of finds to collection
units. Handheld computers with attached barcode readers have been in use in archaeological excavation
and survey since the mid-1980s, and the addition to the system of a barcode reader on a CompactFlash
card is trivial.
A downside of the current system is that it can only record GPS locations of the kit itself, forcing the
operator to walk along the features that are to be mapped. The mapping capabilities of the system would
be clearly enhanced if it were possible to record the locations of distant features as well (eg, suitable areas
for further work, caves seen in cliff faces). The time consuming and often strenuous task of following
field boundaries could also be replaced by either of the following methods:
• by manual on-screen mapping as proposed in section 3.1, using a georeferenced large-scale
topographic map as a reference; in this case the position of the features is estimated; or
• by attaching laser range-finder binoculars, which allow the measurement of distance and bearing.
Commercially available models can be as accurate as +- 2m at distances up to 2000 m.
As the functionality of the digital fieldwork assistant increases, more hardware components are added,
leading to a shortage of ports. Many components require a serial connection so not all of them can be
connected to the PDA at once without some form of intermediate switch; and even if they can, the
additional cables and connectors make the system increasingly cumbersome. Even with a well-designed
harness to keep these under control, there is an ever-present danger of catching cables on trees and other
obstacles. We feel that the most practical solution to these problems lies in the adoption of wireless
(radio) communications between system components. Ideally, each device would contain its own radio
and would collaborate with the others to form a ‘Piconet’or ‘Body Area Network’. The long-awaited
arrival of Bluetooth devices (www.bluetooth.com) which are intended to support short range (<10m)
wireless networking may provide a solution.
An ever-present concern with mobile equipment is battery life. Battery technologies are improving but
this is at least partly offset by increasing power demands as handheld computers become more powerful.
The major limitation here is that most manufacturers design their systems for occasional and brief usage,
whereas field computers are typically used more frequently and for longer periods. Many devices that are
aimed at a consumer PDA market do not have adequate battery life for a full day’s work in the field,
particularly when heavy use is made of their serial interface to receive GPS data. Other similar machines
7 -
12
EDUCATING THE DIGITAL FIELDWORK ASSISTANT
intended for industrial/commercial use are, however, equipped with higher capacity batteries that have
proved equal to our demands.
The power requirements of GPS receivers have reduced considerably in the last few years as
manufacturers strive to develop receivers that can be embedded in other equipment, such as PDAs and
mobile phones. At present, a few receivers are available in PC card format, and smaller CompactFlash
devices may appear in the near future. Whilst these have the advantage of reducing the number of cables
used in the system, they rely on the PDA as a power source and therefore put an additional strain on its
batteries. Whilst we intend to experiment with integrated components such as these, it will probably
remain necessary to carry spare or external battery packs to support the combined load.
Communication for data exchange between handheld and desktop machines typically uses serial or wired
Ethernet connections. For those devices capable of using PC cards, local communication over a range of
about 150m is possible using conventional ‘Wireless Ethernet’cards. As yet, however, no such card is
available in the CompactFlash format more commonly supported by PDAs. Away from the team’s base,
mobile phones provide a suitable communication medium, provided that there is adequate network
coverage in the survey area. In uncovered areas, other devices such as wireless modems might be used.
The main limitation of current mobile phone technology as a data transfer medium is cost. At the time of
the SIBA2000 campaign, data calls on GSM digital network still required the pretence of analogue
transmission and hence the use of a modem. As a result, the exchange of any amount of data, no matter
how small, requires a lengthy negotiation phase as the modem attempts to establish a connection with an
ISP. It is this rather than the low data rates (typically 9600bps) that constitutes the main limitation.
FieldNote transfers often involve sending and receiving only a few hundred bytes, but long connection
times mean that any transfer takes a minimum of about 70 seconds.
Since the SIBA2000 campaign, several telecom networks have begun to roll-out a GPRS service. Whilst
still using the basic GSM technology, this provides a fully digital connection, similar to that of ISDN
systems, albeit at a much lower data rate. The mobile phone can maintain an ‘always-on’IP connection to
a network server and charging is by data volume rather than usage time. Initial experiments with such a
system in the UK suggest that this may become a viable communications medium for future field
campaigns. Over the next few years, further developments are scheduled. The next major advance will be
the so-called ‘third generation’networks which will offer far higher data rates (up to 2Mbps) and the
possibility of live multimedia and video links.
In parallel with infrastructure developments, a convergence of mobile phone and PDA technologies is
under way. Eventually, the question of whether to add mobile communications to our field tools may
become irrelevant as these will be part of the normal function of a PDA.
3.3
USER AND NETWORK INTERFACES
Extending the functionality of the PDA in all the directions suggested above will require some rethinking
of its user interfaces, which will have to allow full configuration of task and display options and efficient
ad hoc switching between tasks and displays. As more intelligent use of the system will require the
simultaneous display of more different types of information and the availability of more options on a
limited screen size, the design of intuitive and effective user interfaces will become essential.
A complex multi-component system will also require extensive configuration. Here we envisage that the
design phase of a typical fieldwork project would include the preparation of configuration scripts which
are uploaded onto each individual kit. An on-screen menu will then give access, firstly, to the list of
available configurations (tasks such as ‘create new grid’or ‘re-locate features’), and secondly, to a list of
the configurable variables for that task (e.g., ‘set grid size’or ‘set alert distance’). The configuration data
7 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
can also specify which options will be available as on-screen buttons and/or as voice actuated options.
Given that the system already contains components for handling data exchange in XML format, and that
XML techniques for managing software configuration are becoming widespread elsewhere (see, for
example, Austin et al. this volume), it is likely that this approach will be used to provide these
configuration scripts.
At present, we see on-line communications between field crews as something of a luxury, but it has the
potential to significantly improve the reliability of the system. If newly recorded data is rapidly mirrored
on a remote server, we can overcome the ever-present fear of losses arising from device failures in the
field. There may also be benefits in making preliminary environmental and archaeological maps available
to crews as they are produced. As the costs of use and features offered by the mobile networks become
more favourable, we anticipate exploring ways in which such connectivity might be exploited, particularly
as the field teams are already carrying mobile phones for voice communications.
Work is also under way to provide a mechanism for automatically generating a web interface to notes,
maps and photographs recorded in the field. This has the obvious benefit of making the field data
accessible in a widely used form but, with on-line field communications, it would enable all participants in
the field, at the campaign base, and at their home institutions to have direct, near real-time, access to the
progress of the campaign. This opens up the possibility of 'remote' specialists taking part in the field
work, for example providing determinations of enigmatic finds.
Further development of the desktop data management tools developed for the current and earlier
versions of the FieldNote system is required. At present, these tools support conversion to and from
appropriate GIS data formats and simple display and editing of data collected in the field. A revised
version of the desktop component should include facilities for managing the configuration scripts
discussed above. We envisage that responsibility for the dFA will become part of the Data Manager’s
task, and that procedures for the acquisition and distribution of data on dFA’s will need to be fully
integrated with the broader survey design.
4
CONCLUSIONS
The experiments described here have confirmed the potential of the dFA system for both speeding up
field recording procedures and reducing the number and size of errors made during the recording
process. The system’s potential for easing navigation and the sharing of information during surveys has
not been fully explored, but our experience in (re-)locating archaeological sites mapped in the 1960's
indicates that it will also prove useful in that area.
With the limited enhancements to functionality discussed in section 3.1, and the further improvements to
the standard accuracy discussed in section 2.4, the system can profitably be used in any type of
archaeological survey. With full technical and procedural integration of a professional version of the kit
into fieldwork methodology, along the lines suggested in sections 3.2 and 3.3, dFAs will begin to
transform fieldwork practise. This will require further extensive testing of system components, software,
and field procedures.
In recent years, the use of professional GPS surveying equipment in archaeological fieldwork has become
much more popular (cf. De Wulf et al. 2000a,b), and some teams are adapting commercially available
products in order to obtain GIS-like capabilities in combination with GPS (Johnson & Wilson n.d.).
These high-powered approaches, while providing very high accuracy and versatility, require considerable
7 -
14
EDUCATING THE DIGITAL FIELDWORK ASSISTANT
expense and highly skilled personnel, and cannot provide a true field information system4. The advantages
of the digital Field Assistant system described here over such alternative approaches can be summarised
as follows:
• Inexpensive - it is possible to fully equip a fifteen to twenty person team for the price of one typical
professional survey kit;
• Immediate feedback - the data collected in the field are made available for use straightaway in a
process contributing to and enhancing the available pre-loaded data, rather than being taken away for
later processing and use. They can also be made context-aware, presenting themselves actively rather
than passively as in field GIS;
• Portability and unobtrusiveness - the equipment weighs very little, will fit in your trouser or vest
pockets, and requires very little training. It is designed to avoid distracting the user from the tasks at
hand;
• Utility - the equipment, as envisaged, performs typical and frequently occurring archaeological
fieldwork tasks such as setting out grids, mapping collection units and points of interest, and
recording finds information.
With respect to the GPS component of the system, the availability of a good location device is a crucial
feature in the recent shift in emphasis of archaeological survey work away from the well-mapped and
well-controlled coastal zones of Italy, to the more rugged and less well-mapped inland zones. For
archaeological applications where the accuracy requirements are higher than what can be achieved with a
single receiver, the addition of a GPS base station for differential correction would give the most
satisfactory results. Post-processing, of course, would not offer any improvement in real-time positioning
in the field but so far we have not identified any reason why this should be a high priority. Should it
become necessary, corrections could be broadcast from the base station and received at the rovers by
using conventional wireless-modems.
REFERENCES
Attema, P. A. J., G.-J. Burgers, M. Kleibrink & D. G. Yntema 1998
Case studies in indigenous developments in early Italian centralization and urbanization: a Dutch
perspective, European Journal of Archaeology 1(3):326-381.
De Wulf, A. et al. 2000a
GPS Surveying Techniques in Archaeology: Topographical Survey of the Thorikos Region (Greece), in
Vermeulen, F. & M. de Dapper (eds), Geoarchaeology of the landscapes of classical antiquity (Babesch
Suppl. 5):189-196.
De Wulf, A. et al. 2000b
Analysis of the Efficiency in Archaeology of GPS Satellite Surveying versus Classical Surveying using
Totalstations: Applications in the Thorikos Region and on the Pyrgari (Greece), in Vermeulen, F. & M. de
Dapper (eds), Geoarchaeology of the landscapes of classical antiquity (Babesch Suppl. 5):197-208.
Johnson, I. & A. Wilson, n.d.
Making The Most Of Maps: Field Survey On The Island Of Kythera.
ESRI offer the field mapping system ArcPad (which runs under Windows CE); Trimble offer the Pro XRS system with Asset
Surveyor and Pathfinder Office software for use on notebook computers. This will view and edit pre-loaded maps and interface
to ArcView for subsequent query, analysis and presentation. Total cost for this system, used by the Australian Paliochora-Kythera
Project to map survey units and landscape/archaeological features and to upload and correct the data in ArcView at the end of
each day, is $12K. De Wulf et al. (2000a,b) investigated the use of survey quality GPS in an archaeological topographical survey
of the Thorikos region in Greece, concluding that it is more than twice as efficient as a total station (EDM) survey but also twice
as expensive at the highest accuracy class; moreover, the processing, quality control, and interpretation of GPS measurements
required a higher degree of skill than was needed with a total station.
4
7 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
Kleibrink, M. 1993
Religious activities on the “Timpone della Motta”, Francavilla Marittima, and the identification of Lagaria,
BABesch 68:1-49.
Kleibrink, M. 2000
Early cults in the Athenaion at Francavilla Marittima as Evidence for a pre-Colonial Circulation of nostoi
Stories, in Gassner, V. et al. (eds), Die Ägäis und das westliche Mittelmeer; Beziehungen und
Wechselwirkungen 8. Bis 5. Jh. V. Chr. (Archäologische Forschungen 4): 165-192. Wien: Österreichische
Akademie der Wissenschaften.
De Rossi, G. M., L. Pala, L. Quilici & S. Quilici-Gigli 1969
Carta Archeologica della piana di Sibari. Roma: Società Magna Grecia.
Pascoe, J., D. R. Morse, & N. S. Ryan, 1998
Developing personal technology for the field, Personal Technologies 2:28-36, August 1998.
Peroni, R. & F. Trucco 1994
Enotri e Micenei nella Sibaritide. Tarento.
Ryan, N. S., J. Pascoe & D. R. Morse 1998
Enhanced Reality Fieldwork: the Context-aware Archaeological Assisstant, in Dingwall, L. et al. (eds),
Computer Applications in Archaeology 1997 (BAR Int Ser 751):269-274. Archaeopress, Oxford.
Ryan, N. S., J. Pascoe & D. R. Morse 1999a
FieldNote: extending a GIS into the field, in Barcelo, J. A., I. Briz & A. Vila (eds), New Techniques for Old
Times: Computer Applications in Archaeology, 1998, Proceedings of the Barcelona Conference, March
1998 (BAR Int Ser 755). Oxford: Archaeopress.
Ryan, N. S., J Pascoe & D. R. Morse 1999b
FieldNote: a Handheld Information System for the Field, in Laurini, R. (ed), Proc. TeleGeo'99, 1st
International Workshop on TeleGeoProcessing, Lyon.
Van Leusen, P. M. 1998
The Viroconium Cornoviorum Atlas: high resolution, high precision non-invasive mapping of a Roman
civitas capital in Britain, European Journal of Archaeology 2(3): 313-325.
Van Leusen, P. M. & P. A. J. Attema, in press
Regional Archaeological Patterns in the Sibaritide: Preliminary results of the RPC field survey campaign
2000, Paleohistoria 42/43 (2000/2001).
7 -
16
C HAPTER 8
T H E R P C F I E L D S U R V E Y S,
1998 – 2000
1
INTRODUCTION
The aim of this chapter is to gather together the significant results, both archaeological and
methodological, obtained through the RPC project surveys conducted in the period 1998-2000. The
surveys conducted by the RPC project have all been aimed at a better understanding of the spatiotemporal distribution of archaeological materials within the three study regions, and have generally been
targeted at marginal landscape units in order to compensate for the fact that previous research has mostly
been concentrated on urban settlements and their immediate surroundings1.
The 1998-9 fieldwork near Ninfa in the Lepine footslopes originally aimed to map Republican rural villa
sites in an area adjacent to that of the earlier Norba survey (King 1995) in an effort to enhance an earlier
topographic survey by Vittucci (1968), but following the discovery of a large number of Archaic sites its
aims were broadened to include sites of protohistoric date. The 1998-9 fieldwork near the Fogliano
lagoon aimed to establish to what extent the coastal landscape of ancient beach ridges was indeed
‘marginal’with respect to contemporaneous developments (centralisation, urbanisation) in the Alban hills
area, as argued by Attema (1993). Whereas the latter area and its fringes showed a clear development
towards urban forms from the late Iron Age onwards, no such development had been visible elsewhere in
the Pontine region. The 1999 fieldwork in the Salento Murge near the town of Ostuni was likewise aimed
at studying an area thought to be geographically and economically marginal to the developments ongoing
in the Salento Isthmus zone, studied intensively by researchers at AIVU since 1980. No rural surveys had
yet taken place in this landscape of LBA and IA centralised settlements and scattered Hellenistic and
Roman farm sites. The 2000 fieldwork east of the town of Lauropoli in the footslopes of the Pollino
range was rather different from previous campaigns, in that it expressly aimed to test aspects of the
distribution of known sites established by Quilici in the 1960s (De Rossi et al. 1969). Specifically, our aims
were to test whether the large-scale patterning of mainly Hellenistic-Roman scatters which Quilici
correlated to a hypothetical infrastructure was not due to the manner in which he conducted his survey;
and to establish the nature of the small-scale variation in the archaeological record in between the sites
discovered by Quilici.
The Ninfa field survey was planned and conducted as part of the Pontine Region Project, precursor to the RPC project at the
GIA, which aimed to map the Republican villa landscape in the immediate neighborhood of the early towns of Segni, Lanuvio,
and Sezze. The preliminary report reproduced here in chapter 9 therefore contains no references to the extensive Italian literature
which is available on the ‘formazione delle città’in this area. Starting with the 1998 campaign near Fogliano, however, field
surveys in marginal landscapes of the Pontine region, the Salento Isthmus, and the Sibaritide were integrated with the existing
research programs at the GIA and the AIVU. The surveys in the Fogliano area served to complement the long-term excavations
conducted by Prof. Kleibrink at the Archaic center of Satricum; the Ostuni survey complemented the intensive diachronic
research conducted by Prof. Yntema in the Brindisino since the early 1980’s; and the Sibaritide survey was part of an ongoing
program of surveys in the surroundings of the protohistoric settlement and cult place at Francavilla Marittima, excavated by Prof.
Kleibrink since 1991.
1
8 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
1.2
BACKGROUND
Interest in the theory and method of field walking surveys, famously introduced into Italy in 1956 by
Ward-Perkins, peaked in the late 1970’s and 1980’s, as can be seen from such edited volumes as those of
the Society of Antiquaries one-day seminar on archaeological field survey in Britain and abroad (1983,
Macready & Thompson 1985), the 1986 Theoretical Archaeology Group conference in London
(Schofield 1991), the 1988 conference “La struttura agricola romana: il contributo della ricognizione
archeologica” organised by the British School at Rome (Barker & Lloyd 1991), the 1989 conference on
archaeological landscapes in Eastern England (Leicester, Parker Pearson & Schadla-Hall 1994) and the
1981 colloquium on archaeological surveying in the Mediterranean area (Athens; Keller & Rupp 1983).
Monographs in this area included those by Shennan (1985) on collection and analysis experiments and
Ebert (1992) on distributional archaeology. Numerous less significant conferences have been held on the
subject, and a count of journal articles would certainly run to more than one thousand.
After what seems to have been a considerable lull due to delays in publication, another two important
volumes of proceedings on the subject of surveying and plough-soil assemblages were recently published
(Bintliff et al. 2000; Francovich & Patterson 1999); these were based, respectively, on an academic session
at the 1996 EAA conference in Riga (Latvia) and on a 1995 symposium organised by the POPULUS
European network. An important recent monograph was Boismier’s (1997, appearing in the British series
of the BAR) PhD thesis on tillage-induced pattern formation. Further reviews and discussions conducted
in the context of a conference and workshops organised by the RPC Project team in 1999 will become
available shortly (RPC forthcoming). In no particular order of importance, the significant trends arising
from these studies are:
• geopedological research has become a required part of regional projects, not only in order to map
one of the most important factors in past land use, but also to map geomorphological bias in the
survey results (erosion, accretion). The OST99 survey demonstrates that some landscapes can be
seriously affected by soil movement and restructuring.
• The increased use, since the middle 1980s, of field digital equipment, DBMS, and GIS as a data
management tool
• The ‘maturation’ of survey as compared to other forms of research - increased complexity of
organisation, concern for (supra-) regional comparison, leading to calls for standardisation. Cherry’s
(1983) work still aimed to show how important survey was; Cambi & Terrenato, in their introduction
to landscape archaeology (1994:151-8), write that much work is still needed before agreement on
basic procedures for survey is reached, and that the problem of visibility had only recently been
recognised, but nowadays many surveyors are well aware of these issues and have begun to be
concerned to be able to compare results.
• The increasing importance of ‘off-site’archaeology and more intensive collection methods. The
general tendency has been for surveys to become increasingly intensive, both in regard of the distance
between walkers and of the amount and type of material recorded, collected, and described. Schiffer
and Wells (1982) tabulated some typical crew spacings of the ‘80s; Cherry (1983) tabulated the
number of sites produced by survey projects over the previous three decades. Although theoretically
rooted in the early 1970’s, this tendency towards intensification only resulted in the advocacy of nonor off-site survey by Mediterranean archaeologists from the early 1990s onwards.
1.3
APPROACHES
The choice of the area to be surveyed was based on physiographical criteria, in some cases modified by
pragmatic and political considerations. The Ninfa survey area needed to be adjacent to the Norba area
surveyed earlier, and was defined essentially on geological grounds by the steep limestone slopes of the
Lepine mountains on one side, and by the tuff ridges of the Alban massif on the other. The Fogliano
8 -
2
RPC F I E L D S URVEYS , 1998 -2000
survey area was again defined geomorphologically, this time by the landscape of lagoons and ancient
beach ridges that forms the coast of the Pontine plain and is bounded on the landward side by the clayey
deposits of the Latina level. The Ostuni survey targeted two areas, one located within the Murge upland,
the other just below the Murge scarp on which an LBA settlement was known to be located. The
SIBA2000 survey targeted a transect through the fluvio-marine terraces which were thought to expose
most of the detectable physiographic and archaeological variation occurring in the northern half of the
Sibaritide region, but here the choice of area was also limited by the concession obtained from the
archaeological superintendency for Calabria. In all cases, current land use at the time of the surveys to
some extent directed our choice of general area – a sufficient amount of accessible tilled land had to be
available - , and precluded surveying in a considerable number of agricultural fields within those areas.
The general approach used during all the surveys was that of intensive rural (or ‘off-site’) field-walking
aimed at total coverage within the survey area. Methods evolved from those previously used by Attema
(which in turn derive from methods developed by the Agro Pontino Survey project, Voorrips et al. 1991)
and quickly became more intensive when it was found that significant small-scale variation occurred in
the archaeological record in between the easily recognisable classical sites. Because it was intended that
GIS be used for the management and visualization of the survey results, the collection of finds and
recording of variables was organised in geographically defined units rather than archaeologically
meaningful units such as sites; the former units decreased in size from agricultural fields (Ninfa survey) to
hectare-sized (Fogliano survey) and eventually quarter-hectare sized (50 by 50 meters) in the Ostuni and
SIBA2000 surveys.
The intensity of the surveys was also increased by the decision to collect all archaeological materials on
the transects walked by each team, as it was recognised that individual team members could not be
expected to classify and count all materials correctly in the field. Since the typical distance between
walkers was 6 to 10 meters, approximately 20% of the ground surface of each unit was inspected and
approximately 20% of surface material was collected. Primary finds processing took place at the survey
base under the direction of team members qualified to classify the material on the basis of fabric and ware
group; more detailed secondary classifications based on form were carried out later on (sometimes after
the field campaign) by appropriate specialists.
These approaches entailed a lowered ‘productivity’of the surveys as compared with less intensive surveys
conducted earlier. The overall speed of the surveys was reduced to just over 1 hectare per person per day.
Various experiments were conducted in an effort to increase the efficiency of the surveys, firstly by
reducing the number of variables to be recorded for each unit (these could be derived from digitised map
layers in the GIS at a later stage), secondly by reducing the number of hardcopy forms needed for the
paper trail of units and finds prior to their entry into computer databases (see Appendix to this chapter),
and thirdly by skipping hardcopy forms altogether and recording unit and finds data digitally using
portable equipment in the field (see chapter 7).
2
RESULTS
To a remarkable extent, the RPC surveys succeeded in uncovering evidence for the existence of preRoman indigenous settlement, which previous research had either ignored or been unable to detect
because of the low visibility of the archaeological materials involved. Hypotheses about the marginal role
of these landscapes with respect to the large-scale processes of centralisation, urbanisation, and
colonisation have had to be modified not only by the direct results of the surveys, but also by our
increased awareness of the biases introduced by the geopedological and land-use history of the study
areas.
8 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
2.1
PONTINE REGION
The Ninfa survey proved that the Lepine footslopes between Cori and Norba were thinly settled from the
later Iron Age onwards, with a large increase in the number of settlements occurring in the Archaic.
Because the assignment of pottery and tile to the post-Archaic period is not yet secure, the fact that
almost all Republican rural villa scatters proved to include Archaic and post-Archaic material cannot be
used to argue for an essential continuity of settlement in this area; rather, it is likely that the increasing
likelihood of raids forced some of the inhabitants to retreat to central and defended settlements
elsewhere. The Roman Republican villas were established on previously used Archaic settlement locations
in the area sometime after the late 4th century BC pacification, as part of the Roman ‘colonisation’of the
Pontine region.
The Fogliano survey uncovered a protohistoric landscape in some respects very similar to that of the
Lepine footslopes. A relatively small number of ?BA and IA settlements was attested, and these may have
been located in promontory-like landscape units in order to take advantage of cooling breezes and clear
views across the small valleys that disect the beach ridges – whether with a view to hunting the wildlife
that would have concentrated there, or to grazing their cattle, is uncertain. The number of sites increases
steadily throughout the Archaic and post-Archaic periods but no nucleation occurs; it is only in the late
Republican period that the area appears to have been ‘discovered’by the Romans, and a rural village
grows up at the site of present Borgo Grappa, possibly in connection with industrial villas exploiting fish
farming along the coast.
Both these surveys show a remarkable cessation of occupation following the early Roman Imperial
period; it is not clear whether this must be explained by a depopulation of the Pontine plain or by a
concentration of the population in large villae and defended burghs such as those attested for the later
Middle Ages. It is thought that the economic significance of the Pontine plain, as a supplier of grain, olive
oil, and other products to the Roman market, was much reduced when the Empire acquired more suitable
lands.
2.2
SALENTO ISTHMUS
The Ostuni ’99 survey resulted, first and foremost, in the discovery that both the areas surveyed
contained a completely unexpected and large number of MBA ceramic scatters, one of which was nearly 6
ha in size. The same material was also found in abundance in the coastal plain during a later (unpublished)
campaign by Burgers, and poses interesting problems of interpretation – for now, a period of shifting
cultivation is proposed to explain the widespread occurrence of the undiagnostic coarse impasto. In fact,
finds dating to the MBA were in absolute majority because both areas contained only a few classical
(hellenistic-Roman) sites. No LBA and only a few possible IA finds were made, but for these periods the
settlement system is known to have been strongly nucleated on hilltops on average about 12 kms apart.
The absence of Archaic and Classical finds indicates that this nucleated pattern continued to dominate
until the area was incorporated into the Greek colonial and, later, Roman Imperial sphere. For these latter
periods, there is a remarkably regular (although thin) pattern of farms present in both the Murge uplands
and the coastal plain, which presumably formed part of the rural hinterland of the town of Ostuni. As in
the previously discussed surveys, no material dating to the high or late Empire was found, and only a few
late antique sherds.
2.3
SIBARITIDE
The SIBA2000 survey, preliminary analysis of which is for the first time reported here in chapter 12,
brought no large surprises to the survey teams in that the results were sufficiently similar to those of the
excellent extensive regional surveys conducted in the 1960’s by Lorenzo Quilici and his team in advance
of the excavations at Sybaris. Our survey was aimed at establishing whether the patterns, densities, and
nature of the archaeological record reported by Quilici would be upheld in an intensive survey of a
representative section of his research area; the quality of his work is evident from the fact that we
8 -
4
RPC F I E L D S URVEYS , 1998 -2000
must make only minor corrections to it. Our work confirms that most of the archaeological record in the
foothill zone consists of poor, and poorly datable, Hellenistic (occasionally Roman) farms; it also showed
that the virtual absence of ‘rural’protohistoric materials is real (although perhaps partly caused by very
low visibility), and again points to a nucleated hilltop settlement pattern – several of which are known to
lie just outside our survey area. Again, no late Imperial or Byzantine wares were found, but these periods
will be the subject of a separate research programme in the future.
In both the Ostuni ’99 and the SIBA2000 campaigns we found ample evidence for the occurrence of
significant local bias factors. In the calcareous area around Ostuni deep soils are so rare, that where they
occur they are often dug away and redeposited eldewhere on the owners’estate, or even sold for use by
other farmers; the ‘shortfall’is made up with large chunks of broken-up limestone which are then covered
up with a thin layer of soil. Such destructive work was evident in many places from the bright red color of
the subsoil as it lay on the surface, with obvious consequences for the archaeological record. In the
Sibaritide, agricultural improvement again caused a bias in the survey, because in many areas the plough
had ripped up and disintegrated chunks of the underlying conglomerate rock. Being full of pieces of
pebbles of many colors and bits of conglomerate cement, it was extremely difficult to discern any but the
most obtrusive archaeological materials in the ploughsoil – possibly contributing to the dearth of reported
non-classical material.
3
3.1
DISCUSSION
FINDS COLLECTION AND PROCESSING
As a point of principle, and because the survey teams consisted of a mixture of experienced and
unexperienced walkers, the policy of the RPC surveys has been to collect all non-natural material
observed in the transect, and not to make any decisions individually about discarding any particular class
of material. Because the surveys took place in marginal rural areas, this strategy only broke down
occasionally as high density scatters of Roman tile and dolium fragments were encountered; in these
cases, smaller representative and diagnostic samples were taken. In addition, team leaders were
responsible for setting a slow pace that allows proper scanning of the surface. It was found that even
under these conditions, great quantitative as well as qualitative differences between individual collections
remained. The general effect of these was throughout to de-emphasise the presence of coarse, earthcoloured and fragmented pre- and protohistoric ceramics and lithics, while emphasising the larger,
brighter, and more obviously artificial (in other words, the classical) finds. This problem could not be
addressed within the context of the RPC surveys but is flagged up here as one that needs much more
attention if collections are to be representative of the surface record. Among the examples cited in the
individual survey reports are the well-known phenomenon of the lithic specialist picking up almost all of
the lithics (Ninfa 98), and highly motivated individuals being the only ones to detect barely visible
protohistoric wares (Fogliano 98, SIBA2000).
Ceramic finds for all the above-mentioned surveys were processed either by, or under the direct
supervision of, Peter Attema, and use a system of classification based primarily on fabrics and ware
groups. Specialists were called in to classify the lithics which were collected more or less as a by-product
(but no less systematically), and those ceramics which allowed closer dating and form description. Where
a typology of the local ceramics had already been produced, as in the Pontine Region, this allowed finds
to be assigned to categories with a fairly restricted period of use (typically a century or two), although
there are still considerable uncertainties associated with some classes and periods. Examples include the
highly worn protohistoric impasto body sherds of the Fogliano survey, which may be ascribed to the
Bronze Age or Iron Age, the Roman coarse and depurated kitchen wares which last from the Republican
into the Imperial period, and various late Republican or Imperial tile categories.
8 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
3.2
DATA PROCESSING
In addition to the question of what data to record in a survey, there is also the question of how to record
and process it prior to analysis and interpretation. In general, the recording practise during the RPC
survey campaigns 1998-2000 documents my attempts to progressively exclude the informal recording of
information and to preclude the recording of ambiguous information – the aim being to ensure that all
forms would be comparable, while at the same time reducing recording and transcription error and
inputting all form data into an RDBMS as quickly as possible. While the starting point in the Ninfa 98
survey was provided by the paper forms that had been in use in earlier survey campaigns by Attema, these
had to be substantially altered to make use of the fact that many topographic variables such as slope class,
relief class, and soil type no longer had to be estimated in the field but could be derived afterward from
appropriate GIS data layers. A second major alteration consisted in relinquishing the individual transect as
a recording unit, and replacing it with an areal unit surveyed by a team of walkers.
Since it was intended for the survey data to be linked to digital maps in a GIS and to be analysed in this
form, further changes became necessary in the system of identifiers being used for collection units
(‘blocks’and ‘sites’) and collections (‘bags’), and in the numeric encoding of variables that were to play a
role in the process of correcting for biases. Finally, it was found that the physical movement of (bags of)
finds through the various stages of cleaning, classification, and storage could only be kept relatively free
of errors if the forms were designed to be passed along with the finds to the next responsible person in
line.
In the end, the SIBA2000 survey teams worked with only two different paper forms: one for recording
information about collection units, the other for information about collections (i.e. the contents of single
bags). The former is sufficiently flexible to allow both the recording of a typical off-site unit and of a
‘site’; the latter ‘bag’form accompanies the bag from the moment it is created in the field through the
processing, and information is added to it at every stage. Finally, only one transcription step is needed
when the form data are entered into a digital database. These forms and the accompagnying user notes,
which represent the latest (2000) stage in the ongoing development of satisfactory field administration
procedures, are included here as an appendix.
It must be admitted here that, although the survey teams and directors were happy with this system, it is
by no means perfect, and relies on thorough instruction and discipline of all those involved in using them.
Inevitably, errors and omissions creep in as the day wears on and people get hot and tired, so it remains
as important as ever to reduce as much as possible the demands made by the forms in the field, while at
the same time ensuring the integrity of the data entered on them. It is evident that these tasks can, in
principle, be better performed by digital forms than by paper forms, and in our last campaigns we have
therefore begun to experiment with the use of handheld field computers or PDA’s (see chapter 7). These
have the advantage that they can present only the relevant fields to the user, can prompt them to fill these
fields out, make it easier to do so by offering option lists, and reject illegal values; they can also be
‘context aware’by automatically providing information such as the current time, date, temperature, name
of administrator, and location (through an attached GPS); finally, they have the advantage that the
information in them can be downloaded rather than transcribed into the RDBMS. An area still to be
explored is that of using such handhelds not only for the recording of alphanumeric unit attributes, but
also for the digital mapping of the boundaries of collection units or even the individual transects walked
by the surveyors, obviating the need for the time consuming practise of laying out measured collection
units in the field. Shortwave radio or satellite phone communications could be used for the purposes of
creating an instant backup of the data thus acquired on the expedition’s computer system.
Once the data are safely within the database, preliminary quantitative processing can produce such
descriptive information as the counts and weights of classes or combinations of classes per collection
unit, which can be transformed into densities and displayed in map form once the boundaries and
identities of the collection units have been entered into the GIS. Since the recording and correction of
biases plays such an important role in my research, I must here go into some detail regarding the
8 -
6
RPC F I E L D S URVEYS , 1998 -2000
correction procedures used. The method developed for the RPC surveys corrects for three factors – area
of collection unit, percent coverage, and estimated total visibility.
Since we used relatively small collection units (typically of 0.25 ha), even relatively minor errors in
mapping their boundaries could potentially have a large effect on the calculation of finds densities,
causing spurious highs and lows to appear in our density maps. We therefore had to take great care in
noting and excluding such features as verges, paths, and gardens intruding on our units, which do not
appear even on the most detailed and up-to-date topographic maps. When I compared the actual total
area surveyed in the Ostuni ’99 campaign as calculated on the basis of our detailed field maps with the
apparent area as derived from field boundaries on the 1:10,000 scale topographic map, the latter turned
out to overestimate by as much as 15% the average surface area of any unit. This was in a relatively wellmapped region of Italy, and indicates that in more poorly mapped areas, or where maps of coarser scale
are being used, this source of error can become even more significant. In the SIBA2000 survey I found
that the actual arable part of some agricultural fields (i.e., the area open to survey) may be less than half
the area contained within its mapped boundaries, because the maps do not show the scrub which usually
takes up all the steeper parts of the landscape. Thus, one cannot rely on sketch maps but needs to
estimate, or better measure, the actual area surveyed.
Since the area open to survey was typically walked by us at an interwalker distance of between 6 and 10m,
depending on the type of land use encountered, the correction must take into account the percentage area
observed (percent coverage). We therefore recorded this factor whenever it deviated from our standard
10m. In order to calculate the percent covered per unit and normalise this to its equivalent of 100%, I
estimated the effective width of the ‘swath’observed on a typical transect to be 2m. This is perhaps a
slight over-estimate but since the same swath width was used in all surveys no errors result from it. Thus,
in a typical case, the percent coverage would be 2 / 10 is 20%, and the correction factor would be its
inverse, or 5.
A similar procedure was used to correct for estimated total visibility, although here the objectivity of our
methods for estimating both the size of the bias factors and their effect on finds recovery may be
questioned (see chapter 5 on bias modelling for details). Three areas for future research have been
identified here: firstly, the need to establish objective measurement scales for bias factors, next, the need
to distinguish the effects these bias factors can have on the recovery of different classes of archaeological
materials; and thirdly, the need to avoid multiplying out of proportion random variations in low density
finds data – the ‘low numbers’problem.
Using this method, the raw counts per finds category are normalised into a continuous variable (densities
per hectare), assuming total coverage and optimal visibility, and displayed as GIS raster map layers for
interpretation. At this point they can be compared both with other data layers in the GIS (generally
holding environmental variables), and with qualitative archaeological data regarding finds and sites
observed during the campaign and in earlier compilations. Although this work could (and perhaps should)
be done during the field campaign itself so as to be able to help ‘steer’the work away from trouble, our
relatively low-budget campaigns did not allow us to do so, and all GIS processing and most database
processing was done at the Groningen Institute of Archaeology after the campaigns ended.
3.3
INTERPRETATION: (RE-) CONSTRUCTING S ETTLEMENT AND LAND U SE HISTORIES
Although each of the RPC surveys has helped to answer the immediate questions posed about each
region, it is to be hoped that their impact will be wider than that. Bit by bit, field surveys are contributing
to a database that should allow us to begin to see similarities and differences through space and time. The
scope of typical field survey data may be limited by its generally low diagnosticity and temporal resolution,
and its interpretation further bedeviled by a host of post-depositional biases, but broad patterns in space
and time tend to emerge quite well from surveys, and their main use is therefore in qualifying the ‘stories’
told of the long-term history of settlement and land use within regions. Thus, they lead away from the
generalising ‘stories’of centralisation, urbanisation, and colonisation processes, towards the particular
8 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
history of the region or even landscape within a region. Yet because of the low quality of the data much
depends on the comparison between data sets collected in the different regions and landscapes, and
therefore on the comparability of the methods by which these data have been collected. The beach ridge
landscape around Fogliano, and the colluvial slopes of the Lepine margin, can only be called ‘marginal
landscapes’in the protohistoric and Archaic periods if some aspect of the survey results – whether this
simply be the quantity of finds or of sites, or qualitative indicators of urbanisation - , can be shown to
demonstrate this. As will be shown in chapter 13, the necessary comparisons between different data sets
lacking the elementary precondition of standardization in definitions and recording methods makes this
an extraordinarily difficult task. At the very least, our research question should now becomes: from the
perspective of which system (in what sense) should these landscape units be called marginal?
Conversely, it is not clear that survey data have much bearing on the ethnicity of the protagonists. Can
Roman colonisers in the Pontine Region be distinguished from their Latin allies and sometime enemies?
Are the majority of the Hellenistic farmers in the Sibaritide likely to have been of indigenous descent or
from the pan-Hellene colony? The material culture of these rural areas is just too poor, and in some ways
too standardised, to distinguish the two purely on the basis of data gathered by fieldwalking survey.
Questions of ethnicity, cultural affinity, etc., can only be studied through excavated evidence of cultural
practise.
4
CONCLUSION
The RPC fieldwork has contributed to the aim of ‘elucidating the complex nature of archaeological
reality’as expressed in chapter 1, by showing that marginal landscape units can have their own settlement
and land use dynamics, sometimes in line with general regional trends but sometimes independent as well,
and that the archaeological record is indeed severely biased in favour of the classical landscape and of
‘high culture’. While aimed at understanding the nature of ‘marginal’areas within the study regions, it has
also been important in providing a sense of the landscape which cannot be obtained from studying maps
and literature alone. The same is true of an appreciation of the nature and significance of the biases
introduced by geopedological processes, historical and present land use and land cover, and the history of
archaeological research in each study area.
Further work will be directed at the intensified exploration of highland economies as begun in the
SIBA2000 campaign, with the attendant development of appropriate methods and techniques to increase
surveying efficiency and quality of the data collected.
8 -
8
RPC F I E L D S URVEYS , 1998 -2000
A P P E N D I X : S I B A 2 0 0 0 F I E L D F O R M S A N D U S E R NOTES
UNIT FORM
A UNIT can be any geographically defined collection unit (such as an agricultural field, measured grids, a
site, or a “string square”); the UNIT form is used to record information about the UNIT itself, and about
the archaeological samples taken from it. Large-scale landscape characteristics (such as land use/land
cover and geomorphology) are no longer recorded for each UNIT individually, but are mapped separately
on copies of 1:10,000 scale topographic maps. If the UNIT is a standard nonsite area, only those
characteristics particular to the UNIT (soil colour, evidence of working, etc.) must still be recorded in the
NOTES box and on the reverse of the UNIT form where necessary. If the UNIT is a ‘site’(however
defined), the mandatory additional characteristics to be recorded are a) the location of the core and b) at
least one contour line for the halo, with density in finds/m2. Recording of the conditions affecting the
recovery of material from the UNIT has been streamlined so that a five-point scale must now be ticked
for each of six factors. There must be prior agreement on the meaning of the scale.
A SAMPLE is defined as any set of finds put together in one finds bag. An RPC standard sample consists
of all non-recent objects found along 10 m interval lines (for a swath width of 2 m, this gives a standard
20% coverage). Additional, non-standard, samples may be collected for a variety of reasons; four nonstandard RPC sample types are recognised:
• Grab sample (Gs) – an unsystematic collection of ‘typical’artefacts aimed at obtaining a quick
impression of the surface material in the UNIT, usually made when circumstances do not permit the
collection of a better sample;
• Diagnostic sample (Ds) – a systematic collection of artefacts selected for their diagnostic value,
usually made in order to obtain a closer dating in cases where the Standard sample was not
sufficiently diagnostic, or where the overall finds density is very high;
• String square sample (Sq) – a total collection of artefacts within a 4 by 4 m area marked out with a
specially prepared rope, usually made on-site in order to obtain accurate finds densities and to ensure
that unobtrusive finds categories are not overlooked;
• Total sample (Ts) – a total collection of all artefacts in the UNIT; usually made in the case of small
and low-density scatters.
Each UNIT form has space for recording three samples, so an additional UNIT form will be needed if
more than three samples are taken from the same UNIT. To avoid any confusion between standard
samples taken on different visits to the same UNIT, samples must always be recorded straightaway (so
the bag number will tell you which sample was taken first); as an additional safeguard, you can include ‘2nd
visit’etc. in the Notes section of each sample record.
At a minimum, the ticket in each bag (written in waterproof marker) should have the UNIT, BAG, and
Sample ID numbers filled out. The bag weight and contents information is included to provide additional
control against loss during processing. The bag itself should also have (written on the outside in
waterproof marker) the UNIT, BAG, and Sample ID numbers.
The page number box on the UNIT form is filled out at the end of the campaign in order to facilitate
annotation of, and referral to, UNITs.
8 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
BAG FORM
Finds bags containing SAMPLEs, after preliminary recording on the UNIT form, are brought back to the
survey base for processing at the end of each day. BAG forms are written out in the field whenever finds
bags are produced, and are passed to the finds processing supervisor along with the bags themselves. The
purpose of the BAG form is to record, in summary form, the contents of a single finds bag, and to
provide the necessary control during all stages of processing. The form is A5-sized and contains
information about a single bag, allowing the forms to be stored in the order of the UNIT/BAG ID
number. The form is not intended for the recording of specialist processing involving individual objects.
Each BAG is identified by its combination of UNIT and BAG numbers, eg 0136-02, which must also
appear on the ticket included in the bag and, for security, written in indelible marker on the outside of the
bag.
Four stages of finds processing are recorded on the BAG form, facilitating the identification and tracing
of bags that go missing at any stage:
• From Field (date) – records the date on which the bag was brought back from the field and became
available for preprocessing.
• Washed/Dried (date) – records the date on which the washed and dried finds were put back into
their bag, and became available for processing.
• Processed (date, person) – records the date on which and the person by whom the finds were
classified, counted and weighed. Non-artefacts and recent artefacts are thrown away at this stage;
broad find categories (lithics, protohistoric to archaic ceramics, classical to post-antique ceramics) are
bagged separately each with a copy of the original ticket (see below) and put into crates for storage.
• Entered into PC (date, person) – records the date on which and the person by whom the processing
information were transferred into the project RDBMS. The data entered is subsequently checked for
errors and omissions by a second person.
The description of the finds during the processing stage follows a classification system which may vary
depending on the region where the survey takes place. The number of sherds larger than 1 cm2 within
each class is recorded, as well as the total weight per class so that average sherd size can be calculated.
Both the condition of the material (weathering, patination, abrasion, rounding) and the occurrence of
diagnostic features is noted on the form.
The writable plastic ticket included with each bag contains spaces for:
• (compulsory) UNIT (4 chars) and BAG (2 chars) numbers; Find category code; Estimated Weight (4
chars)
• (optionally) Date; Sample type (Sq, Ts, Gs, or Ds) and ID (2 chars); Site y/n.
Information regarding the find categories and estimated weight is included on the ticket as a precaution
against the loss or omission of these same data on the UNIT form, and to allow the identification of the
bag in cases where the BAG ID has been recorded incorrectly or incompletely. The GIA address is
included for reference.
8 -
10
RPC F I E L D S URVEYS , 1998 -2000
Campaigna
Teamb
Adminc
Dated
Timee
Weatherf
Page#g
/
ñ
UNIT#h
SIBA 2000
Stonyj 1 2 3 4 5
1 2 3 4 5 Tillage
Dustk
Shadyl 1 2 3 4 5
1 2 3 4 5 Recent
Matm
Veget 1 2 3 4 5
Covern
1 2 3 4 5 Final
Visibo
ï Indicate size of one grid box in meters
% Covq
BAG#
p
Wt.t
Sample
IDs
v
Recent /
Glazed
Fine
Pure
Coarse
Impasto
BAG#
Tile
Stone
u
Lithic
Notes
Sample Ds Gs Sq Ts
typer
w
%Cov
Sample Ds Gs Sq Ts
type
Sample
ID
Wt.
Recent /
Glazed
Fine
Pure
Coarse
Impasto
Tile
Stone
Lithic
Notes
Entered byx
on (date)y
/
Checked by
on (date)z
8 -
/
11
RPC F I E L D S URVEYS , 1998 -2000
UNITaa
Box
Nohh
Find
Category
BAGbb
Count
S?cc
Weight
(g)ii
From Fielddd
Washed / driedee
Processedff
/
/
/
Conditionjj
Bygg
Noteskk
Processing Remarksll
(continue on reverse if needed)
Date into PC
By
/
Date checked
By
/
a
Campaign code, eg SIB2000
b
Team ID, eg 200
c
Initials of team administrator
d
Format DD/MM
e
24 hour format
f
Brief statement such as ‘AM cloudy, PM clear’
g
Page number to be added later on, when forms have been ordered by UNIT#
h
All geographical collection units get a 4 digit UNIT#; administratively therefore there is no difference between a
site (however defined) and a block of land.
i
Other forms; Land Use; Soil; Relief; Levelling etc.; Erosion/Inflation; Drawings; Photos; Unusual legend entries;
Visits, Local informants
j
Stoniness from 1 none to 5 very much
k
Find visibility as affected by tillage (ploughing conditions, harrowing etc), rain and dust, from 1 not affected to 5
very much affected
l
Amount of shade interfering with visibility from 1 none to 5 very much
m
Amount of recent material (post-medieval) affecting find visibility, from 1 none to 5 very much
V A N L EUSEN : P A T T E R N T O P ROCESS
n
Amount of vegetation cover obscuring the soil, from 1 none to 5 very much
o
Final visibility factor (estimate of all other factors combined), from 1 very low to 5 optimal
p
The bag itself gets numbered with the UNIT #, the BAG #, and the SAMPLE ID
q
In the case of a standard sample this is 2 divided by the interwalker distance (assuming a swath of 2 meters), eg
2/10 = 20%
r
Only if % coverage is not filled out. Ds = Diagnostic sample; Gs = Grab sample; Sq = StringSquare sample; Ts =
Transect sample
s
If two diagnostic samples were taken, write DS 2 / 2.
t
In grams, estimated; this serves as a control during finds processing
u
Flint and/or obsidian
v
Add categories as required
w
For example: Includes material from a diffuse scatter of PARCH ceramics, see UNIT 136.
x
Initials required
y
Format DD/MM
z
Format DD/MM
aa
3 digit Unit ID, from ticket in bag
bb
2 digit Bag ID, from ticket in bag
cc
Mark this box if BAG is from a Site.
dd
date (DD/MM) that bag was brought in from the field; must be filled out by team administrator
ee
date (DD/MM) that finds were dried; must be filled out by supervisor of washing team
ff
date (DD/MM) that finds were determined; must be filled out by processing supervisor
gg
Initials of person processing the finds
hh
Box number may be added later if finds are sent to a specialist first
Weight should always be recorded, since it provides a better basis for comparing proportions of different materials
than do the counts
ii
jj
Condition of the finds (eg, abraded, fresh breaks, surface treatment), numeric scale to be determined
kk
ll
For example, regarding condition of finds, wares, dating, etc
For example, “all finds recent & thrown away”, “finds lost during processing”
C HAPTER 9
ARCHAIC SETTLEMENT AND
EARLY ROMAN COLONISATION OF
THE LEPINE FOOTHILLS∗
1
INTRODUCTION
The protohistoric and early Roman settlement history of the Pontine Region were most recently studied
by Attema (1993, 1996, in prep.). The current view is that nucleated settlement seems to have originated
in the Iron Age around the Alban lake (with nuclei such as Ardea, Lanuvium, and Velletri), and to have
developed slightly later on the higher ground around the Pontine plain proper (with sites such as
Satricum, Cisterna di Latina, and Caracupa/Valvisciolo). Many of the Archaic nucleated settlements in
the latter area disappear however sometime during the later 6th century (the late Archaic), and seemed to
be replaced after 500 BC (the post-Archaic) by small dispersed settlement in the volcanic tuff hills only
(Attema 1993:122).
For the colluvial plain deposits in contrast, Attema reports that Iron Age and Archaic materials are very
poorly represented and no discrete protohistoric sites were identified at all in the Cori survey transect,
although protohistoric pottery was found among the predominantly Roman finds, indicating continuity of
some sites (Attema 1993:117-8). A field walking survey south of Sermoneta again yielded very little early
material for the plain, where conditions must have been generally unfavourable for settlement. Again, in a
similar survey near Sezze, finds from the Iron Age and early Archaic were absent (Attema in prep.).
Whereas colluviation may well be implicated in decreasing the visibility of archaeological remains in the
plains, we would not expect this to have such an impact on the western slopes of the Lepine mountains.
Since 1993, several field walking survey campaigns have been conducted on these slopes in order to
establish whether this view is correct.
The regional chronology for the current study departs from the generally accepted chronology (see fig.
9.2) in substituting a post-Archaic period (roughly 500 - 350 BC) for the early Roman Republican
period, as archaeological indicators for the Republican colonising movement (colonies, villas, amphoralike forms, black glaze) appear in the area only after 350 BC. Other important aspects of the regional
chronology are the shift from Late Iron Age orientalising pottery styles (7th century) to Archaic red
pottery, and the shift (around 500-490 BC) from the latter to the post-Archaic, which is visible in the
increased production of pale and orange firing pottery and tiles made from more highly purified clay and
the introduction of more substantial building styles (with heavier roof tiles). It should be noted here that,
in the absence of excavations, site dating in most cases does not permit us to distinguish whether a site
was occupied in the post-Archaic, early Republic, or both.
∗
This chapter was previously published in Assemblage, online journal of the graduate students of the Archaeological Institute at
the University of Sheffield, at http://www.shef.ac.uk/~assem/4/. The fieldwork was conducted as part of a wider project by
the Groningen Institute of Archaeology, aiming to map the Republican villa landscape in the immediate neighborhood of the
Roman colonies of Cora, Norba, and Sezze.
9 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 1: Map of the Pontine Region, with survey areas, from Attema
1993, fig. 2.
Figure 2: Chronological chart of central Italy, after Nijboer 1998,
fig. 1. See especially the column labelled ‘Latial chronology’.
9 -
2
F I E L D W O R K I N T H E L E P I N E F OOTHILLS
2
THE LOWER LE P I N E S L O P E S
According to the established view, it appears there was very little settlement activity on the Lepine slopes
until the later 4th century, when a system of Roman Republican villas, related to the new colonies of
Cora, Norba, and Setia appeared (cf. Attema 1993:233ff). This villa system apparently represents a clean
break with what went before, not just because of the change in building technique and pottery styles and
techniques, but also because large-scale olive tree and vine culture began to be practised. The period 500350 BC apparently brought an abrupt end to the evolution of the Archaic landscape - most likely due to
the troubles associated with the Volsci wars – and was an archaeological ‘dark age’.
2.1
EARLIER SURVEYS
A topographic survey conducted in the early ‘60s in the Cori and Artena map sheet area (some 10 by 15
km) by Paola Vittucci Brandizzi was published in the Forma Italiae series (Vittucci 1968). Little further
work was done until the late ‘80s when, in the course of the Pontine Region Project at Groningen
University, field walking surveys were conducted in adjoining parts of the Lepine footslopes near Norba,
Cori, Sermoneta and Sezze, and at the Archaic proto-urban site of Valvisciolo (see Figure 9.1). This work
was reported on by Attema (1993). The most recent work, leading up to the 1998 survey near Ninfa, was
done near Sezze (Attema 1994 in prep.) and Norba (King 1995). The goal of these two surveys was to
locate and assess sites belonging to a hypothetical colonial system of Roman Republican villas, as
suggested by Attema (1993:233ff and fig. 148; see Figure 9.4). Generally these sites are easily detectable
in the landscape not only because of concentrations of pottery and tile, but also because their platform
retaining walls and associated cisterns, constructed in polygonal dry stone and -later- cemented work,
have withstood the ravages of time until very recently. For those same reasons these villas also figure
prominently in the earlier topographic and desk-based survey by Vittucci. A summary of the results is
presented here.
Table 1 – Descriptions of Vittucci sites in the Ninfa area.
Site
ID
33
34
35
45
46
47
48
49
50
51
52
53
59
Description
villa, wall in opus caementicium, many pieces of tile and amphora especially to the south of the
wall, and reticulate bricks
terracing in 2nd polygonal style
terracing in 2nd polygonal style
nd
cisterna in opus caementicium and wall in 2 polygonal style. Other similar structures are
supposed to have existed nearby but may have been overbuilt. Some remains of coarse
‘grossolana’sherds in the area
villa platform ca. 4m high in 2nd polygonal style and terracing below it. Covered entrance to
platform. Some ceramics and tile found. Lead pipe reportedly found nearby in 1915 but later lost
terracing, polygonal walls, ca. 65m of a longer stretch that was recently destroyed remaining
tile, marble architectural remains, remains of tombe a cappuccina. Site recently disturbed due to
viticulture
top and bottom walls of pedemontana in polygonal work
villa rustica (cisterna in tuff blocks, terracing in 3rd polygonal style), around it numerous tile, imbrex
and vase fragments
cisterna in opus caementicium, some remains of polygonal style terracing below
top and bottom walls of pedemontana, ca 6m wide, just below the modern path
long stretch of pedemontana walls, curving in order to cross a small fosso
villa (3-sided polygonal wall in 3rd style, small cisterna in opus caementicium), ca 20 by 5 by 2.5m.
Numerous limestone architectural fragments reported; reticulate wall reportedly found during
construction of nearby house
9 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
Vittucci’s (1968) map and description of sites in the Cori-Norba area relies heavily on the compilation of
contemporary archaeological records and a follow-up on reports by local inhabitants. It is therefore no
surprise that she reports no prehistoric activity at all, and instead concentrates on mapping Roman rural
villa architectural (cisternae, platform retaining walls, field terracing) and infrastructural (the pedemontana
road, about which more later on) evidence. The table below gives her descriptions, with Site ID referring
to the numbers in Figure 9.3.
The Lepine footslopes just south and southeast of Norba were surveyed by Attema and his team on two
occasions, first as part of the 1987 Norba transect survey (see Figure 1) and again in 1995 in order to
further map possible Roman sites along the pedemontana. Attema noted that, as in the Vittucci survey,
“Roman scatters [are] in all cases related to architectural remains, either polygonal masonry, opus
caementicium finished with opus (semi) reticulatum or a combination of both. The latter are an
indication of continued use of the site from early Republican times into the late Republican and early
Imperial period.” For the protohistoric period however, the evidence he gathered indicated that
settlement began in the later Iron Age and intensified in the Archaic with both a proto-urban nucleus
(Valvisciolo) and a dense distribution of smaller sites developing. Most of the Republican villas and other
Roman sites seem to have developed out of these earlier Archaic settlements and none of the sites
appears to be occupied after the early Empire.
Attema also noted that sites identified as Roman villas are distributed fairly regularly at distances of
about 1 km along the lower slope deposits, suggesting the possibility of a ‘villa colonisation’of the
landscape, directed from the colonial towns. One Republican villa was unusual in that it is situated much
further upslope inside the Valvisciolo gap; Attema suggests this could indicate that there was some level
of hierarchical organisation to the rural villa system, and this site could be an example of a ‘top-level’
villa.
2.2
THE DOGANELLA DI NIN FA SURVEY
In April 1998 a new team led by Attema conducted a further survey survey of the lower Lepine slopes,
this time in the Doganello di Ninfa area which directly adjoins the area surveyed in 1995. This survey
area, of about 5 km by 750 m and dominated by the Monte Arrestino massif (862m asl), is bounded by
the Canale delle Acque Alte (Canale Mussolini) and Fosso del Cavone on the west and north sides, by the steep
uncultivated slopes and cliffs of the Lepine hills, by the area surveyed in the 1995 campaign below Norba
and by the Cori gap in the north (see Figure 9.3). Nowadays it consists mostly of large and small fields
with olive trees, with smaller areas devoted to fruit trees, viticulture, and grazing. The slopes are cut in
two or three places by (nearly) dry gullies, and drainage at the foot of the slope is NW-SE because of the
elongated tuff hill geomorphology there. The area is dotted with small farming cabins and, increasingly,
with modern houses. The two main rubble-metalled tracks running through the area perpendicular to the
slopes, and many minor paths too, are in heavy use and we saw much evidence of fields being worked
with machinery in order to remove stones. This included the removal of terrace walls and remains of
Roman architecture. Increasingly, the steeper slopes are also cleared, ploughed and enclosed.
The team systematically surveyed a total of 27 fields, and discovered or reconfirmed the presence of 20
sites. Figure 3 shows the locations of these fields and sites on a topographical background. Although the
stated purpose of the survey was to map further parts of the Roman system of platform villas, associated
with olive culture and the strategic colonies of Setia and Norba and connected by the pedemontana road,
most of the Ninfa area proved to be rich in Archaic finds, with Roman Republican sites (often without
any platform walls) generally occurring in the same locations. This confirmed that the results of the 1995
survey should not be seen as exceptional and that the Romans could not be seen as colonisers in the
sense that they brought a previously marginally used landscape under cultivation. Instead, a fairly dense
and possibly differentiated Archaic settlement pattern seems to be present, which will force a partial readjustment of current views regarding the settlement history of the area.
9 -
4
F I E L D W O R K I N T H E L E P I N E F OOTHILLS
Figure 3: 1:25,000 scale topographic map of the Doganella di Ninfa survey area.
Fields surveyed by the RPC in green; new Archaic and/or Roman Republican sites
are represented by small red dots with site ID’s in red; black numbers highlighted in
red are sites identified by Vittucci (1968). Just left of center is the larger site S14
that hints at the potential complexity of the Archaic settlement system.
9 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
Table 2 - preliminary site identifications resulting from the Ninfa98 survey.
Site ID
S2
S4
S5
S6
S7
S8
S9
S10
S11
S12
S13
S14
S15
S16
S17
S18
S20
S21
Description
diffuse scatter, early Archaic
Archaic farm to Republican platform villa, some IA material present
small scatter of Archaic sherds
Archaic to Republican farm
Archaic to Republican farm, some IA material present
Archaic to Republican farm
Archaic farm to Republican platform villa, some IA material present
Archaic farm to Republican platform villa
diffuse scatter, Archaic
IA ceramic scatter, no precise date available yet
tomb ‘a cuppucina’, Republican
Archaic to Republican hamlet
Archaic to Republican farm
small scatter of Archaic and post-Archaic sherds
scatter of Archaic and post-Archaic sherds
Archaic and Republican ceramic scatter
Archaic to Republican farm
Archaic to Republican farm
2.3
SETLLEMENT HISTO RY
Ceramic finds were classified according to fabric, ware, and form. Although many of the pottery and tile
types found cannot be dated very closely, it is possible here to present a first rough chronological
description of the settlement history of the Ninfa slopes.
The earliest datable material from the area is a red firing impasto from the early Iron Age, which occurs
as offsite material on a possibly levelled tuff ridge near the Lepine slopes proper. No material from the
middle IA was found. The late IA (phases IVA and IVB, dated about 730-630 BC and 630-580 BC
respectively) is represented by the presence, in phase IVA, of common red slip ware occurring in very
low numbers in the southern half of the study area and, in phase IVB, of a coarse red firing fabric with
FeMn (manganese) temper at many locations. It would seem, then, that LIA sites occur about every
kilometre, with single occasional finds in between. The regular if not very dense distribution suggests
that, as in the 1995 survey area, we may trace the beginnings of the exploitation and settlement of the
Lepine footslopes in the Ninfa area to this period, and especially to the IVB phase.
The Archaic period (580-490 BC) is represented in the study area by coarse red firing pottery with augite
temper, occurring in thick (dolium) and thin (olla) forms. The latter may also date to the post-Archaic,
but as the distribution pattern and even the densities at which both forms occur are very similar this has
no effect on the settlement history of the area. The crude red augite tempered pottery in fact occurs at all
sites and fields surveyed and therefore indicates that a strong intensification of settlement and land use
with respect to the previous period took place (even after correcting for the fact that the Archaic period is
twice as long as LIA phase IVB).
The post-Archaic period (490 - 350 BC, following Attema 1993) is less well recognised in the finds but
must be represented by pale and orange firing augite tempered crude pottery (mostly tile), and by orange
firing coarse ware. We find such materials occurring in fairly high numbers, probably as off-site material,
in most of the area. Nearly all of the Archaic sites also contain a lot of these post-Archaic ceramics,
which suggest that there may have been a continuation of previous land use patterns. This is also
supported by the fact that the lion’s share of the post-Archaic materials concerns the highly visible tile,
whereas the orange coarse ware occurs in comparable densities to the Archaic coarse ware. We therefore
must assume that the only differences between the two periods are an increased use of tile for roofing
9 -
6
F I E L D W O R K I N T H E L E P I N E F OOTHILLS
and a shift in pottery production technology to a more controlled process resulting in lighter coloured
products consisting of a more highly purified clay matrix and augite temper. Presumably the shift in
roofing construction is indicative of a more general Latial change in building construction. The lack of
more dramatic visible changes in the archaeological record for this period of historically attested Volscian
inroads is remarkable.
The Roman middle and late Republican periods (350 – 50 BC) are represented in the finds by a growing
diversity of pottery technology (depurated, reduced, and hard wares) and styles, by the use of lava temper
in tiles, and by the introduction of fine wares such as black glaze. These materials occur at all of the
larger sites occupied in the previous post-Archaic period, but a detailed look shows that there were
changes in emphasis nonetheless - some sites seem to have been abandoned while others show a sharp
reduction or growth in the number of finds. Also, the period we are dealing with has a ‘stratigraphic
advantage’with respect to, and is more than twice as long as, the post-Archaic, and absolute numbers of
finds must be interpreted with this in mind. It would seem, then, that we must interpret the finds from
the Republican period to indicate certainly not a growth in density of occupation - rather a concentration
of settlement onto a lesser number of larger farm sites. In partial support of Attema’s (1993) suggestion
that this period saw the establishment of a system of rural platform villas, we indeed found such
platforms in at least five of the sites.
Again echoing the results of the 1995 survey, we found very little evidence for the continued use of the
area into the Empire, other than the occurrence of small pieces of Terra Sigillata and African Red Slip
ware in some of the Republican sites. Coupled with the fact that we could not identify any full Imperial
or early Medieval sites, this argues for the assumption that the area was essentially abandoned by the early
Empire. However, with Barker (1996:67) we should exercise caution for the later period: “The failure of
many field surveys to locate early medieval settlement effectively is [] the predictable outcome of a
combination of negative factors: a much sparser population, a nucleated settlement system, greater use of
perishable materials in buildings, and of poorly made pottery that is much less precisely dated than
Roman ceramics.”
3
DISCUSSION
In total, the Doganella di Ninfa survey uncovered 16 new site locations, ranging from a single probable
‘tomba a cappuccine’to an Archaic and Republican hamlet, and from the early Iron Age to the late
Republic. No sites dating definitely to the Empire or to the Middle Ages were found. Perhaps most
importantly, it turns out that the whole of the western Lepine slopes between about 60 and 150 metres
high appears to have been intensively used in the Archaic and post-Archaic periods.
3.1
SETTLEMENT PATTE RNS
THE ARCHAIC SETTLEMENT SYSTEM
There is now some evidence for three levels of settlement among the Archaic sites in the western Lepine
slopes. In between the ‘top’level represented by the single large proto-urban site of Caracupa/Valvisciolo
and the ‘bottom’ level of the multitude of small and scattered single-family farmsteads, site S14
represents a middle level of aggregated settlement consisting of a few households without any evidence
of centralised or ‘proto-urban’function – a hamlet.
This system was decapitated with the abandonment of Caracupa/Valvisciolo at the end of the Archaic. It
may be that the incipient urbanisation of the Pontine region, referred to in the introduction, was reversed
by the end of the Archaic as circumstances became less favourable through sporadic warfare, and the
inhabitants resettled into smaller and more easily defended sites on the Lepine scarp. It would have been
9 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
the latter that were targeted for the early (i.e., early 5th century) Roman colonisation reported by the early
historians.
9 -
8
F I E L D W O R K I N T H E L E P I N E F OOTHILLS
CONTINUITY IN THE POST-ARCHAIC?
The existence and nature of any Roman colonisation in the wester Lepine slopes is to a large extent
predicated on the presence or absence of a post-Archaic ‘gap’in the settlement history of the area – the
period called a ‘dark age’by Attema (1993:17). But if there was no settlement continuity, and the area had
been more or less deserted for five or six generations, then how is it possible that the republican farms
are all located on Archaic settlement sites? Certainly dispersed post-Archaic settlement existed in the
Cisterna di Latina area and there is no reason to deny it to the footslopes. We must therefore assume that
there was settlement continuity throughout the post-Archaic, and therefore we must ask what was the
nature of the change from slight buildings and thick augite tempered pottery in the Archaic, to tile-roofed
platform buildings with cisterns and amphora and fine pottery in the Republic. Was the indigenous
population moved away to make place for the new settlers, whose farms were constructed according to
some colonial base plan? Or have the changes visible in the archaeological record occurred over a longer
period, allowing indigenous farmers to take on certain Latin (and later, Roman) habits? Certainly the very
fact that ceramic dating becomes very uncertain during these centuries points to the lack of recognisable
development, and therefore to a disruption to normal production and cultural exchange, that we may
ascribe to the Volsci wars.
THE REPUBLICAN ‘COLONIAL’VILLA SYSTEM
Roman style ‘urbanisation’only begins in the middle of the 4th century with the establishment of formal
coloniae at Cora, Norba, Circeo, Terracina, and Setia. But the establishment of these colonies did not
take place in a vacuum. Control over the landscape and its inhabitants also took the form of a rural
colonisation of which we can most reliably trace the villas. A morphological characteristic linking both
parts of this system is the use of polygonal dry stone walls; physically they are linked by the via
pedemontana and other tracks. The recent surveys seem to indicate that these villas were not as closely tied
to the colonies in a spatial sense as was thought around 1992; rather they seem to occupy all of the
available calcareous soils on the Lepine margin (see figure 9.4). It is possible that closer dating of these
villas may show how the villa system expanded from initial settlement areas around the colonies, to
eventually fill the landscape.
Among the villas found during the survey, Attema’s villa hypothesis (Attema 1993:233ff) suggests that
we may be able to distinguish three groups:
• firstly, a group of platform villas built according to the same plan around the middle of the 4th
century, and strung out approximately 1km apart along the pedemontana. The system might consist
of as many as 20 villas between Cori and Sezze, and each may have a number of associated minor
structures nearby;
• secondly, a much smaller group of ‘controlling’platform villas, which can be distinguished by their
commanding positions and much greater investment in platform architecture. This group would be
contemporaneous with the first; and
• thirdly, a group of non-platform villas occurring in the lowland or lower Lepine slopes, and
developing out of pre-existing Archaic farm sites, or alternatively, ’filling in’available space.
Evidence for the hierarchical organisation of the Roman Republican villa system is intriguing but the
distinction between the three levels will need verification by excavation, which could uncover functional
differences between them. The hypothesis that early platform villas were built in a communal colonial
effort and are located at regular intervals of about 1 km could be further tested in future surveys and by a
careful comparison of the dimensions and masonry styles of the platform architecture.
9 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 4: The Roman Republican ‘colonial’villa system of the western Lepine slopes (adapted from
Attema 1993, fig. 148). Before the recent surveys Republican villas appeared to be spatially tied to
the colonies of Cora and Setia (open symbols); the newly located villas however (closed symbols),
seem to occupy all of the available Lepine margin.
3.2
LOCATIONAL CHARA CTERISTICS
BURIAL
Very little evidence for burial was found by any of the surveys in the area. The nearest known Archaic
necropolis is that of Valvisciolo, and Republican necropoli are of course associated with the Roman
colonies. Possible tombe ‘a cappuccina’(cremation burials) were found in two locations some 1 km apart,
but it seems reasonable to assume that burial took place normally at the edges of each individual
settlement’s core area. Many more graves are therefore likely to exist in the area, but they are very hard to
detect from surface survey alone, as the remains of such burials could easily go unnoticed among the
general off-site noise surrounding the Republican sites in the area. No evidence was found for the
existence of a burial ground on the higher slopes of Monte Arrestino.
9 -
10
F I E L D W O R K I N T H E L E P I N E F OOTHILLS
SPACING AND COMMUNICATION
Although the pedemontana is generally thought of as a Roman construct, forming a system with the towns
and villas built in the mid-4th century, it seems likely that its route and many lesser routes besides had
been in constant use since at least the early Archaic - connecting the large proto-urban settlements while
avoiding the difficult and unhealthy terrain below. There would certainly also have been communication
lines with settlements (both Archaic and Republican) on the Lepine scarp, similar to or identical with
paths that are still in use today.
It was suggested by King (1995:12) that sites are spaced at regular distances (approximately 500m judging
from her map) along a road. Such a pattern could also be observed in the Ninfa area, with villa sites
continuing from S to N along the presumed line of the pedemontana. The distance between large platform
villas is consistently around 1000 meters here, but if we include the other sites producing Republican
material the typical distance between sites again becomes 500 meters or less. One cannot imagine access
to the road to have been of such great importance that sites had to be located very near it; all sites we
found are within 400 meters of the pedemontana line anyway.
HEIGHT AND SLOPE
The natural conditions in the area mean that, without modern farming machinery, only part of a typical
slope could have been used for farming and olive culture. Archaic and Roman Republican sites are
generally located between about 60 and 150 metres asl. Above this point the soil becomes too steep and
stony to be used for anything but extensive grazing by, as is currently the case, goats and horses. At the
lower end of the slope it is less clear that conditions are unsuitable for farming, but we may suggest two
reasons not to situate farms lower than 50 metres asl - firstly, the view over the Pontine and Alban areas
is dramatically reduced, and secondly, the heavy clay soil may not have been workable through the year.
Slope does seem to be a factor affecting many sites - there is a preference for flatter areas (which of
course correlate with height) and often such areas appear to have been man made already in the Archaic.
Platform villas also occur on steeper slopes and, in the case of villa Vittucci 46, significant walls had to
be constructed to retain the villa platform (over 4 metres high). The occupants were rewarded with a
magnificent view ranging all the way from the Monte Circeo and Monti Ausoni in the south to the Alban
volcano in the north – a distance of more than 20 km both ways! It may be suggested that this villa was
one of the ‘controlling’group mentioned above.
SOURCES OF WATER
Distance to surface water is not clearly a locational factor, although recent changes made to the landscape
mean that this aspect must be further studied. Currently wells have water at about one meter below the
surface, which means that early farmers were certainly not restricted to living near surface water.
Unfortunately it is not possible to date any of the numerous wells in the area that are not obviously
modern.
REFERENCES
Attema, PAJ in prep.
Early Roman Colonialism in South Lazio. A survey of three landscapes.
Attema, PAJ 1993
An Archaeological Survey in the Pontine Region. A contribution to the early settlement history of south
Lazio 900 - 100 BC. 2 Vols. PhD thesis, Archeologisch Centrum Groningen.
Attema, PAJ 1996
Inside and outside the landscape. Perceptions of the Pontine Region in Central Italy. Archaeological
Dialogues 3 (2):176-195.
Barker, G 1996
Regional archaeological projects. Trends and traditions in Mediterranean Europe. Archaeological
Dialogues 3 (2):160-175.
9 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
Gaskell, R 1998
The Archaic-Republican Landscape of the Lepine Foothills. Unpublished report, University of Groningen.
King, C 1995
An Archaeological Survey in the Pontine Region, Italy (June 11th - 24th 1995). Unpublished field work
report, University of Groningen.
Maaskant-Kleibrink, M 1987
Settlement Excavations at Borgo Le Ferriere <Satricum> 1. Groningen: Egbert Forsten.
Maaskant-Kleibrink, M and PAJ Attema forthcoming
Pottery technology and the question of pre-urban and early urban transformations in southern Lazio, in
Paleohistoria.
Nijboer, AJ 1998
From household production to workshops: archaeological evidence for economic transformations, premonetary exchange and urbanisation in central Italy from 800 to 400 BC. Groningen : RUG, Department
of Archaeology.
Salmon, ET 1969
Roman Colonisation under the Republic. London: Thames & Hudson.
Terrenato, N forthcoming
The visibility of artefact scatters and the interpretation of field survey results, in: R Francovich and H
Patterson (eds), Methodological Issues in Mediterranean landscape archaeology. Artefact Studies.
Oxford.
Terrenato, N 1996
Field survey methods in central Italy (Etruria and Umbria). Between local knowledge and regional
traditions. Archaeological Dialogues 3 (2):216-230.
Terrenato, N and AJ Ammerman 1996
Visibility and site recovery in the Cecina valley survey, Italy. Journal of field archaeology 23:91-109.
Verhoeven, A 1991
Visibility factors affecting artefact recovery in the Agro Pontino survey, in: Voorips, A, S Loving and H
Kamermans (eds), The Agro Pontino survey project, methods and preliminary results (Studies in Praeand Protohistorie 6). Amsterdam.
Vittucci, P Brandizzi 1968
Cora, Forma Italiae regio I, Vol 5, Roma.
Zaccheo, L and F Pasquali 1972
Sezze dalla preistoria all’età romana, Historia Selecta Setina 1. Sezze.
9 -
12
C HAPTER 10
A MARGINAL LANDSCAPE:
FIELD WORK ON THE BEACH
RIDGE COMPLEX NEAR FOGLIANO
(SOUTH LAZIO)∗
PETER ATTEMA, ESTHER VAN JOOLEN & MARTIJN VAN LEUSEN
ABSTRACT: This paper reports on the results of archaeological and geographical field work conducted in 1998 and
1999 by members of the Regional Pathways to Complexity project in the coastal landscape of South Lazio (Italy).
The protohistoric (Bronze Age to Archaic) settlement history of this area is dominated by its marginal position with
respect to the proto-urban settlements of the Alban hills and the Lepine margin, and in the historic period (postArchaic to Roman Imperial) by the strategic and economic interests of Rome. A preliminary reconstruction of the
dynamics of both the physical and the human landscape is given for both periods. The interpretation of the results
of the field walking survey and the reconstruction of potential protohistoric and Roman land use are both informed
by an analysis of recent changes in the physical environment – especially the Bonificà carried out in the 1920s.
KEYWORDS: Archaeological survey, land evaluation, central Italy, Pontine region, protohistory, Roman period
1
1.1
INTRODUCTION
MARGINAL LANDSCA PE UNITS IN THE RPC PROJECT
The Fogliano pedological and archaeological field work reported on here is part of the research program
of the Regional Pathways to Complexity (RPC) project conducted jointly from 1997 to 2001 by the Groningen
Institute of Archaeology and the Free University of Amsterdam. The RPC project studies landscape and
settlement dynamics in three protohistoric Italian regions with the processes of centralisation, early
urbanisation and colonisation as its main themes (Attema et al. 1998).
Whereas in all three regions much fieldwork has been carried out in the past decades to tackle these
themes, such work has invariably neglected the marginal areas in the landscape. By ‘marginal’we mean
those landscape units that, on account of their environmental characteristics (low fertility, inaccessibility,
and distance from core areas), were not especially favoured for permanent settlement during the protoand early historic periods. In spite of their correspondingly less-than-impressive archaeology, we feel that
these parts of the ancient landscape should not be excluded from a regional study. On the contrary, they
should be considered integral parts of the ancient human landscape because they may have a specific role
in the economic system, may constitute a potential zone of demographic expansion, or may function as
an area of refuge.
Having been thinly settled in the past, and put to low intensity use, marginal areas pose particular
methodological problems on account of the low density of surface material that constitutes the
archaeological record. A total of four weeks of survey by the RPC team in the lagoonal environment near
∗
This chapter was previously published in Palaeohistoria, the annual journal of the Groningen Institute of Archaeology.
10 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
the lake of Fogliano in the Pontine Region in the summer of 1998 and the spring of 1999 were aimed at
assessing the influence in this marginal landscape unit of the early urbanisation process (7th and 6th
centuries BC) and of Roman expansion in the subsequent period (5th to 4th centuries BC and later). It was
found that the area attracted substantial settlement only in an advanced stage of the Roman Republican
and early Imperial periods, while featuring continuous but sparse human presence during all earlier
periods. We tentatively link this late Republican and early Imperial settlement increase to a socioeconomic development very specific to these coastal margins - the establishment of wealthy maritime
villas that managed large marine fishponds, as well as a substantial pottery industry exploiting the high
quality clay banks found along the coast and now exposed by the sea.
1.2
OUTLINE OF THE P HYSICAL AND HUMAN LA NDSCAPE OF THE PONTI NE REGION
In order to relate the findings of the archaeological surveys of the RPC project to the characteristics and
dynamics of the contemporary landscape, physical geographical mappings are carried out in combination
with the surveys. Such fieldwork took place in conjunction with the archaeological survey, in the area
between Borgo Sabotino and Borgo Grappa, the coast, and the beach ridges of the Minturno level. The
primary aim of this study was to check the detailed (scale 1:25,000) soil maps of this region by
Kamermans et al. (1979) and Bouman and Rot (1982), and to compile these into one map. A subsidiary
aim was to create a soil unit description according to the guidelines of the Food and Agriculture
Organisation (1977), in order to do a land evaluation for agriculture from the late Bronze age till Roman
times, reconstructing the potential suitability of a specific land use type for a specific physiographical unit.
In this study only soil properties and characteristics relevant for early agricultural land use have been used,
which is reflected in the criteria used in the legend. A final aim of the study was to reconstruct the
landscape by examining the units, and especially the recent anthropogenic influences, which drastically
changed this landscape.
Figure 1: The location of the Fogliano field work
area within the Pontine region. The four major
physical geographical regions are outlined on a shaded
elevation model.
The landscape of the Agro Pontino has been fully described by Sevink et al. (1982, 1984) and Kamermans
(1991). In broad outline, it consists of four physical geographical units (see figure 1):
1) limestone mountains (Monti Lepini and Monti Ausoni), folded during the Middle and Late Miocene;
2) tuff hills originating from the Volcano Laziale which was active between 700,000 and 45,000 BP;
10 -
2
F I E L D W O R K O N T H E P O N T I N E C OAST
3) a graben formed in the Plio-Pleistocene as a result of vertical movements along NW-SE running fault
lines, subsequently filled with fine textured and often organic sediments and draining toward the
south-east; and
4) a horst system along the sea coast, consisting of four sandy clayey marine terraces with a local aeolian
cover, developed as a consequence of world-wide sea level rises, although locally other factors played
a role. From oldest to youngest the terraces are named Latina level, Minturno level, Borgo Ermada
level and Terracina level (Sevink et al. 1984). This system as a whole drains to the south-west.
The Minturno level has been dated by fission track, K/Ar and amino acid racemization to about 125,000
years BP. During the next sea level rise in the early Würm (about 90,000 years BP) the Borgo Ermada
level was formed (Kamermans et al. 1991). Only in the Holocene were the beach ridges of the Terracina
level developed, incised later by rivers filling up the valleys with fluvial and marine sediments (Bouman
and Rot 1982). All units have been locally covered by aeolian sands during dry phases from the Würm to
the early Holocene. As part of the fight against the malarial mosquito the Lago di Fogliano was partly
deepened and salted, partly filled in during the 1930s (J. Sevink, pers. comm.). Surplus sediment was also
dumped around the lake and further inland, these units being classified as anthropogenic.
The settlement dynamics pertaining to the protohistorical and early historical period in the Fogliano
survey area should be evaluated in the light of developments in the core area of Latium Vetus, i.e. the
Alban hills including Rome and its environs (Attema 1993). The major developments in this area can
conveniently be presented as three distinct settlement phases, each implying an exponential growth in
agricultural, building, and industrial activities affecting ever more landscape units, including marginal areas
such as Fogliano (see figure 1).
1) Centralisation of settlement during the Bronze age and early Iron age (ca. 1500 - 700 BC) featuring a
very gradual infill of the volcanic landscape of the Alban hills and the core area around Rome, and
marginal presence on the slopes of the limestone mountain range of the Monti Lepini;
2) Proto-urbanisation during the late Iron age and Archaic periods (ca. 700 - 400 BC), a process which
saw the formation of early towns, and which included the growth of a rural landscape along the
slopes of the Monti Lepini and along the ancient beach ridges into the Pontine region proper;
3) Romanisation of the proto-urban landscape and full colonisation of the Pontine plain (400 - 100 BC),
a process which gradually also began to affect the more marginal areas, such as the beach ridges along
the coast of which the Fogliano survey area forms a part. This phase also sees the growth of
industrial activities.
Following this brief sketch of the regional context of our research, the remainder of this article will be
used to describe the physical geographical units and their agricultural potential (section 2), to discuss the
archaeological results and some problems in their interpretation (section 3), and, lastly, to consider how
our research may influence our interpretation of the regional context (section 4).
2
EV ALUATING THE AGRICUL TURAL POTENTIAL OF T HE LANDSCAPE
Evaluating the agricultural potential of the protohistoric and Roman landscape of Fogliano is a three-step
process. First a physical geographical map of the area is made; then follows an assessment of any
substantial changes the landscape may have undergone through natural and human agents; and finally an
agricultural land evaluation is carried out. Figure 2 shows the physical geography of the study area as
mapped by the RPC project. Table 1 summarises the process by which these units were constructed from
two earlier partial mappings. In the following paragraphs each legend unit is briefly discussed and a
preliminary land evaluation is given. A complete land evaluation of the Pontine region is being prepared
by Van Joolen (in prep.)
10 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
2.1
DESCRIPTION OF T HE UNITS
Beach ridge unit. The beach ridges of the Minturno level (M1) and those of the Borgo Ermada level (B1 to
B6) can be classified into the same unit for land evaluation; despite the fact that elevation levels can differ
significantly (respectively 13 m and 6 m asl), the texture, soils and drainage class are more or less the
same. The ages of the different beach ridges are not relevant for the land evaluation, because in all units
more or less the same soils have developed. However, since the older beach ridges had a longer period
for soil development, they have a more clayey texture and more chromic properties (darker colouring).
Lagoonal unit between beach ridges. This unit consists of fluvial and / or lagoonal sediments. The unit is
relatively narrow (less than 150 m). North of the Strada Litoranea the units B7 and B9 (deposits between
and alongside the beach ridges, Kamermans et al.) and the units B4 (valley-units, Bouman and Rot)
coincide. South of the Minturno beach ridge M1 the unit B9 (Kamermans et al.) coincides with M4 of
Bouman and Rot. South of the Strada Litoranea the units B7 and B9 of Kamermans et al. coincide with
unit B6 and T4 (Bouman and Rot). The aeolian unit B6 could not be differentiated from the beach ridge
deposit B1; it lies in the same position as the B7 and B9 deposits and has the same texture. Unit T4 just
west of the village of Fogliano could not be differentiated from the beach ridge unit B1, and is therefore
classified alike.
Level lagoonal unit. Despite the fact that Bouman and Rot classify T7 as anthropogenic, it seems justified to
consider T7 similar to T2, since it has the same position in the landscape and (probably) the same fluvial
genesis.
Figure 2: The physical geography of the Fogliano field work area. Fields outlined in bold were
surveyed by the RPC project in 1998-9; those outlined in grey were surveyed by the Agro Pontino
Survey project in the 1980s.
Aeolian unit. North of the western side of Lago di Fogliano the aeolian units (B6) of Bouman and Rot are
classified as lagoonal between beach ridge deposits, because the texture (clayey loam) resembles that of
the lagoonal deposits. North of the Strada Litoranea the aeolian deposits are classified as beach ridge unit.
Anthropogenic unit. All anthropogenic units of Bouman and Rot are classified as level lagoonal unit, except
for the one north of Canale Allacciante at the western side of Lago di Fogliano, for which insufficient
information was available. The anthropogenic unit B13 of Kamermans et al. is considered to be a beach
ridge unit, because it forms a well drained elongated ridge in the landscape at an elevation of 10 to
10 -
4
F I E L D W O R K O N T H E P O N T I N E C OAST
11 m above sea level. On the map the other anthropogenic units from which the genesis could be
reconstructed are hatched.
Units
Kamermans
et al. 1979
B1 to B6;
B13, M1
T4; B7; B9;
B8; B10 to
B12
Units
Bouman &
Rot 1982
B1; B6; M1
Units Van
Texture
Joolen (this
article)
Beach ridge unit Sandy clay
loam or sandy
loam
T4; B4; M4 Lagoonal unit
Sandy loam or
between beach clay
ridges
T2; T7; B2 Level lagoonal Sandy clay or
unit
clay, sand on
clay
B6
Aeolian unit
Fine, wellsorted sands
T7
Anthropogenic
unit
-
Structure
N/A
Soil classification
(FAO 1977)
Chromic, Orthic,
Albic and Gleyic
Luvisols
Cracks Chromic Vertisols,
Gleyic and Orthic
Luvisols
Cracks Solodic Planosols,
Albic Luvisols,
Pellic Vertisols,
Fluvisols
N/A Eutric Cambisols,
Gleyic, Orthic and
Chromic Luvisols
-
Drainage
class
Elevation Slope
(m asl)
class
4-6
2 - 13
Gently
sloping
2
2-5
Almost
flat
1-2
5 - 10
Flat or
almost
flat
4-5
Variable
Gently
sloping
-
-
-
Table 1: Description of new legend units on the basis of the partial mappings by Kamermans et al.
(1979) and Bouman and Rot (1982). N/A = Not Applicable; asl = above sea level. Drainage classes
according to FAO (1977): class 1: poorly drained; Class 2: imperfectly drained; Class 3: moderately well
drained; Class 4: well drained; Class 5: somewhat excessively drained; Class 6: excessively drained. Slope
classes according to FAO (1977): 0 - 2% flat or almost flat; 2 - 6% gently sloping.
2.2
PRELIMINARY LAND EVALUATION
The method of land evaluation compares the soil requirements of different land utilisation types with the
land qualities and characteristics of the described units, to determine their agricultural suitability. The
results of our preliminary evaluation are given here.
Beach ridge unit. Because of the medium textured sediments and the drainage varying between well drained
and excessively drained, the soils of this unit need fertilisation (by manuring) and if needed irrigation to
cultivate crops. Such practices were not in widespread use in the Bronze age, but were used in Roman
times (Spurr 1986). In the latter period millet (“can grow almost anywhere”, Spurr 1986), fodder crops,
and vines were best adapted to these circumstances in a mixed farming rotation system.
Lagoonal unit between the beach ridges. Because of its clayey texture this unit may have been relatively difficult
to cultivate. The deposits may have been suitable for small scale agriculture of wheat or grass for hay,
together with dry and wet meadows. During the Bronze age the unit probably was not attractive because
of the heavy texture and possibility of waterlogging and may have been only used for transhumant grazing
of cattle, but in Roman times the biennial system of wheat cultivation followed by fallow was possible. It
could also be used for long and short fallow and pratum (maintenance of dry and wet meadowland, Spurr
1986).
Level lagoonal unit. Because the presence of lagoonal sandy clay or clay at shallow depth (less than 45 cm),
which acts as an impermeable layer, causes waterlogging and / or cracking of the soil, cultivation of
crops is nearly impossible on this unit. It is only suitable for the cultivation of hay from dry or wet
grasslands or as grazing lands for cattle. Pascuum (long fallow, Spurr 1986), pratum, short fallow, and
transhumance may have been practised in both Bronze age and Roman times.
Aeolian unit. Because of the high fertility (fine, well sorted sands) these deposits are suitable for a wide
variety of crops. The unit is flat or nearly so, and would have been suitable for cultivation of barley, millet
and vine, perhaps with occasional irrigation in dry periods, in both the Bronze age and in Roman
times. Usage of these soils for grassland would not have been logical. So cultivation practices such
10 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
as mixed farming of barley and vine in both periods would have been possible, also because of the easily
workable soils for manual tools and light ploughs.
3
THE ARCHAEOLOGICAL S U R V E Y S
3.1
METHODOLOGY
BIASES INHERENT IN THE ARCHAEOLOGICAL SURVEY
As always, a discussion of the results of a field walking survey must begin with an assessment of the
factors that bias the samples that were taken. In the Fogliano area these may be grouped under the
headings of visibility factors, post-depositional processes, and surveyor biases.
Artefact visibility was generally low to very low throughout the 1998 Fogliano survey, due to the dusty
conditions and intense sunlight. Excepting in a very small number of cases, the fields had been dry for
several weeks and had apparently been harrowed several times so that a 10 – 15 cm thick top layer of dust
resulted. The direct sunlight also caused large differences in artefact visibility during the day, because of
the variable angle of the sun above the horizon (altitude), its angle to our walking direction (azimuth), and
the sharp shadows thrown by high crops and trees. A control survey conducted under better conditions
in the spring of 1999 confirmed the importance of these visibility conditions, especially for pre-and
protohistoric sites. It appears that even our intensive surveying could not consistently identify the Bronze
and Iron Age material present in the study area in low numbers. With Archaic and post-Archaic ceramics
the danger is slightly less because the sherds can be more easily distinguished (by size, color, and texture)
from the soil matrix; and with Roman Republican and Imperial material visibility did not appear to be a
factor at all. Artefact visibility bias has therefore mainly affected our observations of protohistoric
ceramics.
A second factor affecting artefact collection are the post-depositional processes. Three centuries of Roman
agricultural exploitation of the area during the late Republican and early Imperial period (200 BC – AD
100) constitute the first major post-depositional factor in the study area, significantly reducing our
chances of finding pre-Roman ceramics. In addition to this, the Bonifica Integrale - the wholesale land
reclamation of the Pontine plain brought about by the Italian government in the late 1920’s and early
1930’s – finally brought about the long wished-for transformation of the natural landscape of the
Fogliano area into farming land. The characteristic landscape of beach ridges disected by natural streams
disappeared under the plough, and the lagoons of Fogliano and Dei Monaci were reduced in size and
partly embanked. Interventions in the hydrography of the area had begun on a very small scale some 15
years earlier, as part of local private enterprises. Until then the size of the lagoons, having no natural
outlets to the sea, was variable and depended on the water supply by rivers from the interior, such as the
Rio Cicerchia (since regulated) and the Bracciolo (since filled in).
In many areas peaty and clayey soil was added to improve the less fertile sandy fields, while sandy soil may
have been added both to improve workability and to raise the level of some of the lower-lying heavy
clayey fields. Many other fields, especially those on ridges, may also have been levelled to make them
more easily workable with modern farm machinery. Given the original relief of the area, marginal
increases of the ground water level and of the water level in the lagoons will have resulted in significant
decreases in accessibility, with marshes and lagoons closing in on the sandy ridges, as indicated by the
presence of clayey deposits in between these ridges. This will have had great significance for the use of
the smaller landscape units in antiquity. An analysis of the relief maps produced for the Bonificá of the
1920s is currently underway (Feiken & Van Leusen, forthcoming) and may in future allow us to correlate
water level rises with the location of the archaeological sites in the protohistoric and Roman landscape.
Taken together, these alterations may have significantly and non-randomly lowered artefact visibility in
the area. However, no precise record of such activity was kept, and we therefore must rely on our own
10 -
6
F I E L D W O R K O N T H E P O N T I N E C OAST
observations and on the memories of the local peasants.
The third and last factor influencing artefact recovery are inter-walker differences in training and in visual
acuity. To minimise the effect of these, walkers were instructed to recover all flint and ceramic objects
encountered; but of course the differences in experience did express themselves in one walker
preferentially finding flint, and another mainly red, black, and orange wares. Whilst differences between
the surveyors generally did not affect the highly visible Roman Republican finds, the older protohistoric
ceramics were often caked in sand and had approximately the same size as many of the natural sand
clumps produced by the fine ploughing in the area; and so it was quite easy to miss individual shards and
even small scatters.
Because of the importance of these three visibility factors, our analysis of the survey results must take
them into account. Our field recording methods allowed us to do this with two factors (dust conditions
and vegetation cover, and soil addition/levelling) but not with inter-walker differences. In addition, we
attempted to control and quantify biases by re-surveying representative fields and by comparing our
results to those obtained earlier by the Agro Pontino project (Voorrips et al. 1991).
All recovery factors identified in the Fogliano surveys have tended to bias results away from the
protohistoric periods. The re-surveying of selected fields in the Fogliano survey area has indeed shown
that the visibility of proto- and early historic sites is highly dependent on survey conditions, land use and
weather conditions being the two most important factors. From our comparison of the results obtained
on the ploughed but very dusty fields of August 1998 with those obtained on the recently planted but
moist fields of April 1999, two major consequences follow for the interpretation of settlement history.
Firstly, the find of even a single sherd of pre-Roman ceramics is likely to indicate the presence of a small
site. Given that this single sherd is relative unlikely to be picked up even in an intensive systematic survey
such as was used by the RPC project (with typically about 30% coverage), we must assume that our
surveys can only identify some of the Bronze Age and Iron Age sites present; for Archaic and postArchaic sites this danger is less great because the ceramics are not so similar to the soil matrix; whereas
for the even more easily visible Roman Republican and early Imperial material, occurring at much higher
densities, we may assume all sites present in the survey fields have been discovered. Unfortunately the
much higher density of the later material also masks the presence of small and dispersed scatters of early
ceramics; in such a case the latter are much less likely to receive the attention and follow-up that the same
sherds would get when found in an otherwise sterile field.
Modern land management and agricultural practices have been another major factor influencing the
recovery of artefacts from the Fogliano study area, and research attempting to map and quantify this
factor is currently underway (Feiken & Van Leusen, forthcoming).
DEFINITION OF SITE AND OFF-SITE
The density at which we put the distinction between site and off-site material varies according to the type
of material being analysed; in general, ‘off-site density’can be taken to mean a density of less than 10
percent of the typical site density. For some materials that only occur in very low quantities and for which
there are severe visibility problems, this can mean that a single ceramic find is interpreted as a probable
site; at the other extreme, it can mean that high densities of Republican sherds occurring within 150
meters of a Republican site are classified as off-site material.
The main characteristic of the protohistoric off-site landscape as it appears from our survey is the
extremely low finds density of 1-5 finds per hectare. A comparison with equivalent figures for the
contemporary Alban Hills which are 30 kms away (ca. 50 finds per hectare), or even the Lepine margins
at less than 20 kms distance (ca. 15 finds per hectare), reveals that human use of the coastal landscape
must have been relatively marginal. In the late to post-Archaic period finds densities increase somewhat,
and four distinct areas of use become apparent on the map. The non-use (or at least non-intensive use) of
large parts of the landscape is evident from the percentage of fields with no finds from this period at all
(80%).
10 -
7
Aeolian /
sandy
body
Borgo
Ermada
beach
ridge
Imperial
> AD 100
Late
Republican &
Early Imperial
200 BC
– AD 100
Early & Mid
Republican
350 – 200 BC
Post-Archaic
500 – 350 BC
Unit
Site ID
Minturno 225
202
beach
201, 231
ridge
101
Archaic
600 – 500 BC
BA – EIA
< 800 BC
MIA & LIA
800 – 600 BC
V A N L EUSEN : P A T T E R N T O P ROCESS
102
F227/3
219
227
226
005
211
214
210
205
217
215
216
103
105, 106, 107
104
208
218
204
212
213
003
002
206
207
004
228
230
222
110
001
220
109
Figure 3: Summary results of the 1998 and 1999 Fogliano survey campaigns indicate the main
trends in the development of settlement patterns from the LBA onwards in the three major landscape
units. The width of the columns is proportional to the duration of each period. Light grey – presence
of site uncertain.
This can be contrasted with the situation developing throughout the middle and late Republic, in which
more or less all the available land was taken up for settlement and agricultural or other use. This is
reflected in the occurrence of small amounts of Roman Republican or early Imperial ceramics in nearly all
of the fields surveyed by us.
3.2
SUMMARY RESULTS
Figure 2 shows the location of the fields that were surveyed by the RPC project and the earlier Agro
Pontino project in the Fogliano area. For the purposes of our discussion of the results, the landscape is
divided into units on the basis of its physical geographical characteristics as described in section 2, the aim
being to show more clearly the continuity in occupation of the basic terrain units and the changes in
occupation density over time. The Minturno level beach ridge is ca. 400 m wide with an average elevation
of 13 m asl. The aeolian and sandy body connects with and overlies part of this ridge; it is several
10 -
8
F I E L D W O R K O N T H E P O N T I N E C OAST
kms deep and provides easily the most space for settlement and agriculture. The younger beach ridges,
just inland from the lagoons and separated from the two other units by bands of clayey alluvial and/or
lagoonal sediments, are less wide (about 150 m) and less high (maximum 9 m asl) than the older Minturno
level beach ridge. Figure 3 sets out the main trends and chronology of settlement in these three units.
Traces of human occupation of the Fogliano area are present in an unbroken sequence from the Middle
Palaeolithic onwards. The lithic material collected during the survey, which dates from the middle
Palaeolithic to the Neolithic, will be the subject of a separate publication; discussion here will be restricted
to the ceramic finds dating from the Bronze Age to the Roman Imperial period. The earliest ceramics,
which are of a friable reddish brown fabric with sand temper, probably date to the Bronze Age/Early
Iron Age (circa 1000 BC) and occur in at least two of the three landscape units; certainly all of the area
was regularly used from the advanced Iron Age (ca. 800 BC) onwards, although precise dating of the
material only becomes possible with the appearance of red slipped wares in the 7th century BC. A detailed
fabric analysis of the protohistoric material will be carried out in the near future. Figure 3 shows that, by
the Iron Age/Archaic transition (around 600 BC), essentially all of the coastal beach ridge and lagoonal
landscape
Figure 4: Density of pre-Roman (late Bronze Age to Archaic) ceramics in the Fogliano survey area.
was in use, with site numbers doubling and low levels of ceramics (1 - 5 sherds per hectare) occurring in
almost all fields. Small sites are dotted across the elevated parts of the landscape every few hundred
meters, avoiding only the clayey hinterland (see figure 4).
10 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
On the beach ridges, no major changes in this picture seem to have occurred, although a gradual
incorporation of the area into the expanding Roman Republic did lead to changes in building style (tiled
roofs) and in pottery consumption (wheel-thrown industrial pottery and fine wares) during the 5th and 4th
centuries BC. The number of sites gradually rises until the early Imperial period. The picture is
dramatically different for the aeolian and sandy body. Here, the relative lack of multiperiod settlements
throughout protohistory seems to indicate the unconstrained nature of the terrain, and the number of
sites doubles in the post-Archaic period. The early and middle Republican periods see a slight fall in the
number of sites, but some are re-occupied when, in the late Republican period (from ca. 200 BC), a
comparatively sudden five-fold increase in the number of sites reflects demographic growth perhaps
brought about by an influx of labour from other areas. Although difficult to quantify, this period sees the
growth of a rural village at Borgo Grappa, with isolated rural villas dotting the rest of the beach ridge
landscape. This development may perhaps be understood in the context of the establishment of a
production economy centring around the exploitation of the lagoons for fish farming, the importance of
which is indicated by infrastructural works to regulate water levels and salinity (see section 4 for a more
detailed discussion).
Figure 5: Density of Roman (post-Archaic to early Imperial) ceramics in the Fogliano survey area.
The late Republican flowering of the area lasted perhaps until the early 2nd century AD, after which
period our surveys found very little evidence for continued occupation into the middle Empire. The
virtual absence of finds from the late Empire and the post-Roman period might indicate that the area
reverted to occasional and marginal uses, but an alternative hypothesis suggests that the population was
concentrated into a small number of larger centres associated with Imperial villas, such as the one
belonging to Alfeius Ceionius Julianus Camenius, mentioned in an inscription from Fogliano (Cecere
1989:22).
10 -
10
F I E L D W O R K O N T H E P O N T I N E C OAST
3.3
SITE INTERPRETAT ION
Whereas the finds assemblage at sites dating to the Roman period, with its building stone, roof tiles, and
storage vessels, indicates fairly clearly that we are dealing with various types of (farm) building, such
clarity cannot often be obtained at the pre-Roman sites. For one thing, no traces of roof tiles were found
before the post-Archaic period; but even if the custom of tiling roofs had not yet reached the coastal
landscape, one would expect settlement sites to possess at least some thick-walled vessels for storage
purposes. One of the relevant results of the Fogliano survey was therefore the discovery of late Archaic
ceramic scatters consisting of small to medium pottery shapes mostly intended for food preparation. In
addition, nearly all of these sites occur as isolated scatters, that is, at locations that were not re-used for
later Roman settlements. It is possible that these are the remains of ploughed-out graves, an idea
supported by the fact that some of them occur in fields from which a considerable amount of topsoil is
said to have been removed (F. Gardosi, pers. comm.). Such graves would be likely to date to the late and
post-Archaic period (500 – 350 BC), when such burials are known to occur elsewhere. However, an
alternative explanation seems more likely at this stage – namely, that the early inhabitants of the coastal
areas lived in simple huts until well into the 4th century BC. This is in striking contrast to the proto-urban
tendencies present in the core areas of south Lazio.
4
DISCUSSION
The following paragraphs offer a preliminary interpretation of the results described above, in the context
of regional and supra-regional processes.
4.1
THE PROTOHISTORI C LANDSCAPE
For all of the protohistoric period from the Bronze age to the end of the Iron age, the archaeology of the
Lago di Fogliano survey area suggests a non-hierarchical settlement organisation consisting of small
settlements dotted over the landscape. Bronze age sites (supposing that our sandy fabrics are diagnostic
for this period) are very small and few in number. In this respect the Lago di Fogliano survey confirms
the results of the Agro Pontino surveys which recorded only six small sites with material dating to the
second millennium BC for the whole of the Pontine plain. Of these only one was said to contain more
than the usual few sherds (Voorrips et al. 1991:125). In his analysis of Bronze age settlement in South
Lazio, Guidi (1986) notes that the Bronze age period in general is characterised by a preference for
lakeside locations with, in its later phase, a more specific preference for the volcanic lakes situated inland.
The substantial and rich Bronze age site known as the Villaggio delle Macine (village of the grinding
stones), found submerged in the Alban crater lake, suggests that here indeed a process of centralisation
had begun by the Middle Bronze age (17th/16th centuries BC; Chiarucci 1985). The site of Casale Nuovo
along the Astura river with traces of metallurgy and the find of a Mycenaean IIIB shard indicates the
development of trade routes connecting the coast to the mountainous hinterland (Angle and Gianni
1985). The beach ridge complex of which our study area forms a part appears as yet to have been
excluded from these developments.
This must have remained so well into the Archaic period. Although human presence seems to have
increased judging from the increase in the number of sites and the quantity of shards found per site, the
finds assemblages do not contain roofing tiles or industrial-size storage vessels. In combination with the
small size of the scatters this indicates that in the 6th century BC the beach ridge complex was still very
much peripheral to the development of large proto-urban settlements and substantial rural infilling in the
Alban hills (cf. Attema 1993:219 – 224). It therefore seems likely that during all of the protohistoric
period the economic basis of life was intimately tied up with the exploitation of the lagoonal milieu, with
subsistence farming taking place in the immediate vicinity of simple hut dwellings. This is confirmed by
the agricultural evaluation, which indicates that, before technological advances made possible exploitation
10 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
of the heavier soils, only the aeolian deposits were suitable as arable; most of the remaining dry land
would only be suitable for extensive use, e.g. as grazing lands.
It is interesting to note the parallel between this protohistoric settlement pattern in the beach ridge area,
and that which can be deduced from maps that show a system of lestre (simple hut settlements and their
yards) used by fishermen and transhumant pastoralists, dotting the beach ridge area as late as 1851
(Attema 1993:51 and fig. 13).
4.2
THE ROMAN LANDSC APE
As is shown by figure 3, the number of sites rises in all three topographical units from the Iron Age /
Archaic transition onwards. Although the data are not very explicit yet on the post-Archaic period (5th /
4th centuries BC) pending closer ceramic dating of the survey material, there is no reason to believe that,
except for the introduction of farm houses with tiled roofs and maybe a corresponding shift in the
economy towards farming, substantial changes occurred in settlement intensity in the Lago di Fogliano
area. The largest increase in site density – from 3 to 16 occupied sites – occurs in the main aeolian / other
sandy body by the late Republican period, but it was argued in section 3 that this may be partially the
result of visibility biases. The agricultural land evaluation does indicate that technological advances
introduced by the Romans would have made both the beach ridge unit and the lagoonal unit between
ridges suitable for arable use – multiplying the area available for such use – but at the same time more
efficient farming practices would obviate the need for additional workers.
A substantial economic interest from outside the core area of Latium Vetus only took shape in the late
Republican period and is probably to be connected with the establishment of large and luxurious villas
along the coast, some of which exploited large artificial fishponds, and with industrial pottery production
also along the coast near the mouth of the Astura river (Picarreta 1977). This interest was backed up by
means of infrastructural works such as the Via Severiana, a Roman road that ran along the seaward or
landward side of the lagoons (Egidi 1980), and by interventions in the hydrography of the area, such as
the digging of the Martino canal (the Rio Martino). In his study of artificial fishponds in Roman Italy,
James Higginbotham notes that in Roman Republican times (3rd and 2nd centuries) inland bodies of water
were favoured for the farming of fish, but that at the beginning of the 1st century BC a change in taste
occurred towards sea fish. It is from this period onwards that elaborate seaside fish ponds were
constructed, a number of which are found along the coast between Nettuno and Circeo (Higginbotham
1997:4-5). It is against this background that we should evaluate the Roman site and off-site distribution
found in the Fogliano survey.
The original hydrography of the Fogliano area would have been much different from its current state.
This is illustrated by historical maps, and it is probable that the Romans were the first to make substantial
changes in the landscape. The Martino canal, draining the graben, was supposedly already dug through
the horst system in Roman times, thereby flowing into the coastal watershed area. It appears already on
the earliest maps dating around 1600 AD. In between the Lago di Fogliano and the Lago dei Monaci the
landscape changed gradually into marshland and rushes. Here the Rio Martino - coming from the Via
Appia - dispersed, partly discharging into the Lago dei Monaci and unable to reach the sea. Cecere
(1989:22) suggests that the Rio Martino may not primarily have been aimed at draining the Pontine
marshes lying north of the Via Appia, but rather was meant to provide the lagoons with additional fresh
water to improve conditions for fish farming in the Republican period. In any case, interventions in the
hydrography of the coastal landscape go back to Roman times as was proven by the German
archaeologist Elter (1884), who reports on an inscription found on a 1st century AD Roman villa terrain at
San Donato immediately to the north of the Lago dei Monaci. This inscription attributes to one
Phaenippus the carrying out of construction or maintenance at water management works built earlier in
the area. Cecere suggests that these works are related to the Rio Martino canal itself and were intended to
regulate the supply of fresh water to the lagoons.
Elter relates the archaeological remains found on both sides of the Rio Martino in the vicinity of presentday Borgo Grappa to a vast villa complex at San Donato that specialised in fish farming as well.
10 -
12
F I E L D W O R K O N T H E P O N T I N E C OAST
According to him this villa was continually enlarged between the 1st and 4th centuries AD. Although the
latter identification is not certain (Egidi 1980 provides an alternative identification as the way station Ad
Turres Albas on the Via Severiana), there seems sufficient reason to believe that the increase in settlement
intensity during the late Republican and early Imperial period recorded in the Fogliano survey must be
connected to the developments in the fish farming industry. Although the latter may have eventually been
monopolised by the large maritime villas, the increased level of economic activity would have attracted
others to set up as independent farmers in the area, or in a ‘service industry’to the Roman villa owners.
4.3
CORRELATING THE PHYSICAL AND HUMAN L ANDSCAPES
If we look at the most stable sites listed in figure 3, we see that the Minturno beach ridge and the aeolian
landscape were probably occupied in all periods from the late Bronze Age to the high Empire, while on
the younger beach ridges evidence for the very earliest and latest occupation phases is probably lacking
only because of the stochastic nature of the survey.
In order to illustrate the interplay between the human and natural landscape of the Fogliano region, figure
6 shows the combined results of the agricultural land use potential mapping and the archaeological
surveys for part of the study area. We will first discuss the diachronic changes in land use in relation to
the changing agricultural potential of the units in this area, then follow with a discussion of the correlation
between settlement location choice and the topography of the terrain.
As was stated above, the aeolian soil unit remains the best suited for agricultural use throughout the
whole period, whereas the suitability of clayey landscape units such as the lagoonal unit between the
beach ridges increases in the Roman period through soil improvement (adding sand) and technological
change (using heavier ploughs). The suitability of the beach ridge unit itself also increases with the
introduction of irrigation and manuring practices. It is only the level lagoonal unit which remains too
clayey for intensive agricultural use in any period. We would therefore expect a much wider agricultural
use of the landscape in the Roman period as compared to the protohistoric period, and the wide
distribution of Roman sites and offsite ceramics confirms this. On the level lagoonal unit we expect to
find no settlements of any period, and little if any original off-site material. Again this is largely confirmed
by the results of the Agro Pontino Project surveys, which report the majority of fields in this unit to be
aceramic, with indeterminate Roman finds occurring in one field where the unit borders on a beach ridge.
In addition to soils, the topography of the area also patterns the human landscape in a non-random way.
Factors such as exposure to the weather, viewshed, accessibility, and access to natural resources (water,
fish?) determine the precise location for settlement and other activities. We cannot hope to be able to link
the location of individual settlements to environmental factors operating at such small scales, but we are
nonetheless fortunate to possess detailed maps of the relief, hydrography, and general vegetation type of
the Fogliano area during the late 1920’s. These allow us to reconstruct, to some extent, the potential of
the landscape for travel, settlement, and various types of exploitation. Mesolithic to Archaic sites, for
example, nearly all occur at dry elevated parts of the landscape that strategically overlook lagoonal or
fluvial valleys, whereas the many late Republican settlements are preferentially located some distance away
from such locations and rather seem to be centered on prime agricultural land. The effects of exposure
and viewshed should be most clearly visible on the Minturno beach ridge, the south side of which
receives much more sunlight and sea wind than the north side (the other ridges are probably too low for
these factors to be significant). Although our own work indicates that sites dating from the Iron Age to
the Roman period were indeed situated on the south-facing slope of this ridge, the Agro Pontino Project
surveys found at least one late Iron Age site on its north slope.
The scope of field walking surveys does not allow us to extend our models of the correlation between the
physical and human landscapes by including site function as a variable. Neither is the area large enough to
allow quantitative analysis, but our analysis does illustrate in a qualitative sense how the history of
settlement and land use in the Fogliano region may be understood by reference to the physical landscape.
10 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 6: A detail of the Fogliano Survey area illustrates the value of the combined approach to the
physical and human landscapes. Both off-site finds densities and sites of all ceramic periods are
depicted against the background of the potential arable land use. For the legend to the finds
densities, see Figure 5. Surveyed fields with continuous outline: RPC project; with dotted outline:
Agro Pontino project.
5
CONCL U S I O N S
The surveys of the RPC project in the Fogliano area have been very successful in establishing a basic
settlement history that can probably be extrapolated to the whole Pontine coastal landscape. Continuity
of occupation has now been proven from the 8th century BC down to the 3rd century AD, and among the
protohistoric ceramic finds there may still be hidden a significant amount of Bronze Age and early Iron
Age material. Further study of these fabrics will be undertaken in order to reliably recognise this material.
The RPC approach of combining geographical work with archaeological survey has proved important not
only for a more detailed reconstruction of past land use than would otherwise have been possible, but
10 -
14
F I E L D W O R K O N T H E P O N T I N E C OAST
also because it has helped identify areas of anthropogenic disturbance. A more detailed formal land
evaluation using the FAO system is in progress (Van Joolen in prep.).
Similarly, our emphasis on understanding the biases that occur during an intensive systematic field
walking survey has prevented us from jumping to some unwarranted conclusions and has generally
helped us in our interpretation of the results. We feel this aspect of our research is so important as to
deserve further study - the Pontine landscape has been subject to so many recent changes, especially since
the 1930s, that a more detailed reconstruction of its original geomorphology and hydrology is needed
before the results of the archaeological surveys can be fully understood. We obtained elevation contour
maps of the area as it was during the Bonificá of the 1920’s, and by studying the differences between these
and the modern relief we have mapped the main landscape changes and modeled the effects on the
accessibility of the landscape of fluctuations in the ground and surface water levels (Feiken & Van
Leusen, forthcoming).
The results of the Fogliano survey are best interpreted in the context of the developments in the wider
region by relating them to processes of centralisation and early urbanisation in the core areas around the
Alban massif and Rome. The development there during the Bronze Age, Iron Age and Archaic period
(i.e. to the 6th century BC) of, first, centralised settlements and, later, peer polity city states is reflected by a
similar, but late and stunted, development of marginal polities such as Caracupa / Valvisciolo on the
Lepine footslopes and Cisterna di Latina on the south-eastern margin of the Alban massif. During the
post-Archaic and Republican period the growing political, military and economic influence of Rome
expressed itself archaeologically first in the establishment of colonies on the Lepine margin and, later, in
the exploitation of the coastal landscape for fish farming, pottery production and leisure industry, and
mixed farming on the colluvial slopes and (though much less so) along the Via Appia. The apparent
dismantling of the Lepine olive culture and the near abandonment of settlement there and in the coastal
area following the early Empire indicates that the Pontine region generally became economically
marginalised as the Roman Empire moved its large-scale agriculture and industry elsewhere.
ACKNOWLEDGEMENTS
We would like to thank the students from the universities of Gent, Amsterdam, and Groningen who
joined the RPC-team in carrying out the field work. Special thanks are due to Bas Bijl for his help in
conducting the physical geographical survey, to Massimo Alvisi for his logistic support, and to
Michelangelo La Rosa for his preliminary analysis of the lithic finds. We are grateful to the inhabitants of
Borgo Grappa who allowed us to set up camp in their school buildings, and to the farmers around Borgo
Grappa who gave us access to their fields. Finally we would like to thank the superintendent of the
Archaeological Service for Lazio, D.ssa Annalisa Zarattini, who made the survey possible.
REFERENCES
Angle, M. and A. Zarattini 1987
L’insediamento protostorico di Casale Nuovo, in Archeologia Laziale VIII, QuadAEI 14, 250-252.
Attema, P.A.J. 1993
An Archaeological survey in the Pontine Region, A contribution to the early settlement history of South
Lazio (900-100 BC), Groningen (Ph.D. thesis).
Attema P.A.J., G.J. Burgers, M. Kleibrink and D.G. Yntema 1998
Case Studies in Indigenous Developments in Early Italian Centralisation and Urbanisation: a Dutch
Perspective, in European Journal of Archaeology, Vol. I: 326-381.
Bouman, D. and G. Rot 1982
Bodemkaart omgeving Fogliano. Unpublished report, Fysisch- Geografisch & Bodemkundig Laboratorium,
Universiteit van Amsterdam.
Cecere, C. 1989
La Villa Caetani A Fogliano, Il Luogo, L’Architettura, La Storia, Arti Grafiche Fratelli Palombi, Roma.
Chiarucci, P. 1985
Materiali dell’età del Bronzo nelle acque del lago Albano, in Archeologia Laziale VII, QuadAEI 11, 18-25.
Egidi, R. 1980
Una statio romana sulla via Severiana: ad Turres Albas, in Archeologia Laziale III, QuadAEI 4, 123-125.
Elter, A. 1884
10 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
Antichità Pontine, in Bollettino dell’Istituto di Correspondenza Archeologica, ??. Roma.
FAO 1977
Guidelines for soil profile description (2nd edition). Rome.
Feiken, H. A. and P. M. van Leusen, forthcoming
Interpreting Field Survey Results In the Light of Historic Relief Change: the Fogliano beach ridges (south
Lazio, Italy), in CAA 2000. Proceedings of the 2000 Meeting of Computer Applications and Quantitative
Methods in Archaeology, Ljubljana (BAR Int. Series).
Guidi, A., 1986
Gli Insediamenti perilacustri di riva d’età protostorica nel Lazio centro-meridionale, in Quaderni di
Protostoria, I, 239-247.
Higginbotham, J. 1997
Piscinae, Artificial fishponds in Roman Italy, The University of North Carolina Press, Chapel Hill and
London
Joolen, E. van, in prep.
The Changing Landscape: Land evaluation of three central and south Italian regions from the late Bronze
Age to the Roman period, 1400 BC – AD 400. PhD thesis, University of Groningen.
Kamermans, H. 1991
Faulted land: the geology of the Agro Pontino, in A. Voorrips, S.H. Loving, and H. Kamermans (eds), The
Agro Pontino Survey Project (Studies in Prae- en Protohistorie 6). Amsterdam: Instituut voor Pre- en
Protohistorie.
Kamermans, H., A. Stierman, P. Vos and W. Westerhoff 1979
Bodemgesteldheid in de Agro Pontino ten Z.W. van Latina, Midden Italië. Unpublished report, FysischGeografisch & Bodemkundig Laboratorium, Universiteit van Amsterdam, June 1978.
Picarreta, F. 1977
Astura, Forma Italiae I, vol. XIII, Firenze.
Sevink, J., P. Vos, W.E. Westerhoff, A. Stierman and H. Kamermans 1982
A sequence of marine terraces near Latina (Agro Pontino, Central Italy), Catena 9, 361-378.
Sevink, J., A. Remmelzwaal and O.C. Spaargaren 1984
The soils of southern Lazio and adjacent Campania. Publication no. 38 of the Fysisch Geografisch en
Bodemkundig Laboratorium, Universiteit van Amsterdam.
Spurr, M.S. 1986
Arable cultivation in Roman Italy, c. 200 BC - c. AD 100 (Journal of Roman Studies Monograph no. 3),
London.
Voorrips, A., S.H. Loving and H. Kamermans (eds) 1991
The Agro Pontino Survey Project (Studies in Prae- en Protohistorie 6). Amsterdam: Instituut voor Pre- en
Protohistorie.
10 -
16
C HAPTER 11
WALKING THE MURGE: INTERIM
REPORT OF THE OSTUNI FIELD
SURVEY (APULIA, SOUTHERN
ITALY)∗
Gert-Jan Burgers, Martijn van Leusen, and Peter Attema
I
1.1
INTRODUCTION
COMPARATIVE SETT LEMENT ANALYSIS
Although studies of ancient landscapes have a long established tradition in mediterranean archaeology,
one may observe in recent decades a definite break-through, encouraged in particular by the development
of intensive field survey techniques. One of the merits of this trend is that it has allowed us to question
traditional generalisations emphasizing uniformity in Graeco-Roman culture throughout the
mediterranean basin. By stressing internal, and regionally specific, factors of change, the projects involved
create the possibility of identifying variability in the dynamics of regional cultures and landscapes.
However, the disadvantage of this approach is perhaps an overemphasis on regionally specific
explanations. Whereas many regional data sets have become available in the last decennia, few attempts
have yet been made to formulate new questions and syntheses on a supra-regional level (see notably
Alcock 1994, Bintliff 1997). This is especially the case in Italy, notwithstanding the fact that the number of
surveys here is comparatively large. To confront this issue, in 1997 three longstanding Dutch regional
fieldwork projects joined forces to establish a new project aiming at a comparative analysis of centralization and early urbanization processes in three regional landscapes in Central and Southern Italy. This
umbrella project was named Regional Pathways to Complexity, Landscape and Settlement Dynamics in Early Italy RPC project for short. It is carried out by the Archaeological Institutes of the Vrije Universiteit at
Amsterdam (AIVU) and Groningen University (GIA), and is subsidized by the Netherlands Organization
for Scientific Research (NWO). The Italian regions investigated are, from south to north, the Sibaritide in
Calabria on the Ionian sea, the Salento Isthmus in Apulia on the Adriatic, and the Pontine Region in Southern Lazio on the Tyrrhenian sea, south of Rome (see figure 1). The data proffered by the fieldwork
projects carried out in these regions are analysed both within their specific internal social and
environmental contexts, and in the context of the supraregional networks these regions were embedded in
(see notably Attema et al. 1998a/b).
In order to fully understand variability within and between the regional contexts, we hold that it is
imperative to investigate the total landscape of the regions concerned, i.e. all landscape units represented
within them. Thus, in the context of the RPC project additional field surveys are carried out in those
landscape units that have so far received little attention from archaeologists (see also Van Leusen 1998;
Attema, van Joolen and van Leusen 2001). This report presents the results of one of these surveys,
focused on the uplands of the Murge, the northern part of the Salento isthmus.
∗
This chapter will be published in Studi di Antichità, 2002. Introduction and Discussions by Gert-Jan Burgers; Methodology and
Data Processing by Martijn van Leusen; Impasto quality study by Peter Attema.
11 -
1
V AN L EUSEN - P ATTERN TO P ROCESS
Figure 1: The Salento peninsula, with locations of major sites mentioned in the text. The shaded areas
are those covered by the Ostuni’99 survey and by the previous surveys of the AIVU. Inset: Italy,
with three RPC regions indicated.
1.2
AIMS OF DUTCH RE SEARCH IN THE SALENT O ISTHMUS
The ‘Salento Isthmus’is the common denomination of the stretch of land between Taranto and Brindisi,
connecting the Salento peninsula to the rest of Italy (figure 1). Apart from the limestone plateau of the
Murge in the north, the major landscape units in the region are the Taranto plain in the southwest and the
larger Brindisi plain, or piana messapica, to the southeast. Since 1981 the AIVU has carried out a series of
excavations and surveys in various rural and urban units throughout this region, but focusing in particular
on the Brindisi district (figure 1)1. This fieldwork aimed to define the 1st millennium BC occupational
history of the region and to analyse it in the light of three major supra-regional processes:
1. Processes of centralisation of settlement during the late Bronze Age and early Iron Age (ca.
1400-600 BC)
2. Early urbanisation, which in Salento is attested notably during the late Archaic and early
Hellenistic periods (550-250 BC)
3. Romanisation of the early urban landscape, which proceeded after the Roman conquest of
Salento in the first half of the third century BC.
See notably van Alberda et al. 1999; Boersma 1990; 1995; Boersma et al. 1991; Boersma/Yntema 1987; Burgers 1998; Yntema
1993-1/2. Preliminary reports have been published annually in Bulletin Antieke Beschaving and Taras, Notiziario delle attività di tutela.
1
11 -
2
F I E L D W O R K I N T H E S A L E N T O M URGE
Accordingly, the general aim of the Ostuni surveys was to evaluate the settlement dynamics of the Murge
landscape unit in the context of these three processes.
1.3
THE OSTUNI SURVE Y
The Murge plateau is part of the Apulian karst. The landscape is marked by rolling hills and ridges,
alternating with dolines and other forms of enclosed depressions characteristic of karsts. The altitude of
the Murge averages about 420 meters above sea level. The plateau extends from the Bari district in the
northwest to the Salento peninsula in the southeast. At its eastern edge, it drops abruptly towards the
coastal plain bordering the Adriatic, whereas in the south it slopes more gradually towards the Gulf of
Taranto. In the west it gradually merges into the Appennine mountain chain.
Intensive surveys in the Murge area have been limited to its western edge, at the Bradano trench between
Apulia and Basilicata (Vinson 1972; Small 199l, 1998). Little detailed knowledge is available on the
archaeology of the Salento part of the plateau, some 100 km to the east. This situation contrasts with that
in the adjacent Brindisi plain, which mainly consists of fertile light arable and clayey soils. Here, intensive
surveys by the AIVU have resulted in the identification of a densely and hierarchically settled landscape at
least in Hellenistic and Roman times (see note 2). A major question regarding the Murge was therefore, to
what degree the lack of data in this area reflected a lack of intensive, problem oriented research, or a
Figure 2: Geomorphological map of the larger part of the Salento Murge, with locations of major sites
mentioned in the text and of the Ostuni’99 survey areas.
11 -
3
V AN L EUSEN - P ATTERN TO P ROCESS
marginal human exploitation.
To confront this question, a survey was carried out in 1999 by staff and students from the two universities
participating in the RPC project in two landscape units typical of this part of the Murge, located in the
territory of the modern town of Ostuni (province of Brindisi; figure 1). To deal with the possibility,
discussed above, that these areas were only marginally exploited, the survey method was specifically
designed to document the low density and low visibility artefact scatters characteristic of low intensity
settlement and land use. However, as we shall describe below, although we did in fact detect very diffuse
and low density ceramic scatters within a general ‘blanket’of off-site material, it has also become clear that
such scatters do not constitute the norm in either of the two investigated landscape units. Both contain
large and dense sites dating to the middle Bronze Age, and settlements from the Hellenistic and Roman
periods were also discovered, suggesting an intensive human exploitation in these phases.
1.4
METHODOLOGY
To investigate the possibility of differentiation in the human exploitation of the local landscape, 1.5 km²
sample areas were located in two of the landscape units defined by the land evaluation (figure 2; see
section 2). The first sample area (‘Area 100’) is situated near the cliff-like interface between the Murge and
the coastal plain, just below the large Bronze Age site and cult cave at Masseria Risieddi; the second
sample area (‘Area 200’) lies some six km inland, on the Murge plateau around the post-Medieval Masseria
Cervarolo. Both areas were systematically surveyed in gridded units of 0.25 ha (50 by 50 m) at 10 m
intervals (~ 20% coverage) by teams of student surveyors, and all non-recent objects were collected per
unit. Further passes were employed in order to define, map, and sample any finds concentrations (‘sites’)
encountered during the first pass. Samples from these further passes were generally taken in order to be
able to quantify the density of material at various locations within a concentration, and in order to collect
additional diagnostic material for dating. Both survey unit boundaries and finds density contours were
mapped on 1:10,000 scale cadastral maps of the Comune of Ostuni for later digitisation. The finds collected
during the survey were sorted, weighed and counted after washing and drying, then passed to lithic,
protohistoric, and classical material specialists for more detailed classification; these data were then stored
in a database system (MS-Access) attached to a GIS (GRASS) containing the digitised field maps.
POST-PROCESSING
The raw counts and weights resulting from the survey in themselves cannot easily be interpreted because
the collection units themselves may vary in surface area, finds visibility, and surface percentage covered.
The results are therefore presented here in the form of density distribution maps corrected for unit area
and aggregate finds visibility (a correction for percent coverage was not needed in this case because this
factor was kept invariant during the survey). Raw counts per unit were transformed into normalised
densities (counts per hectare) by calculating the true surveyed area of each unit (excluding landscape
features such as paths and verges, which can occupy a large part of the area of a unit) in the GIS, and
dividing the raw counts by the recorded percent coverage and the true unit area in hectares. For example,
if 6 impasto sherds were found in a unit of 0.25 hectares and 20% coverage, the normalised impasto
density will become 6 / 0.25 / 0.2 = 120 per ha. Finally, the normalised finds densities were corrected for
factors causing differences in finds visibility (and therefore biasing recovery rates). Although it is generally
agreed that such a correction is necessary before the results of a survey can be properly interpreted, there
is no such agreement on objective methods for recording visibility factors, nor for using these to correct
uncorrected finds densities. A full discussion of these issues will be the subject of a separate publication
(Van Leusen, in prep.); the method followed by us will be described here.
An aggregate visibility estimate on a 6-point ordinal scale, based on tillage, vegetation, weathering, and
stoniness, was recorded by us in the field for each collection unit (figure 3). In order to quantify the effect
of this visibility estimate, each of the points on the scale was assigned a percentage value based on our
estimate of its relative significance (for example, high = 100%, normal = 90%, low = 50%, and very low =
25%). Following this, the algebraic correction for this recovery bias was applied in MS-Access by dividing
the normalised finds densities by the aggregate surface visibility percentage. In the example used earlier, if
11 -
4
F I E L D W O R K I N T H E S A L E N T O M URGE
visibility had been ‘normal’, the corrected impasto density would be 120 / 0.9 = 133 per hectare. An
unsolved problem with this approach, particularly when working with low raw counts and visibility
percentages, is that the significance of random variations in finds density is exaggerated, while the diversity
(in the statistical sense) of the finds assemblage is systematically underestimated. See, for example, the
discussion in Orton (2000:172-6), and the discussion following Bintliff’s recent re-evaluation of Boeotia
survey data (Bintliff et al. 1999, 2000).
Figure 3: Aggregate surface visibility in the Ostuni survey areas (aggregate of ploughing conditions,
vegetation cover, dust/rain, stoniness).
LAND EVALUATION AND LAND USE HISTORY
In order to relate the findings of the archaeological surveys of the RPC project to the characteristics and
dynamics of the contemporary landscape, physical geographical mappings are being carried out in
conjunction with the surveys. Such fieldwork also took place in conjunction with the Ostuni ’99
archaeological survey (figure 2), with the aim of creating a soil unit description according to the guidelines
of the Food and Agriculture Organisation (1977). Such a description was needed in order to do a land
evaluation for agriculture from the late Bronze Age until the Roman period. The method of land
evaluation compares the soil requirements of land utilisation types (e.g., of Roman arable farming) with
the land qualities and land characteristics of the physiographic units defined during the physical
geographical fieldwork, in order to determine their agricultural suitability. The results, to be published in
full elsewhere, are here summarized in section 2.
While the land evaluation is intended to reveal correlations between the archaeological record and the past
agricultural suitability of the land, the more recent history of land use within the survey areas has a directly
observable impact on that record. Surface artefact recovery rates are of course biased by a host of factors
besides those directly relevant to the density calculations discussed above. Those factors considered most
relevant in the study area were recorded during the survey - the presence or absence of features indicating
ongoing surface erosion (figure 4), the amount of recent or subrecent material on the surface (figure 5),
and the condition of protohistoric sherds (see below, figure 10). The significance of each for the
interpretation of the results of our survey will be briefly discussed below.
The presence of erosive features in the study area, including slope wash, small gullies, and patches of
bedrock, indicates the instability of soils under current tillage, and is therefore an indication that soils
including archaeological material may have been eroded at some areas and redeposited elsewhere.
Evidently geomorphic processes might be responsible for distorting the recovered pattern, especially of
protohistoric finds, but because erosive features were not consistently recorded for all collection units the
resulting map (figure 4) must be seen as indicative and no formal analysis along lines similar to the Cecina
survey (Terrenato 1996, Terrenato and Ammerman 1995) could take place. However, this factor does
enter into our discussion of the interpretation of the protohistoric ‘off-site’material in section 3.
11 -
5
V AN L EUSEN - P ATTERN TO P ROCESS
Recent or subrecent material (a category which is composed mainly of glazed or very hard baked white
ceramics, but also contains glass, rubber, metal and other materials) occurs throughout the study area and,
even in low densities, is likely to distracts walkers from noticing archaeologically more significant find
types (figure 5). Where it occurs at higher densities, this factor is likely to have a significantly dampening
influence on recorded finds densities. Figure 6 shows that the largest recorded density of (sub-)recent
material was present in the southeastern part of Area 100, nearest the modern town of Ostuni, and it
seems likely that refuse from the town was dumped in these nearby fields. The possibility that
archaeological material, especially from the classical periods, might have been overlooked by the walkers
in these fields led to a targeted resurvey, the results of which will be discussed in the appropriate section
below.
Figure 4: Presence of erosive features in the Ostuni survey areas.
Figure 5: Corrected density distribution maps of recent finds in areas 100 and 200.
Some months after the original field work at Ostuni, the protohistoric impasto ceramics forming the bulk
of the finds were studied in more detail by one of us. The aim of this study was to identify whether postdepositional disturbance had occurred among finds groups recorded as ‘sites’in the field or, in other
words, whether such find complexes were likely to have been found in situ or not. Sherd fragmentation
and wear were taken as an indicator of land use intensity and sherd displacement, reflecting the degree to
which the impasto has been affected by fluviatile movement, frequent tilling and weathering. The main
outcome of this study, discussed in more detail in section 3.2, is that there is no simple correlation
between the density and the quality of the protohistoric material, casting doubt on the validity of site/offsite distinctions made in the field.
11 -
6
F I E L D W O R K I N T H E S A L E N T O M URGE
2
L A N D S C A P E , S E T T L E M ENT AND AGRICULTURE
2.1
EVALUATION OF THE PHYSIC AL LANDSCAPE 2
From the Adriatic coast to the interior, the three major physiographic units identified during the land
evaluation are the coastal plain, concavely sloping land, and rolling land on the Murge plateau (figure 2).
The coastal plain is separated from the sea by a narrow line of dunes. The plain rises very gradually from 2
m asl behind the dunes to 70 m asl some 2-3 km inland, at which point it gives way to the concavely
sloping land unit. At intervals of 1 to 2 km the coastal plain is incised by deep, canyon-like river valleys
(lame) which originate in the Murge upland. The soils of the plain reach a thickness of 50 cm on average.
Along the coast, the upper calcareous horizons are more coarsely textured than in the remainder of the
unit, where they consist of sandy loam or sand. Most other soils consist of an upper horizon of brownishred loam, with a second more fine-grained horizon underneath. The main crop cultivated in the plain is
wheat, interspersed with olives on the higher parts.
The second landscape unit consists of a very wide concave slope at the base of the cliff-like edges that
separate the lowland from the Murge plateau. The soils of this unit vary in thickness between 10 and 160
cm, consisting mostly of sandy to silty loams. Most commonly, an upper calcareous red A-horizon is
followed by an even brighter red non-calcareous B-horizon. In thinner soils the B-horizon is absent and
the A-horizon lies directly on top of the bedrock. Stoniness is rather high in this unit, and occasionally the
bedrock even comes to the surface. Although wheat growing does occur in this unit in conjunction with
olive culture, the latter dominates to such an extent that natural vegetation has almost completely
disappeared.
Our third unit, the rolling land of the Murge plateau, largely consists of hills alternating with depressions,
dry valleys and valley floors. The upper parts of the hills generally have thin soils, consisting of loamy clay,
silty loam or loam. These soils all lie directly on the limestone bedrock, which here, too, frequently comes
to the surface. The degree of erosion is high on these hills, which are planted mostly with almonds and
occasionally with olive trees, plants that need only thin soils to survive. Accumulation takes place on the
lower parts of the slopes and in the depressions and valleys. In the latter, soils are at least 150 cm thick
and consist of clayey loam or loam. In contrast to the other units, drainage in these lower areas is relatively
poor and they are mainly used for viticulture and horticulture.
For each of these units, the suitability for specific (pre-) historic land use types was determined on the
basis of an evaluation of a range of land characteristics (Kamermans 1993; Foeken & Gietema 2000).
According to this classification, the lowland units can both be defined as suitable for wheat growing, even
without the use of ards. This also holds good for the lower parts of the slopes and the river valley floors
on the Murge plateau. The major factors of influence here are the relative flatness, ample nutrient
availability and workability. The agricultural potential of the river valley floors, with their relatively clayey
texture, is likely to have improved with the introduction of a drainage system and of ploughing. Of all
units in the study area, these valleys are least suited for olive growing and best suited for horticulture and
viticulture, for which ample foothold is a major requirement. Because of their steepness, stoniness, thin
soils and excessive drainage, the Murge hills were less attractive for wheat cultivation. However, they can
be defined as (marginally) suitable for vine and olive growing, although the climate factor should not be
underestimated (see above). Indeed, from the point of view of climate, the lowland units have much more
to offer olive culture, and olive trees are nowadays found even on the lower slopes of the otherwise bare
Murge cliff facing the coastal plain.
2 This summary is largely based on the preliminary physical geographical fieldwork conducted in parallel with our survey, and
reported in Foeken and Gietema 2000.
11 -
7
V AN L EUSEN - P ATTERN TO P ROCESS
2.2
SETTLEMENT AND L AND USE
A brief review of settlement hypotheses for the Salento part of the Murge must start with the work of
Prof. Coppola at Rome and Lecce University, whose interest in cave, lithic and early ceramic sites provides
us with a scatter of such sites (Coppola 1977; 1983; 1985). Coppola has dedicated most of his research to
the archaeology of the caves, which were formed by two different processes in the limestone geology of
the area. Percolation of rain water along angled bedding planes on the one hand has resulted in caves with
entrances where such bedding planes intersect with the land surface; other caves were opened up by wave
action along successive coastlines as the land was lifted up and the sea subsided in stages. Many of these
caves were inhabited from the early Palaeolithic onwards – perhaps even by hominids3. Lithic sites were
also found along lame (fossil river valleys) in the coastal plain.
In contrast to the relative abundance of known prehistoric sites in the Murge, information on occupation
in protohistoric and historic phases was virtually lacking before the start of our survey. A few larger
settlements dating to the Bronze and Iron Age were known to occur on the edges of the Murge, some 6
or 7 km from the present coastline, and another handful of Roman ‘villa’sites were known for the entire
area under consideration. This led to the hypothesis that the high Murge was, at most, marginally used
before the Middle Ages. In the Middle Ages themselves, and for most of its more recent documented past,
the Murge plateau was never intensively exploited or settled, being defined as selva (or forest - as opposed to
the coastal plain, known as marina). Peasants of nearby villages enjoyed rights to put animals to pasture in the
selva, to spend the night, cut wood, and to draw water from catchment basins (Galt 1991:69). From the 19th
century onwards, these forests have gradually been reduced by large scale deforestation projects, and today
only 5% of the region is still covered with woods or macchia, notably on the cliff-like edges mentioned earlier.
The 19th century saw a remarkable rural infill of the Salento Murge, with isolated farms (trulli) and hamlets
(jazzèlere) appearing all over the plateau. This reclamation of previously marginally exploited lands, with
roots in the later 18th century, can be intimately linked with the rise of viticulture at a time of increasing
international demand for wine4. Vine growing became a major occupation of the peasants colonising the
plateau. External markets also played an important role in the flourishing of olive culture in the coastal
lowland, which has long been dominated by olive growing. As early as the 16th century the area was
described as ‘a forest of olives’(Galt 1991:71-72). By contrast, olive culture is a marginal phenomenon in
the Murge because of its abundant winter precipitation and even the occasional snow storm. Frost damage
is therefore an ever present danger making olive culture a risky business.
3
RESULTS OF THE ARC HAEOLOGICAL SURVEY
3.1
GENERAL OBSERVAT IONS
Nearly 850 blocks (collection units), for a total area of 243 hectares, were surveyed during the Ostuni’99
campaign. Figure 6 shows the locations and approximate sizes of the 36 sites recorded in this area. Our
survey method aimed explicitly at recovering marginal occupation, and its success is demonstrated by the
fact that we did indeed identify even very small and diffuse scatters. After processing, 238 kg of ceramics
remained, about evenly divided among three categories – impastos, depurated wares, and other pottery.
Apart from a general thin lithic scatter, to be published in detail elsewhere, ceramic material datable to the
Neolithic was found in only one spot, while nearly all of the landscape was found to be blanketed in
Concurrently with the Ostuni survey an excavation was carried out at a newly discovered cave at Ceglie Messapico by a team led
by Prof. Coppola and Dr. Biagio Giaccio. If the geological ante quem date of 560,000 BP is correct, this cave was inhabited by
Homo Erectus.
3
The anthropologist Galt (1991) has dedicated a detailed ethno-historical account to this process, focusing on the interaction between
landlord and peasant strategies in the territory of Locorotondo, some 25 km northwest of Ostuni.
4
11 -
8
F I E L D W O R K I N T H E S A L E N T O M URGE
Bronze Age coarse impasto ware. Equally as remarkable as the amount of Bronze Age ceramics found, is
the almost absolute lack of Iron Age to Classical material: only a handful of (possibly) Iron Age sherds,
and no Archaic or Classical material at all, were found. While sites from this period might in theory go
unrecognized because of a lack of diagnostic fine wares and roof tiles, one would expect to find at least
thick walled (impasto) storage wares.
Some of the most evident scatters of artefacts found during the Ostuni survey can be dated to the
Hellenistic and Roman periods; their visibility is high because of the relatively large amounts of tiles and
brightly coloured wares (eg, orange coarse ware, black gloss and ARS). Interpretable as farmstead and
graveyard sites, in area 100 they occur on lightly sloping terrain not far from the edge of the Murge; while
in area 200 they are situated fairly high up the slopes overlooking the larger valleys.
Figure 6: The gridded survey transects of the Ostuni’99 survey, with site areas shaded.
3.2
THE PROTOHISTORI C PERIOD
Figures 7 and 8 show the raw and corrected densities of protohistoric materials (almost exclusively coarse
Bronze Age impasto ceramics) recorded during the first pass of the Ostuni survey. It may be observed
that this material occurs over most of both areas, in densities varying from 1 to 570 per hectare (median
density: 13). In Area 100 the relative absence of protohistoric material in the northeastern half may be
partly related to the local topography, with a reduced use of the less accessible terrain directly below the
Risieddi terrace, but it also seems likely that recovery rates were substantially lower here because of the
presence of much recent material (above, figure 5). In Area 200, zones of low finds density appear not to
be systematically correlated with land form, but we do suspect that high recent erosion and deposition
rates may be involved. The large number of very small (diam 20 m) to small (diam 50 m) impasto sites is
remarkable, and they seem to occur in all types of terrain including, on occasion, hilltops; however, the
larger impasto sites tend to be situated on lower slopes and valley bottoms, in area 200 even clustering
together in one valley5.
The macroscopic fabric of the impasto pottery group, very homogenous in area 200 but less so in area
100, is red firing and the paste contains quartz and/or flint and iron6; no shiny minerals were noted in the
clay, nor does it appear to be very sandy. The pottery has not been subjected to microscopic analysis yet.
A shiny layer caused by burnishing of the clay body was noted on a fairly large number of sherds from
area 200, and appears as a separate layer onto an often friable core. The sherd surface is unevenly burnt
and colours typically vary between red 2.5YR 5/8 and strong brown 7.5YR 5/6 to black. Sherd cores
5
Given the occurrence of iron ore and slag we may have to presume that such sites are exceptional.
6
Clay layers containing FeMn nodules are present in area 200.
11 -
9
V AN L EUSEN - P ATTERN TO P ROCESS
consist of red and black layers or are totally black, indicating firing at a low temperature in an open fire.
Many sherds do not have any surface treatment, though this may be partly due to post-depositional
processes. This ‘standard’impasto fabric is dated to the early part of the second millennium BC (protoApenninico B), implying that the settlement system producing it may have been in use for as brief a period
as 100 to 150 years only (personal communication dott.ssa A. Cinquepalmi). Although diagnostic forms
were barely present, and the general pattern of finds was more diffuse in area 100, reflecting its higher
intensity of agricultural land use, the larger part of the impasto fabrics resemble those of area 200, and all
may be assigned to the proto-Apennine B facies.
Figure 7: Raw density distribution maps of impasto finds in areas 100 and 200.
Figure 8: Corrected density distribution maps of impasto finds in areas 100 and 200.
11 -
10
F I E L D W O R K I N T H E S A L E N T O M URGE
Figure 9: Presence of possible Neolithic and Iron age material in area 100, with location of the
Zaccaria cave.
Area 100 seems not to have been quite abandoned in the preceding and following periods, however,
because two sherds of a possibly Neolithic light brown burnished ware were found just below the
Zaccaria cave site7, while sherds of a harder fired impasto with crushed limestone inclusions, indicative of
Iron Age fabrics, were found in two other locations (figure 9)8.
Encircled densities in figures 7 and 8 indicate the areas of higher relative finds density (‘sites’) mapped in
the field, details of which are listed in Table 1 (site reference numbers can be found in figure 6). While the
site/off-site distinction could be maintained fairly easily for the classical periods, and the material was
generally in good condition with no reason to suspect the presence of extensive manuring or plough
scatters (see below), the variation in density and quality of the protohistoric material made us question the
validity of our site/off-site distinction. An impasto quality study was therefore conducted, the results of
which confirmed our suspicions: on the one hand, some ‘protohistoric’sites (eg, site 4) turned out to
consist partly or wholly of a (probably post-antique) very hard sandy impasto-like fabric, or to be
contaminated by off-site plain coarse wares dating to the Hellenistic and/or Roman periods; on the other
hand, groups of high quality protohistoric impasto sherds had gone unrecognised in the field.
The impasto quality study was carried out in order to determine whether post-depositional disturbance
had occurred among the impasto finds groups. Focusing on variability in sherd fragmentation and wear,
four classes were defined: 1) completely rolled, 2) worn and with rounded edges, 3) worn, but still
recognizable as (body) sherds, and 4) sherds with surface treatment and form characteristics preserved.
The Zaccaria and S. Maria d’Agnano caves were both occupied from the paleolithic onwards; among the finds reported in
Coppola (1983:24, 251-2) are sherds of various types of Neolithic impasto.
7
Coppola (1983:252) suggests that a few Iron Age sherds found at the entrance of the S. Maria d’Agnano cave are derived from
the Masseria Risieddi site located at the top of the Murge cliff.
8
11 -
11
V AN L EUSEN - P ATTERN TO P ROCESS
Figure 10: Summary of the distribution of the four impasto quality classes in areas 100 and 200. The
numbers refer to the four classes mentioned in the text.
Table 1: Protohistoric sites (for locations, see figure 6)
Area 100
ID
2
3
8
9
10
11
12
16
11 -
Description
Small (diam 5 m) and dense scatter
This scatter consists of both impasto pottery and flint. However, the impasto is not of site quality.
Small (diam 5 m) and diffuse scatter; halo (radius 100 m)
This scatter contains 15 thin-walled impasto-like sherds of some 7mm thick, and 6 sherds of medium-walled
impasto, some rather hard fired. However it also contains some late coarse ware (date uncertain). The
scatter contains 10 certain pre/protohistoric impasto sherds (probably BA), which in itself does not warrant a
site, but in conjunction with other finds from the surrounding area (‘halo’; blocks 109-14 to 17) there is
enough standard impasto of site quality. Diagnostic forms are hardly present: Block 109-14 contains a.o. a
large knob? and rim of tazza; 109-15 contains several large fragments and also a thin-walled impasto
tazza.
Elongated scatter of impasto pottery; dense core 150 by 10 m, periphery 150 by 50 m
The impasto within this scatter, of standard type, is very worn and not of site quality; it contains only a few
clear body sherds. Some sherds appear to be fired harder than usual.
Dense scatter, at least 250 m wide but could be larger toward S and W; top 10 m on Agnano terrace
and continues another 50 m down terraced slope
This scatter contains much impasto material of standard type, but some 10% is in fact coarse ware or fabric
X (see following sections). The impasto is generally very worn, but a small percentage is not worn and will
have been ploughed up more or less in situ (e.g. in block 115-01). The scatter may have originally extended
further to the south and west, on and along the edge of the Agnano terrace, but dense vegetation precluded
survey. Site status is beyond doubt.
Diffuse scatter, core diam 80 m, halo over 100 m wide
Much of the impasto in this scatter is very worn, except for blocks 118-15 and 118-A-DS-1 which have
clearly site quality material.
Diffuse scatter, diam 25 m
This impasto scatter was characterized in the field as ‘not convincing’because many sherds were lying on
the surface rather than in it. The quality of the material is good, but the density appears sufficiently low to
regard this as off-site. The scatter is dissected by wide dry stone field boundaries.
Diffuse scatter, 10 by 100 m
The elongated shape of this impasto scatter, and its location alongside a recently dug leat and field
boundaries, indicates that the material may have been recently dug up. This might explain the material has
site quality, but the number of finds is not impressive.
Diffuse scatter, 15 by 25 m
This impasto scatter is located on both sides of a dry ditch, which may mean that the material was
deposited when the ditch was dug. Since the material is very worn, it appears that this cannot have
happened recently.
12
F I E L D W O R K I N T H E S A L E N T O M URGE
Area 200
ID
18
19
20
23
25
26
27
28
30
35
36
Description
small diffuse scatter
Contains only 7 fragments.
Very large (500 by 200 m) diffuse scatter, with core area 100 by 100 m
This impasto scatter, of standard type, has a very complicated shape, with many internal variations in
density. The highest quality material occupies both sides of a dry valley, with lower quality material
extending in lobes along low ridges. However, the difference may be due to current building and
landscaping activities going on in the former; ‘nests’ of high quality impasto in blocks 202-10 and -11
indicate that fresh material is still being ploughed up.
The larger part of the material (95% or more) is between ca. 1 and 2 cm thick. The forms are jar- and bowllike (olle and scodelle, ciotole); most of the simple band handles present among the material belong to this
category of large pottery forms. No complex handles such as are known from the middle and later Bronze
age were present. Only a few sherds of a black burnished thin-walled impasto were recorded that must
have belonged to cups (tazze). .Additional finds within this scatter include some Neolithic pottery and
fragments of naturally occurring iron ore, possibly related to the iron slag found mainly in neighbouring field
211.
Diffuse elongated scatter, 120 by 20 m
This scatter does not contain any diagnostic forms, and all sherds are medium-walled except for one tazza
fragment in block 203-12. It is very similar in type and density of material to site 19.
Diffuse scatter, diam 20 m
Located on a hill top.
Diffuse scatter, diam 20 m, with 30 m radius halo to its N
This scatter initially contained only 7 impasto fragments, but a later revisit turned up an additional 21.
Diffuse elongated scatter, 100 by 40 m
Blocks 13 and 16 to the east of this scatter were disturbed and therefore not surveyed.
Diffuse scatter, diam 40 m
Scatter with dense (2-5 sherds/m2) core of 30 m diam, and diffuse halo
Almost all the material in this scatter is medium-walled, with 10 diagnostic sherds. However, the quality of
preservation of material in neighbouring block 215-25 is higher.
Dense scatter, diam at least 40 m
Located on hill top; the scatter may have continued toward the north into an unsurveyed field. Some thickwalled impasto bases were recorded, but no thin-walled impasto.
Diffuse scatter, diam 50 m
Although the total number of sherds in this impasto scatter is only 11, it includes rather substantial sherds of
thick walled impasto; more-over, it may have been partially masked by Hellenistic site 22.
Diffuse scatter 100 by 100 m, with small core 25 by 50 m at southern end
The core of this scatter seems to be located in blocks 112-14 and 112-13, and contains some very large
lumps of impasto, one teglia lug, and one thick base fragment. Probably the scatter originally extended
further toward the south, but a recent walled garden and buildings prevented further survey in this area.
Figure 10 summarizes the results of this study by giving the highest class occurring within each recording
unit; it can be seen that higher quality generally occurs on the high Murge whereas in area 100 much postdepositional wear and tear has occurred. That the latter play an important role in both areas is confirmed
by the fragmentation of this material, due to frequent turning of the soil and long exposure to sun and
rain: although some scatters still contain fragments measuring 6 by 6 cm or larger, 90 % of the material
was fragmented to 4 by 4 cm or smaller. No recent fresh fractures were noted in the material from area
200, all sherds having been exposed at the surface for several cycles of tillage. No sherds were classified as
class 1 (severely rolled), the bulk of the sherds being of class 2 (worn and fragmented by tilling and
exposure). The quality of some of the finds, classified as class 3 (recognizable fragments) and 4 (sherds
with surface treatment and form characteristics), indicates that new material is still being taken up into the
ploughzone. The quality of the impasto sherds from area 100 was generally lower than that of area 200,
and (parts of) many find groups were classified in class 1, meaning that the sherds have been severely
affected by post-depositional processes. This is probably also reflected in the fact that many of the
impasto sherds no longer seem to be related to clear scatters and were recorded as off-site material.
Impasto finds classified in class 2 and 3, in contrast, are indicative of actual sites ploughed to the surface
within the units where they were collected.
The protohistoric pattern emerging from the Ostuni99 survey poses interpretative problems relevant to
11 -
13
V AN L EUSEN - P ATTERN TO P ROCESS
the ongoing general debate regarding the definition and practical implementation of site/non-site/off-site
concepts. Following the tenets of distributional archaeology (Ebert 1992), we might have only recorded
distributions of archaeological materials at the chosen spatial resolution of 50 by 50 m and coverage of
20%; in practice we recognised, and therefore wanted to record, patterning at higher resolutions - ‘sites’.
Typically, the observation of ADAB (abnormal density above background) for any of the materials
occurring in the area would lead to the definition of a ‘site’in the field, and a more detailed investigation
would follow in order to define the distribution of materials within it and to recover diagnostic materials9.
Of course, some density fluctuations (sites) might not be observed or defined in the field, whereas others,
defined in the field, might subsequently lose their site status during finds processing and analysis.
In the light of this definition, the pattern of protohistoric material revealed by our survey presents us with
considerable interpretative difficulties. Inasmuch as one would expect sites being ploughed up to exhibit
both greater quantity and higher quality of material, it is noteworthy that areas of high quality impasto
(figure 10) in many cases do not coincide with impasto scatters observed during the survey (figures 11 and
12). Two possible explanations for this phenomenon suggest themselves:
a) areas of high quality impasto occur where they do because sites are present in the subsoil only there,
and are being brought to the surface through general tillage. This is a weak explanation because it
relies on the absence of evidence;
b) areas of high quality impasto are indicative of underlying archaeological sites having been ploughed up
recently; conversely, areas of medium or low quality impasto represent similar sites that were ploughed
up earlier or more often. One example of this is site 2, recorded as a recognizable scatter in the field,
but consisting of material ‘not of site quality’. This explanation would imply that areas of high quality
impasto vary from year to year depending on burial and tillage, and leads to the possibility that most
of the landscape qualifies as a ‘site’.
Neither manuring nor plough scatter can be invoked as an explanation for the occurrence of off-site
material in this period, the former because it presupposes high-intensity land use, the latter because
experimental work indicates that ploughing cannot disperse sherds very widely. Whilst material classified
in class 1 (severely worn) may conceivably be redeposited by natural and man-made causes, in most cases
the observed severe fragmentation and wear will be due to the long and intensive cultivation of the olive
groves where the survey took place, and cannot be taken to indicate ‘off-site’uses10. Thus we are forced
back to the position that the generally occurring low densities of worn protohistoric finds must represent
a palimpsest of occasional but recurrent, low intensity, social activities.
The ‘aging’of surface archaeological material through repeated tillage presents us with a problem: should
high quality material be treated any differently from lower quality material that shows no evidence for
rolling? The issue is further complicated by the occurrence of localised soil disturbance and movement in
the course of building work and agricultural improvement, of which there was much evidence in the
region. The occurrence of high quality archaeological material may signal that such work has recently
taken place. If we have evidence that a scatter resulted from ditch-digging nearby or from variations in the
tillage of a deep valley soil, what does that say about the size/shape of the underlying site? Thus, while
concepts of site and off-site are practical in field situations, and are needed in order to compare our results
with those of other surveys in the region and elsewhere in Italy, the two approaches are not well integrated
from a methodological perspective. It is unlikely that these issues can be resolved without further
fieldwork, especially a programme of test excavations directed at understanding the range of relations
between subsoil and ploughsoil archaeology.
9
Conversely, the term ‘off-site’was applied to the occurrence of materials in ‘background’densities (the B in ADAB).
As examples of the former, doubts were recorded in the field about the originality of a number of sites (nos. 11, 12, 16, 21). In
addition, Coppola (pers. comm.) suggests that much of the material currently present on the Agnano terrace may be secondarily
deposited from the cave sanctuary area of S. Maria d’Agnano itself.
10
11 -
14
F I E L D W O R K I N T H E S A L E N T O M URGE
3.3
THE HELLENISTIC AND ROMAN PERIODS
The raw and corrected density distribution maps of Hellenistic material recovered by our survey (figures
11 and 12) are similar to those presenting the protohistoric material in one sense: the material in area 100
appears to be much more scattered and fragmented than it is in area 200, where two clearly separate
scatters (nos. 22 and 24) were identified in the field. In area 100 a targeted re-survey of units which had
yielded some scraps of classical fine wares was needed in order to turn up a sufficient amount of
additional material to allow us to define scatter cores. The resulting pattern may still be somewhat biased
by the preponderance of recent and subrecent material in parts of area 100 (see figure 5), which would
have prevented the relatively inexperienced student walkers from reliably recording similar-looking
classical material. Sites 14 and 15 are within 200 m of each other, and may be parts of the same settlement;
the similarity in the location of the three Murge sites, all on the upper part of slopes with a commanding
view of the valley beneath, is remarkable.
Figure 11: Raw density distribution maps of early Hellenistic finds in areas 100 and 200.
Figure 12: Corrected density distribution maps of early Hellenistic finds in areas 100 and 200.
Comparison of figures 11/12 and 13/14 shows that there is a large measure of continuity in the pattern of
settlement from the Hellenistic to the Roman Imperial period, although the density of finds from the
Roman period is higher. In this phase, in area 100, sites 14 and 15 are still paired and 14 is now more
clearly part of the halo surrounding 15; the small site 5 now appears to belong with a similar partner site
11 -
15
V AN L EUSEN - P ATTERN TO P ROCESS
32, which is situated in a sheltered location just beneath the Agnano terrace and the Zaccaria cave site.
Unfortunately the terrain directly to its south-east was quarried for limestone. Table 2 provides
descriptions of the Hellenistic and Roman sites.
Figure 13: Raw density distribution maps of Roman Imperial finds in areas 100 and 200.
Figure 14: Corrected density distribution maps of Roman Imperial finds in areas 100 and 200.
3.4
POST-ANTIQUE TO RECENT
Although recent and subrecent material was only recorded by us because it is a significant factor in biasing
the recovery rate of other material groups, we did hope to identify Byzantine and Medieval material
despite not having an appropriate specialist on our team. As it turned out, the only probably post-antique
material we could identify turned up as a result of the impasto quality study; as yet unidentified, it was
termed ‘fabric X’. The fabric is sandy, and is fired very hard; it appears to be used for both tiles (typically
between 1.8 – 2.0 cm thick) and pots; its predominant colour is reddish yellow 7.5YR 7/8-6/8. Figure 15
and Table 3 provide an overview of the distribution of this material. Clearly, most of the material forms
two scatters, located on opposing edges of the Agnano terrace; a few sherds in the northern part of the
surveyed area indicate that a third site may be nearby. Further work is needed to identify and date this
11 -
16
F I E L D W O R K I N T H E S A L E N T O M URGE
fabric.
Figure 15: Corrected density distribution map of impasto sherds with fabric ‘X’(per ha) in area 100.
Table 2: Hellenistic and Roman sites (for locations, see figure 6)
Area 100
ID
1
5
14
15
17
32
Description
Diffuse scatter of Hellenistic pottery, diam 50 m
Scatter consists of fine wares (Apulian Black Gloss, late Republican Grey Ware), coarse and plain wares
and amphora (Corinthian A). Amount of tiles very low. Near to the core of the site a Gnathia sherd was
found, indicative of an early Hellenistic burial.
Small and diffuse scatter of Hellenistic and Roman Imperial tile and pottery, diam 10 m
Located on terraced slope, this scatter contains mostly tiles, and only one fine ware sherd (Italian sigillata).
Diffuse scatter of Hellenistic and Roman Imperial tile and pottery, 50 by 20 m
The long axis of this scatter is located along a field boundary. The pottery consists notably of fine slip wares
(mainly Apulian Black Gloss, Red Slip wares), amphoras and coarse – kitchen wares. The major phases
are the early Hellenistic and the middle to late Roman Imperial periods (to well within the 6th century AD;
witness the presence of Byzantine/Palestine amphoras)
Dense scatter of Hellenistic and Roman Imperial tile and pottery, core 50 x 30 m
This scatter has a considerable halo and contains large amounts of diagnostic ceramics. The pottery
consists notably of fine slip wares (Apulian Black Gloss, Italian Sigillata, Red Slip wares), amphoras and
coarse – kitchen wares. The major phases are the early Hellenistic and the Roman Imperial periods (to well
within the 6th century AD; witness the presence of Byzantine/Palestine amphoras). The presence of Apulian
Red Figured sherds is indicative of an early Hellenistic graveyard.
Dense scatter of tile and pottery, 50 x 15 m; overall density not known
The long axis of this scatter is located along a field boundary; it probably represents a Hellenistic farmstead
since multiple Apulian Black Gloss were found.
Diffuse scatter of Roman pottery and tile, diam 25 m
Few diagnostic ceramics, datable to the 2nd/3rd century AD (African Red Slip and San Foca).
11 -
17
V AN L EUSEN - P ATTERN TO P ROCESS
Area 200
ID
22
24
29
Description
Dense scatter of Hellenistic and Roman pottery and roof tiles; core diam 75 m, with 50 m halo (up to
100 m downslope)
Fine wares include Apulian Black Gloss, late Republican Grey Ware, Italian sigillata and early African Red
Slip wares; probable Roman building blocks were re-used in terraces. Same material possibly re-used
during recent soil additions higher up on same spur (Block 9). The scatter can be interpreted as the nucleus
of a hilltop farm that was continuously occupied from the early Hellenistic period to within the 2nd century
AD.
Diffuse scatter of ancient pottery and tile, 50 by 50 m
The scatter is located along a field boundary. In 1999 only undiagnostic material was found here. In 2000, a
revisit to the surrounding area revealed the existence of an extensive and dense Hellenistic-Roman site in
the fields bordering the scatter to the north. In view of this, the scatter can be argued to constitute the
periphery or halo of that site. The material still has to be analysed in detail.
Dense scatter of ancient tiles and pottery, diam 20 m
The material is not very diagnostic and has, as yet, not been analysed.
Table 3: Post-antique sites (for locations, see figure 6)
Area 100
ID
4
9
31
Description
Dense scatter of fabric ‘X’, 25 by 10 m
A large percentage of the sherds from this scatter belong to tiles of which in some cases the rims were
preserved, but thinner potsherds were also recorded. The quality of the sherds is high (class 3) and the
scatter appears to be homogeneous. We probably deal here with a post-antique structure in local fabric,
although the siting on a steep terraced hillside is strange. Provenance of finds unclear; could be from site
on edge of Agnano terrace.
Dense scatter, at least 250 m wide but could be larger toward S and W
The top 10 m of this scatter, containing some 10% coarse ware or fabric X, are located on the edge of the
Agnano terrace; it continues another 50 m down the terraced slope
Diffuse scatter, diam 10 m
Turns out to be all post-antique material.
4
DISCUSSIO N
As stated in section 1, the primary survey data presented in the previous section are analysed in the
context of the three major (supra-) regional developments – centralisation, early urbanisation, and Roman
colonisation – introduced in section 1.2. Below, the outcome of this analysis will be presented
chronologically.
4.1
CENTRALISATION O F SETTLEMENT IN THE LATE BRONZE AGE AND EARLY IRON
AGE
The detail of the survey method used in the Murge proved successful in locating even small, low density
and low visibility middle Bronze Age (18th – 15th centuries BC) artefact scatters. In both survey areas a
large number of such ephemeral scatters was found alongside a series of larger and more dense
concentrations, suggestive of an extensive human exploitation of both up- and lowland areas in this phase
of the Bronze Age (figure 12). With regard to locational preferences, one may observe that the major
Bronze Age scatters in area 200 cluster on the lower terraces at the interface between hill slopes and valley
floors. A similar pattern is attested near the contemporaneous large Bronze Age site of Masseria Carestia,
some 3 km to the east in the same landscape unit (personal communication d.ssa A. Cinquepalmi). This
preference can be explained in reference to the classification of the lower parts of the slopes and the valley
floors as suitable to palaeotechnic wheat farming, due to their relative flatness, nutrient availability and
workability. The same holds true for the concavely sloping land and the lowland plain proper, where
middle Bronze Age sites abound as well. Whereas the sheer abundance of middle Bronze Age material in
11 -
18
F I E L D W O R K I N T H E S A L E N T O M URGE
all areas makes it likely that the settlements were not all permanently occupied contemporaneously
(explanations should be sought in the context of group mobility related to the practice of shifting
cultivation, necessitated by the short term fallowing system practised in the Bronze Age), the homogeneity
of the material argues for a relatively short period of use. Unfortunately, due to the undiagnostic nature of
the finds we cannot assign this settlement period to any particular phase within the MBA.
The highly dispersed middle Bronze Age pattern established by our survey contrasts with that of the late
Bronze Age (LBA; 14th - 12th centuries BC). In the survey areas no definite LBA material was found.
However, this does not mean that the Murge area was abandoned in this phase. Whereas in the Salento
region as a whole, LBA sites are mainly concentrated along the coast line, in the Murge they are also found
at the top of the steep slopes at the interface of the coastal area and the upland. A good example is offered
by the site of Rissieddi, which stretches out on one of the promontories along this edge, immediately
north of area 100 (see figure 2). Now largely built over, until recently sections of a stone fortification
circuit could be observed to enclose a densely occupied area of some 2-3 ha (Coppola 1983:208-213). The
Rissieddi promontory visually dominates the surrounding landscape, as do similar LBA sites in the Murge.
On the basis of the available data one can conclude that, by the LBA, a strongly centralized settlement
pattern had emerged, contrasting with the highly diffused one of the MBA.
Much less is known about the final Bronze Age and initial Iron Age (11th - 9th centuries BC; the ‘Dark
Ages’). For the Salento peninsula as a whole, the known sites are still situated predominantly on the coast.
After the collapse of the Mycenaean world no imports from that region reach Salento, suggesting that
overseas, and perhaps even interregional, networks had collapsed. Coastal communities were probably
autarchic, while the interior may only have been exploited for extensive pastoralism, if at all (Burgers
1998:173-174). This pattern was gradually transformed in the course of the 8th century BC. In the study
area one may observe a shift in locational preferences in this phase, the Rissieddi plateau being abandoned
in favour of a new settlement on the more accessible hilltop at Ostuni (see figure 2). Iron Age occupation
at Ostuni is attested from the 8th century BC onwards (Coppola 1983:235-254). This is congruent with
recent theories on settlement expansion and shifts in site locations in 8th century BC Salento in general
(D’Andria 1991:405; Yntema 1993-1:161; Burgers 1998:186-191). The early Iron Age Salento is
characterized by a gradual increase in site density, an expansion of already existing sites, and an occupation
shift from coastal promontories to inland plains and hills, suggestive of an internal colonisation
movement. In addition to Ostuni itself, a whole series of Murge sites illustrate this phenomenon, such as
Locorotondo (figure 1; De Michele 1986), S. Pietro di Ceglie Messapico (figure 2; Fusco 1964; Coppola
1977:304), and Castello di S. Vito dei Normanni (figure 2; Cocchiaro 1998; Semeraro 1998). Still, judging
by the relatively large distances between these sites and by the absence of Iron Age artefacts in rural
survey areas such as ours, most of the Murge is likely to have been exploited only marginally.
4.2
EARLY URBANISATI ON AND RURAL INFILL
No direct evidence for the emergence of urban sites could of course be obtained from our rural surveys at
Ostuni, so we must turn to previous research in order to put our results into perspective. Artefact scatters
dating to the Archaic/Classical period (600 – 325 BC) were absent in the survey areas, suggesting that
most of the high Murge was still not used intensively in these phases, and that settlement continued to
concentrate in the same areas as in the previous Iron Age. Indeed, it may be hypothesized that the
nucleated settlement pattern that emerged in the Salento Murge in the early Iron Age was consolidated in
the Archaic/Classical period. Unfortunately, due to the scarcity of systematic investigations at major
Murge sites such as Ostuni, Ceglie Messapico, and Martina Franca (see figure 1), no proto-urban intra-site
studies are as yet available that would allow us to evaluate if these sites underwent changes similar to those
detected at the southern lowland sites of Cavallino, Oria and Valesio (see figure 1)11. At Cavallino,
excavations point to a trend towards the replacement of dispersed early Iron Age hut compounds by
aggregate blocks of (partly) stone built houses with tile covered roofs, and traces of such houses have also
On Cavallino, see notably Pancrazzi 1979 and D’Andria 1991; on Oria, Andreassi 1981; Yntema 1993-1; on Valesio,
Boersma/Yntema 1987.
11
11 -
19
V AN L EUSEN - P ATTERN TO P ROCESS
been found at Valesio. At Cavallino, this reorganization of settlement space was shown to have been
accentuated by, amongst others, the arrangement of clearly defined paved roads and public spaces, as well
as by the construction of monumental stone defensive circuits surrounding the inhabited areas. Similar
defenses have been traced at Oria and Mass. Fani (figure 1; Andreassi 1981; Descoeudres/Robinson
1993).
The emergence of early urban features at these sites can be shown to coincide with a reorganization of
religious space. This becomes particularly evident when the wider landscape is taken into consideration as
well; whereas early Iron Age religious activities remain largely invisible in the archaeological record, for the
Archaic/Classical phases one may observe a process of formalization of cult activities, notably in caves in
the territories of the major settlements. Evidence for the occurrence of this process in our study area is
provided by the cave sanctuary of the Grotta di S. Maria d’Agnano, just a little distance uphill from survey
area 100 (see figure 2). The cave is located some four km north of Ostuni, and opens out onto a natural
terrace overlooking the coastal plain, just below the cliff top harbouring the site of Rissieddi.
Investigations carried out under the direction of Coppola have demonstrated that the cave became the
scene of a formal cult dedicated to a female divinity from the 6th century onwards (Coppola 1983:249252). Following the arguments we have put forward elsewhere regarding such sanctuaries (Burgers,
forthcoming), we believe that this reorganisation not only involved a transformation in the perception of
the landscape, but also a formalisation of territorial claims.
In view of the above, it is all the more remarkable that the results of our survey show that it was only in
the early Hellenistic period (325 – 200 BC) that these territories attracted any substantial rural settlement.
In area 100, following a pattern observed earlier at Oria (Yntema 1993-1), these sites appear to be located
along a line running just beneath and parallel to the Murge edge, suggesting the existence of a pedemontana
road (see figure 11). Unlike similar farms in the Oria and Valesio survey areas, some of these sites lack
identifiable tile, which suggests that a much simpler construction method was used for farm buildings. In
both survey areas these early Hellenistic sites are surrounded by low density ‘halos’indicative of manuring,
while in area 200 they occur on slopes that were largely avoided in the Bronze Age. It may be proposed
that with fertilisation (and possibly irrigation), these slopes had been made suitable for agriculture.
Considering the observed emphasis on polyculture in other parts of Salento from this time onwards, it is
tempting to suggest that the slopes may have been taken into use for arboriculture in a manner analogous
to the 19th century situation referred to in section 212.
Although the latter hypothesis can not be proven as yet, these results may be called highly significant. The
Salento Murge was thought to be an archaeologically marginal landscape even in the Hellenistic period
until the sites and off-site material detected by our survey showed that the Murge in fact participated in
the regional trend of agricultural expansion and intensification, established on the basis of the spread of
early Hellenistic farmsteads in urban catchment survey areas on the Salento Isthmus (Burgers 1998:226263). Considering that in the Murge the rural infill is also attested in the upland survey area 200, at 7-8 km
from the nearest urban centers at Ceglie Messapico and Ostuni, one may suggest that this regional trend
extended to the cultivation of even the most outlying, previously untilled, lands outside the urban
catchment areas.
The agricultural transformation indicated by these results is linked to other major contemporaneous early
Hellenistic developments. It can be argued to have supported a demographic increase and urban
rearrangement of the larger settlements, which reached their maximum expansion in this period. Since we
have only cemetery evidence from Murge sites, a parallel must be drawn with other Salento sites where
systematic urban surveys point to a considerable expansion (Yntema 1993-2; Burgers 1998). Excavations
at several of these sites likewise show the emergence of large nucleated insulae with a relatively regular layout (figure 1; Monte Sannace: Scarfi 1961; 1962; Vaste: D’Andria 1991; Muro Tenente: van Alberda et al.
1999). Besides these domestic quarters, specific urban spaces were arranged for public buildings, cult places, cemeteries, warehouses and for intensive horticulture. Fortification circuits were built to enclose
12
On the development of a polyculture system, see Burgers 1998:255-259.
11 -
20
F I E L D W O R K I N T H E S A L E N T O M URGE
entire settlement areas - as was the case at Ostuni and Ceglie Messapico in the Murge (D’Andria 1991:445;
Coppola 1983:269-275). These urban transformations can be shown to have occurred simultaneously at
sites throughout the Salento peninsula. Relating them to the rural infill of the landscape surrounding these
major sites, one may observe the parallel emergence of pronounced local settlement hierarchies
throughout the region (as discussed in more detail in Burgers 1998:226-263). On the basis of our survey,
the Murge area can be considered one such micro-region, with isolated farmsteads even appearing on the
rolling land of the high plateau.
It can be concluded that, with the rearrangement of the wider landscape, the Murge towns became central
places serving extensive rural hinterlands in the early Hellenistic period. Indicative of this development is
the variability in the ceramic repertoires of the rural sites detected in both survey areas - besides tiles and
local coarse kitchen and plain wares, these include fine wares such as Apulian Black Gloss, Gnathia ware
and Apulian Red Figure, suggesting articulation with a market system.
4.3
THE ROMAN LANDSC APE
Neither survey area in the Ostuni transect shows significant changes in the number or location of sites of
the late Republican and Imperial periods, as compared to the early Hellenistic period. A basic continuity
seems to exist for these phases, suggesting that the Murge landscape was not much affected when the
centre of power in the region shifted towards the Brindisi plain as the direct hinterland of the Latin colony
of Brundisium (see figure 1).
Brundisium was founded around the middle of the 3rd century BC as Rome’s satellite in the newly
conquered Salento peninsula. In the late Republican period it grew into one of the major Italian harbours
for communication with the eastern Mediterranean, as well as into a regional centre for the overseas
export of agricultural products, notably wine and olive oil. With the emergence of an export oriented
market economy, there is evidence to suggest the formation of a regional landscape which was
differentiated in various zones of profitability (Burgers 1998:265-292). Market oriented wine and olive oil
production can be argued to have concentrated in the immediate hinterland of Brindisi, as well as along
the major new transport axis of the Via Appia. As a consequence, the role of the former Hellenistic towns
as central places in these areas lost much of its significance, which may be one of the major reasons for
their decline.
Areas further removed from Brindisi and the Via Appia (including the Ostuni survey areas) must also have
become economically peripheral, with subsistence farming likely to have prevailed in the late Republican
period. The aforementioned lack of systematic research at towns in the Murge warns against drawing
definite conclusions regarding their development in the Roman period. However, judging from the
scarcity of Roman finds so far, they do seem to have contracted considerably. On the other hand, the rural
landscape of neither the Murge upland nor the plain seems to have suffered similarly, as the Roman sites
in the survey areas demonstrate, and the presence of considerable amounts of fine wares on these sites
suggests that the area as a whole still had access to a wider market system.
This rural continuity also holds good for the Roman Imperial period, at least for the lowland survey area
(area 100), for which African Red Slip wares are amply attested. There is even some evidence of site
expansion - at site 15 Imperial Roman artefacts are distributed over a much wider area than the Hellenistic
sherds. In comparison to the other Imperial scatters discovered here, this site clearly stands out not only
in extent but also in the amounts of fine wares found. In view of this, it can be interpreted as the central
part (with domestic unit) of a larger estate which also harboured a range of utilitarian outbuildings. A
similar development has been observed in other Salento survey areas, where Roman Imperial sites also
expand progressively (Boersma et al. 1991; Yntema 1993-1:215-226), and may be suggested to reflect a
process of concentration of small, dispersed land holdings into larger, more centrally managed estates. If
this was the case in the lowland around Ostuni, it is tempting to relate it to an expansion of olive culture
in this area, which as we have seen above (section 2.2) is already described in historical documents of the
16th century as ‘a forest of olive trees’and which remained so until the present day. Central estate
management is a prerequisite for large scale olive culture. In this regard, one may also point to the
11 -
21
V AN L EUSEN - P ATTERN TO P ROCESS
existence of the nearby Via Traiana, running all along the coast from the Bari district to Brindisi. Its
course has been studied by Uggeri, who relates the ancient written sources mentioning the road to actual
field observations (Uggeri 1983:228-264). The Via Traiana is likely to have considerably improved the
accessibility of this part of the Murge, not least for the transport of agricultural products. In contrast,
Imperial fine wares are less conspicuous in the upland Cervarolo zone. Other contemporary diagnostic
wares lack as well, suggesting that the area was abandoned in the 2nd century AD and that the Murge
upland now became peripheral to the wider region.
5
CONCLUDING DISCUSSION
From a methodological point of view, the Ostuni’99 survey has had some very interesting results. The
three visibility bias factors discussed in section 1.4 indicate that the recent and subrecent land use history
of the study area has had a significant influence on the results of our survey. The impasto sherd quality
study reported in section 3 revealed a lacuna in our understanding of surface pottery distributions, relevant
to the ongoing general site/off-site/non-site debate. While we may not be able to attach any secure
interpretation to the occurrence of areas of higher sherd quality such as those mapped in figure 10, it is in
our opinion a phenomenon worth looking into by means of a programme of test excavations. These
concerns must be borne in mind when reading the analysis of the use of the landscape in each period.
Despite the unsolved problem of insecure dating of the Bronze Age impasto, the Ostuni survey
conducted by the RPC project has considerably strengthened and widened the basis for generalisations on
the occupational history of the Salento Murge. The sample areas investigated represent two of the major
landscape units of this micro-region. Contrary to expectations, both have given up a relative abundance of
surface material dating to the Bronze Age and the Hellenistic/Roman period. By applying a detailed
survey method, focusing on the documentation of the density and distribution of artefacts rather than
sites, it has become possible to assess accurately the variability in quality and quantity of this surface
material in the light of both cultural and natural formation processes. Moreover, relating the results of the
survey to those of the land evaluation, hypotheses on land use patterns could be more firmly grounded.
On the micro-regional scale of interpretation, one of the major conclusions that can be drawn on the basis
of the surveys is that, in broad outline, both sample areas demonstrate parallel shifts in artefact densities
and distributions from the Bronze Age to within the early Imperial period. Judging by the highly dispersed
pattern of Middle Bronze Age scatters delineated in both survey areas, the Murge plateau and the lowland
zone can be argued to have been extensively exploited during this phase. This rather mobile settlement
system of the MBA contrasts strongly with that attested for the late Bronze Age, which is of the nucleated
type. A similar nucleated pattern, now including the interior of the Murge, came into being in the course
of the early Iron Age, when a number of large sites appeared on selected strategic positions. The latter
pattern can be shown to have continued to well within the 4th century BC, a period in which, according to
the survey results, the rural landscape surrounding the major Murge sites seems to have been exploited
only marginally. Archaeological material dating to these centuries is restricted to the cave site of S. Maria
d’Agnano, where the Archaic formalisation of cult activities can be argued to have supported territorial
claims on the surrounding land.
Our survey indicates that it is only from the late 4th century BC that the rural landscape of both the coastal
zone and the Murge plateau was actually claimed for settlement. In view of this, it can now be argued that
the Salento Murge participated in the general south-Italian trend of agricultural expansion and
intensification that accompanied ongoing urbanization. The recognition that this trend involved even
areas on the high Murge, far from the immediate catchments of the major Murge towns, may illustrate the
scale of the process.
For both survey areas, a basic continuity of occupation throughout the Roman Republican and Imperial
periods can be deduced, with the possible exception of the late Imperial period in the upland area. This
situation contrasts with that known from incidental discoveries at the major Murge sites, which seem to
11 -
22
F I E L D W O R K I N T H E S A L E N T O M URGE
have been largely abandoned in the late Republican period. If this may be taken to indicate that these sites
lost their previous central market role, subsistence farming is likely to have come to prevail in the
surrounding rural territories. As for the Imperial period, the expansion and intra-site differentiation of the
largest site attested in the lowland survey area can be interpreted in the light of a concentration of land
holdings in a more centrally managed estate focused on large scale olive growing.
In line with the central RPC project aim of comparative regional analysis, the settlement history outlined
above is studied from a (supra-) regional perspective. The Salento region has most often been treated in
the context of Hellenization studies, focusing on the demonstration of the diffusion of Greek culture
among non-Greek populations in Southern Italy (cf. notably Whitehouse and Wilkins 1989). This
diffusion is generally conceptualized as a unilinear process of increasing intensity, enhanced in particular
by the installation of Greek colonies along the Ionic shores (notably Taras for the Salento Isthmus), and is
thought sufficient proof of Greek cultural domination over their native neighbours. The recent upsurge of
problem oriented fieldwork in southern Italy, and in the Salento peninsula in particular, allows us to
qualify this strongly culture-historical, diffusionist paradigm. Cherishing a wide chronological and spatial
scope, the fieldwork conducted by the RPC project offers the possibility of research into long-term, supraregional settlement and landscape dynamics. From this perspective, the early Greek intrusions can be
argued to have been not a dominant but rather a peripheral element in a region-wide process of internal
colonisation, settlement expansion and corresponding shifts in territorial perspectives and claims (see
above, section 4.1). This process, which can be shown to have started before the arrival of Greeks, is also
attested in the Murge area, where a range of new sites emerged on strategic locations in the early Iron Age.
Similarly, it has now become possible to overcome the traditional emphasis on Greek polis formation for
the subsequent centuries and to point to the existence of contemporary dynamic processes occurring in
the native world, especially those of early urbanization and agricultural intensification. These processes
involved, amongst others, the emergence of pronounced settlement hierarchies in micro-regions
throughout the Salento Isthmus, and can be related to the formation of centralised socio-political power
structures increasingly integrating previously segmented tribal units (Burgers 1998:195-263).
From the same diachronic and supraregional perspective the Roman landscape of the Murge must be
interpreted in the context of the progressive incorporation of Salento into a developing state and market
system dominated by Rome. This further enhanced political and economic centralisation, favouring
notably the Latin colony of Brundisium and its immediate hinterland in the Brindisi plain. Although the
Ostuni survey results indicate that neither the uplands nor the lowlands were abandoned until the later
Imperial phases, the Salento Murge as a whole can be argued to have become increasingly peripheral from
the late Republican period onwards.
ACKNOWLEDGMENTS
The RPC project is carried out in close collaboration with the Scuola di Specializzazione in Archeologia Classica e
Medievale of the University of Lecce and the Soprintendenza archeologica della Puglia. We would like to thank
our colleagues at both institutes, and in particular prof. Francesco D’Andria, prof. Mario Lombardo and
dr. Grazia Semeraro of Lecce University and dott. Giuseppe Andreassi, d.ssa Assunta Cocchiaro and d.ssa
Angela Cinquepalmi, respectively superintendent and inspectors of the Soprintendenza. The Ostuni
survey was subsidized by the Netherlands Organization for Scientific Research and the Museo delle Civiltà
preclassiche delle Murge at Ostuni. We sincerely thank prof. Vincenzo Capetta and prof. Donato Coppola of
the Ostuni museum for their generous hospitality and constant support. Further thanks are due to prof.
Douwe Yntema of the AIVU and dott. Biagio Giaccio, who respectively studied the classical ceramics and
the lithic material collected during the surveys, and to Rik Feiken and Jaap Fokkema, who carefully
digitized the topographic and distribution maps. The computer drawings in this article were made by Jaap
Fokkema. We are also grateful for the opportunity to use the preliminary results of the physicalgeographical research and land evaluation, carried out by MA students Saar Foeken and Saskia Gietema
under the direction of Esther van Joolen and in close collaboration with dr. Jan Delvigne and Bas Bijl.
Finally, we are greatly indebted to our team of surveyors from Belgium as well as the Netherlands, and to
the farmers who allowed us to trudge their fields.
11 -
23
V AN L EUSEN - P ATTERN TO P ROCESS
REFERENCES
Alberda van K., Burgers, G.-J., Burgers, H., Karel, D., Yntema, D 1999: Muro Tenente. Centro messapico nel
territorio di Mesagne, a cura di A. Nitti, Manduria.
Alcock, S. 1994: Breaking up the Hellenistic world: survey and society,Classical Greece. Ancient Histories and
Modern Archaeologies (ed. I. Morris), 171-190.
Andreassi, G. 1981: Oria – Brindisi, Studi Etruschi 49, 466-468.
Attema, P.A.J., Burgers, G.-J., Kleibrink, M., Yntema, D. 1998-1: Centralisation, early urbanisation and
colonisation in a regional context, dutch excavations and landscape archaeology in central an
d southern
Italy, Saguntum 31: 125-132.
Attema, P.A.J., Burgers, G.-J., Kleibrink, M., Yntema, D. 1998-2: Case studies in indigenous developments in
early Italian centralization and urbanization, a Dutch perspective, Journal of European Archaeology 1 (3):
326-381.
Attema, P., van Joolen, E., van Leusen, M. 2001: A Marginal landscape: Field work on the beach ridge complex
near Fogliano (South Lazio), Paleohistoria 41/42 (1999/2000): 149-162.
Bintliff, J. 1997: Regional Survey, Demography, and the Rise ofComplex Societies in the Ancient Aegean:
Core-Periphery, Neo-Malthusian, and Other Interpretive Models,Journal of Field Archaeology, 24(1), 1-38.
Bintliff, J, Ph. Howard & A. Snodgrass 1999: The Hidden Landscape of Prehistoric Greece,Journal of
Mediterranean Archaeology 12(2): 139-168.
Bintliff, J., Ph. Howard & A. Snodgrass 2000: Rejoinder, Journal of Mediterranean Archaeology 13(1): 116123.
Boersma, J.S. 1990, Oria and Valesio. Dutch Archaeological investigations in the Brindisi region of Southern
Italy, Mededelingen KNAW, afd. Letterkunde (new series),53, no. 3.
Boersma, J.S. 1995, Mutatio Valentia. The late Roman baths at Valesio, Salento, Amsterdam.
Boersma, J.S., Burgers, G.-J., Yntema, D. 1991: The Valesio Project: final interim report,BABesch 66, 115-131.
Boersma, J.S./Yntema, D.G. 1987: Valesio. History of an Apulian Settlement from the Iron Age to the
Late-Roman Period, Fasano di Puglia.
Burgers, G.-J. 1998: Constructing Messapian Landscapes: Settlement Dynamics, Social Organization and
Culture Contact in the Margins of Graeco-Roman Italy, Amsterdam: Gieben.
Burgers, G.-J. forthcoming: Urbanization in Magna Graecia: Settlement, Landscape and Social Dynamics in a
Regional Italic Context.
Cocchiaro, A. 1998: La ricerca archeologica in località Castello a S. Vito dei Normanni (1994-1995), in: L’Area
archeologica di Località Castello a San Vito dei Normanni. La Ricerca come risorsa, 13-18
Coppola, D. 1977: La ricerca paletnologica nel brindisino: storia degli studi e nuove prospetti
ve di indagini,
Brundisii Res IX(2), 261-306.
Coppola, D. 1983: Le origini di Ostuni. Testimonianze archeologiche degli avvicendamenti culturali, Martina
Franca.
Coppola, D. 1985: La documentazione archeologica in alcune grotte del brindisino: contributollo
a studio del
popolamento antico nella Murgia sud-orientale, in: Atti del I Convegno Regionale di Speleologia, 1-30.
D'Andria, F. 1991: Insediamenti e territorio: l'età storica, in:I Messapi, Atti del XXX Convegno di Studi sulla
Magna Grecia, Taranto - Lecce 1990, Taranto, 393-478.
De Michele, V. 1986: Locorotondo. Rinvenimenti archeologici in contrada Grofoleo. Origini di un centro
abitato della Valle d’Itria, Martina Franca.
Descoeudres, J.-P./Robinson, E. 1993: La "chiusa" alla Masseria del Fano: an early Messapian site near Salve in
the Province of Lecce, Lecce: Martano.
Ebert, J. I. 1992: Distributional Archaeology. Albuquerque: University of New Mexico Press.
FAO 1977: Guidelines for soil profile description (2nd edition). Rome.
Foeken, S., Gietema, S. 2000: Farming in the good old days: land suitability classification in Salento (Italy) for
Bronze Age and Roman agriculture. Unpublished report: Free University of Amsterdam.
Fusco, V. 1964: Ceramica messapica in un castelliere sopra Ceglie Messapico,in: Atti del VIII Riunione
dell’IIPP, 187-190.
Galt, A.H. 1991: Far from the Church Bells. Settlement and Society in an Apulian town, Cambridge University
Press.
Kamermans, H. 1993: Archeologie en landevalutie in de Agro Pontino (Lazio, Italië), Universiteit van
Amsterdam, Faculteit der ruimtelijke wetenschappen.
Orton, C.O. 2000: Sampling in Archaeology. Cambridge Manuals in Archaeology. Cambridge University Press.
Pancrazzi, O. 1979 (ed.), Cavallino I. Scavi e ricerche 1964-1967, Galatina.
Scarfì, B.M. 1961: Gioia del Colle - Scavi nella zona di Monte Sannace,Monumenti Antichi 45, 144-331.
Scarfì, B.M. 1962: Gioia del Colle - L'abitato peucetico di Monte Sannace, Notizie degli Scavi 1962, 1-283.
Semeraro, G. 1998: Scavi a San Vito dei Normanni, in:L’Area archeologica di Località Castello a San Vito dei
Normanni. La Ricerca come risorsa, 26-32
11 -
24
F I E L D W O R K I N T H E S A L E N T O M URGE
Small, A. 1991: Late Roman Rural Settlement in Basilicata and Western Apulia, in:Roman Landscapes.
Archaeological Survey in the Mediterranean Region (Barker, G./Lloyd, J. eds.), 204-222.
Small, A. et al. 1998: Field Survey in the Basentello Valley on the Basilicata
-Puglia Border, Echos du Monde
Classique/Classical Views XLII, 337-371.
Terrenato, N. 1996: Field Survey Methods in Central Italy (Etruria and Umbria),Archaeological Dialogues
3(2):216-230.
Terrenato, N. and A.J. Ammerman 1995: Visibility and Site Recovery in the Cecina Valley Survey, Italy.
Journal of Field Archaeology 23 (1996):91-110.
Uggeri, G. 1983: La viabilità romana nel Salento, Mesagne.
Van Leusen, P.M. 1998: Archaic Settlement and Early Roman Colonisation of the Lepine Foothills, in
Assemblage 4 (1998), http://www.shef.ac.uk/~assem/4/.
Vinson, P. 1972: Ancient roads between Venosa and Gravina, in:Papers of the British School at Rome 40, 5890.
Whitehouse, R.D. and Wilkins, J.B. 1989, Greeks and natives in south
-east Italy: approaches to the
archaeological evidence, in: Centre and Periphery. Comparative Studies in Archaeology (ed. Champion,
T.C.), London, 102-126.
Yntema, D.G. 1993-1: In Search of an Ancient Countryside. The Free University Field Survey at Oria, Province
of Brindisi, South Italy (1981-1983). Amsterdam.
Yntema, D.G. 1993-2: The Settlement of Valesio, Southern Italy. Final Report on the Field Survey,BABesch 68,
49-70.
11 -
25
C HAPTER 12
REGIONAL ARCHAEOLOGICAL
PAT T E R N S I N T H E S I B A R I T I D E
P RELIMINARY RESULTS O F THE RPC FIELD SURVEY CAMPAIG N 2000 ∗
Martijn van Leusen and Peter Attema
1
INTRODUCTION
This article presents a preliminary report on the October, 2000 field walking survey campaign of the
Regional Pathways to Complexity (RPC) project in the Sibaritide (northern Calabria, Italy), focusing on
the study of spatial and chronological patterns in the archaeological record and their interpretation in
terms of regional settlement history.
From 1991 onwards, the Groningen Department of Mediterranean Archaeology has conducted annual
excavations at the hilltop sanctuary and settlement areas of Timpone Motta at Francavilla Marittima, with
occasional limited field surveys in the immediate neighbourhood of that site. In 1997, the Sibaritide
became one of three regions being studied in the context of the RPC project run jointly by the
archaeology departments of the University of Groningen and the Free University of Amsterdam (both in
the Netherlands)1. The central aim of the RPC project is to study regional processes of centralisation,
urbanisation, and colonisation which have occurred from the Iron Age onwards, and especially to
evaluate the relative importance of internal and external contributing factors to these processes. Among
several approaches toward that aim is the detailed study of the relations between the regional landscape
and its settlement history at various spatial and temporal scales. This approach utilises the concept of
history developed by the Annales school in the 1930s, operationalised by Braudel after the second world
war, and introduced more recently in archaeology (cf. Bintliff 1991). It requires the modelling of
landscape characteristics within a GIS environment, and the application of careful 'source criticism' of the
available archaeological data. Fieldwork, especially survey, has been used in an attempt to fill gaps and
assess biases in these data.
∗
This chapter is currently in press, to appear in Palaeohistoria 42/43. The campaign was conducted as part of a collaboration
project with the archaeological superintendency of Calabria. We would like to express our gratitude to the Soprintendete, Dr.ssa
Silvana Luppino, for arranging the necessary permissions. Some aspects of the SIBA2000 field campaign, such as the highland
reconnaissance survey and the experiments with context-aware portable digital field assistants, are not discussed here (see chapter
7).
1 The RPC project (1997-2002) is part of the research program ‘
Landscape and settlement’of the Netherlands Foundation for
Scientific Research NWO.
12 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
2
THE SIBARITIDE
The Sibaritide is an alluvial plain on the Gulf of Taranto (part of the Ionian Sea), bounded on the
landward side by the Pollino massif (to the north) and the Sila mountain range (to the south). It is named
after the ancient Greek colony of Sybaris, which lies in the approximate centre of its coastline. Since, on
the one hand, its history of settlement and land use has been largely determined by the possibilities
afforded by the physical and natural environment and, on the other, its archaeological record has been
conditioned (through the medium of recent land use and land cover) by that same environment, a brief
review of the geology, climate, and land use of the study area will be given here.
Figure 1 - Major landscape units of the Sibaritide. The coastal and alluvial plains (D) are
surrounded by terraced (A) and non-terraced (B) marine and fluvial sands and conglomerates
which merge into the lower slopes of the Pollino and Sila ranges (C)
GEOLOGY
The Sibaritide is a roughly triangular coastal lowland, formed mainly in the Quaternary by the
accumulation of erosion products from the Pollino mountain range to the north and the Sila mountains
to the south. A series of stepped fossil coastlines, up to an altitude of several hundred meters above the
present sea level, can be recognised in these loose sediments and conglomerates. Each is represented by a
relatively steeply sloping 'cliff' and a weakly sloping (5%) fan-like 'terrace'. These stepped coastlines
possibly were the result of the interplay of tectonic uplift of the area and changing sea levels due to
12 -
2
F IELD W O R K
IN THE
S IBARITIDE
climate changes in the Quaternary. Fluvial erosion and deposition must have accompanied these changes
in climate through changes in vegetation cover, river regime and sea level. As vegetation cover increased
at the onset of the Holocene, the rate at which sediment was transported must have slowed down.
However, the sedimentation rate in the coastal plain could still have been considerable in human terms,
and the alluvial deposition of several meters of sediments could easily have changed local
hydrogeography. Erosion in the hinterland and sedimentation in the coastal lowlands are ongoing
processes, though now mostly as a consequence of violent winter storms. Their effects, however, have in
historic times been minimised by regulatory works such as river bank reinforcements.
Given the very dynamic geology of the Sibaritide, in which for example the Greek colony at Sybaris lies
below 6 to 8 meters of alluvial deposits, the archaeological record of the plain can only be interpreted
with reference to palaeo-geographical reconstructions. However, the late Holocene deposits are generally
too thick for a manual auguring programme to succeed, although the mechanical augerings taken in the
course of the 1960s search for Sybaris demonstrate that a dedicated coring program is not impossible
(Rainey & Lerici 1967)2. Turning to the mountainous inland parts of the Sibaritide, geomorphology and
land use are such that surveys of large contiguous areas are impossible. These two zones are therefore
generally unsuitable for survey, and the opportunistic study of available sections is the only low-budget
option open to researchers. The location of our survey transect across the transitional zone, consisting
mainly of the aforementioned marine and fluvial terraces of the Raganello river, was therefore largely
determined by the geo-morphological structure of the region. However, as will be made clear below, there
were other archaeological reasons to concentrate research in this zone as well.
The legend of the most recent detailed geological map of the area (CGC 1969) allows us to establish that
there are four main formations in the study transect. From lowest to highest these are:
1) alluvial deposits held in place by vegetation or artificial works; morphologically, these consist of a
single alluvial fan emanating from the Raganello valley;
2) yellow-red sands and pebbly sands associated with conglomerates consisting of well-cemented
rounded pebbles; this formation includes clayey intercalations and, locally, banks of yellow-green
clays and silts; morphologically the formation consists of three terrace levels. The formation is highly
permeable but can resist erosion well locally, depending on the degree of cementation of the deposits;
3) grey-blue clays, often intercalated with sand lenses, and conglomerate sands. This formation occurs
only in one area near Lauropoli, is easily eroded and is prone to slumping especially where sandy
layers abound; the permeability of the deposits is generally low; and
4) well-cemented coarsely stratified conglomerates consisting of large calcareous pebbles and rounded
sandstones, on occasion of pebbly or large-grained sands, locally with silty or clayey lenses. With a
generally high permeability, the local sensitivity to erosion of this formation depends on its degree of
cementation.
CLIMATE AND LAND USE / LAND COVER
A fairly detailed discussion of the current climate and land use of the Sibaritide can be found in D’Angelo
and Oräzie Vallino (1994:785-92), who note that the climate of the terraces and foothills is more pleasant
than that of the coastal plain, with more frequent rains because of the surrounding mountain, and less
oppressive heat in the summer and autumn because of the wind and altitude. The highland is even more
wet and cool, and is suitable for transhumance in summer. The larger part of the survey transect
nowadays is given over to arable land and a mixture of old and new olive groves, the latter giving rise to
extensive soil disturbance (deep ploughing, levelling) and cultivation of previously marginal zones
A recent manual auguring program in the plain, to depths of up to 7.5 m, has obtained absolute (radiocarbon) dates of up to
2100 BP for peaty layers, providing post quem dates for the more recent sedimentation phases (pers. Comm. Jan Delvigne).
2
12 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
(especially steep slopes). The arable is apparently used for grain crops and, in at least one recorded case,
maize for which a previous vineyard had recently been uprooted. Other lesser types of land use in the
survey zone include grassland, a rubbish dump, and small almond groves, but a considerable amount of
land especially in and around the dry gullies is still unused and covered in the local thorny macchia. The
prevalence of the latter increases with elevation, whilst the accessibility decreases. A marked contrast
exists with the land utilisation types mapped in the early 1950s (CNR 1956a,b); at that time, less intensive
forms of agriculture such as olive groves and grassland occurred over larger parts of the survey zone, at
the expense of the arable (macchia appears not to have been mapped).
3
RESEARCH HISTORY
Following the discovery of Sybaris and its Hellenistic and Roman successor towns through the combined
efforts of the University of Pennsylvania Museum and the Lerici foundation (Lerici 1960: 303-337, Rainey
& Lerici 1967, Rainey 1969: 261-273), an approximately 21 by 24 km (500 km2) area3 in the centre of the
region was surveyed by a team under the direction of the young Lorenzo Quilici, who is now a senior
professor at the University of Bologna. Quilici aimed both to provide a context for the excavations
starting at Sybaris in 1969 and to record surface archaeology in advance of land development schemes
funded by the Italian Cassa per il Mezzogiorno (De Rossi et al. 1969:147). In what follows we shall refer to
this survey, for brevity’s sake, as the ‘Quilici survey’, and to its results as the ‘Quilici data set’. Following
their desktop study of the records held by the Soprintendenza archeologica della Calabria, the study area
was divided into three zones, each ‘topographically’surveyed by one or two members of the team. In
addition to revisiting and evaluating the known sites, many new sites and monuments were discovered as
well. The results were recorded on a total of 858 forms and 23 sheets of the 1:10,000 map series
published by the ‘Cassa per il Mezzogiorno’, but were considerably condensed for publication. The
whereabouts of the original survey archive are currently unknown. The nearly 800 sites of archaeological
interest collected from documented evidence or recorded in the field by this team still provide the bulk of
the archaeological record for the region - later additions being of a piecemeal nature - and are therefore
the primary subject of the study reported here (see figure 2).
Additional work on the protohistoric settlement history of the Sibaritide was published by the excavator
of Broglio di Trebisacce, Renato Peroni, based on a series of surveys conducted since 1979 by the
University of Rome in conjunction with the Superintendency for Calabria (Bergonzi et al. 1982; Peroni
1994; Peroni & Trucco 1994)4. By contrast, very little primary research has been done on either the
classical and Hellenistic Greek period, or the subsequent late Republican and Imperial Roman period in
the Sibaritide outside of Sybaris itself; here we must rely on the general literature regarding Magna Graecia
and the Roman Empire. Since 1991, the Groningen Institute of Archaeology has been involved in
excavations at the protohistoric/Archaic hilltop sanctuary of Timpone della Motta and its settlement
(Attema et al. 1998). These excavations have given rise to largely unpublished preliminary and exploratory
surveys in the immediate surroundings of the Timpone Motta, extending on occasion to neighbouring
Monte Sellaro (Feiken & Weterings 1998) and further along the footslopes towards Broglio di Trebisacce
(Haagsma 1996).
In 1994, Haagsma and Delvigne conducted unsystematic field checks on settlement sites identified by the
Quilici survey between the Timpone Motta and Broglio (Haagsma & Delvigne, unpublished notes). These
The study area corresponds to the following (partial) sheets of the IGM 1:25,000 map series, 1958 edition: 221 II NE & SE, 222
III NO & SO, the eastern half of 221 II NO & SO, the north-eastern quadrant of 229 I NE, and just over the northern half of
229 I NO and 230 IV NO. This area is bounded by the following co-ordinates - North: 4410000; South: 4386000; East: 2650000;
West: 2628000.
3
Although this is not clearly stated in the published accounts, it appears that these surveys were largely limited to the fascia
collinara (foothill zone), and were intended to locate protohistoric settlements similar to those at Amendolara, Francavilla
Marittima, and Torre Mordillo.
4
12 -
4
F IELD W O R K
IN THE
S IBARITIDE
revealed that the lowest-lying of the marine terraces were already settled in the Middle Bronze Age while
the upper parts of the coastal plain were not; a Late Bronze Age site was found on the Timpa del
Castello, a rocky outcrop above the present village of Francavilla Marittima5. Their survey yielded no
identifiable Iron Age or Roman material, but did locate some of the 6th century BC Archaic impasto
Figure 2 - Distribution of settlements by period from the Quilici survey, after De Rossi et al.
1969, fig 3. Dithered areas: alluvial deposits. The rectangle indicates the area in which The
RPC survey took place.
The terminology used in this paper is the ‘early’one advocated by Peroni, in which protohistory includes the MBA to IA
periods, and history begins with the early Greek colonisations of the 7th and 6th centuries BC. A more conservative chronology
would have protohistory begin with the late Iron Age (8th century BC), and the historic period with the late Classical and
Hellenistic historians of the 5th and 4th c BC.
5
12 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
missing from the earlier Quilici survey results. Based on this work Haagsma (1996) suggested that Roman
intensive land use may have avoided the foothills and may have been concentrated in the coastal plain. In
1995, and again in 1998, several small field surveys were conducted by students of Kleibrink in settlement
and cemetery zones around the foot of the
Timpone Motta. The 1995 survey by Kleibrink and Waterbolk (reported in Haagsma 1996) found a lot of
material then interpreted as having been “washed off the Timpone”, but only one small Archaic site;
Haagsma and Attema also located a 6th century wall segment and kilns associated with a settlement on the
lowest plateau of the Timpone della Motta. A subsequent reassessment of the stored material from this
survey by Attema and De Haas resulted in the definition of several new sites, including the
Archaic/Classical extension to the Iron Age cemetery of Macchiabate belonging to the settlement on the
Timpone della Motta (De Haas 2001:18-19).
Despite covering only very small areas and few sites, these surveys provide significant additional
information in the form of evidence for periods 'missing' from the Quilici data set, such as the Archaic
(6th century BC), and for continuity and change in the processes of centralisation and acculturation
between the indigenous Oenotrians and the Greek colonists.
4
PATTERNS AND BIASE S
The interpretation of regional archaeological records in terms of a history of settlement and landscape
must be informed by an assessment of the biases that might conceal or produce patterning in those
records. A cursory look at the geological map of the Sibaritide shows that alluviation of the coastal plain
must have been a factor of major significance, resulting in an almost total lack of find spots below a quota
of 25 m asl and in the alluvial plains of the Crati and Coscile rivers extending into the hinterland (dithered
areas in Figure 2). Other factors which previous studies have suggested to be a priori causes of significant
biases in the archaeological record are: the land use and land cover (abbreviated in what follows to
‘LULC’) at the time of the survey, the accessibility given the contemporary infrastructure, and the
selectivity inherent in the methodology employed for the survey.
Accordingly, the main aim of the SIBA’00 survey campaign has been to assess the influence these factors
had on the quality of the extant archaeological record, that is, primarily of the Quilici data set. Since we
were unsuccessful in obtaining the necessary information either from the archives of the Soprintendenza
archeologica of Calabria or from Lorenzo Quilici himself, this assessment had to be obtained through a
combination of targeted desktop and field studies.
In order to assess the influence of land use / land cover around the time the Quilici survey took place, a
land utilisation map depicting the situation of the early 1950’s at a scale of 1:200,000 was digitised and
georeferenced (CNR 1956a,b)6. This map shows that the Sibaritide plain was dominated at that time by
arable (seminativo asciutto) but that a fair amount of grazing land (pascolo ed incolto prodottivo) was also present;
olive groves and arable dominated the foothills (to which fruit- and vineyards may be added especially in
the area around the town of Cassano allo Ionio); while the higher slopes of the Pollino were mostly down
to wood- and grazing land.
The frequency of Quilici sites with regard to land use was tested for deviations from randomness, using
the ? 2 - test after removing all alluvial areas below 25m ASL. Table 1a lists the preliminary results of this
Georeferencing is the procedure whereby digital map data are tied to an area of the Earth’s surface by specifying the map
projection and co-ordinate system. Here, the Gauss-Boaga projection and the Italian national co-ordinate grid according to
European Datum 1940 are being used.
6
12 -
6
F IELD W O R K
IN THE
S IBARITIDE
test, which indicate that there is a strong correlation between land use and site density7. As we can see, the
frequency of sites in woodland (cats 13 and 14 combined) is low with only 9 observed sites where 22
would have been expected given that 3.3 percent of the study area is covered in woodland. Similarly, 52
observed sites in mixed vineyard-olive grove (cat 9) against an expected number of 24 represents a
significant deviation from randomness. Of the land use types that occur most frequently, dry arable with
trees (cat 2) is slightly favoured with 56 sites against an expected 39, while productive grass- and
uncultivated lands (cat 19) is slightly avoided with 75 sites against an expected 103. However, in order to
satisfy all of the requirements for this test, some land use types must be grouped together so that the
expected number of sites per category is less than 5 in less than 20% of categories (see Figure 3).
Site Characteristics
(0)
(1)
(2)
(3)
(4)
(7)
(8)
(9)
(10)
(11)
(13)
(14)
(19)
(20)
(21)
(99)
no data
Seminativo (asciutto)
Seminativo arborato (asciutto)
Seminativo irriguo
Seminativo arborato irriguo
Vigneto
Uliveto
Vigneto-Uliveto
Agrumeto
Frutteto (frutta polposa)
Bosco ceduo
Bosco d'alto fusto
Pascolo ed incolto produttivo
Sterile
Insediamenti ed altre forme
Acque
Totals
Site Characteristics
(0)
(1 & 2)
(3 & 4)
(7)
(8)
(9)
(10 & 11)
(13 & 14)
(19)
(21)
no data
Seminativo asciutto
Seminativo irriguo
Vigneto
Uliveto
Vigneto-Uliveto
Agrumeto / Frutteto
Bosco
Pascolo ed incolto produttivo
Insediamenti ed altre forme
Totals
% cover
Expected
sites
46.4
5.9
0.6
0.5
2.1
20.5
3.5
0.3
0.1
1.1
2.2
15.5
0.2
0.8
0.4
100.0
309.1
39.1
4.2
3.3
13.8
136.8
23.6
2.0
0.5
7.2
14.7
102.9
1.2
5.2
2.4
666.0
% cover
Expected
sites
52.6
1.1
2.1
20.7
3.6
0.4
3.3
15.5
0.8
100.0
350.1
7.6
13.9
137.5
23.7
2.5
22.1
103.5
5.2
666.0
Actual
sites
85
325
56
4
0
12
122
52
2
0
1
8
75
0
9
0
666
Chi
square
Actual
sites
85
381
4
12
122
52
2
9
75
9
666
Chi
square
0.817
7.348
0.010
3.318
0.231
1.597
34.225
0.000
0.494
5.387
3.042
7.568
1.240
2.854
2.449
70.579
2.727
1.683
0.250
1.756
33.721
0.094
7.727
7.838
2.795
58.591
Table 1: Results of a ?2 test of 666 Quilici sites against 1956 land utilisation types. 751 sites in
region of 350 km2. A) Raw results, B) results after reclassification
The results of a second ? 2 - test (Table 1b) confirm that land use and site location correlate very
significantly (far above the 0.1% level of significance). However, they also show that none of the three
most frequent land use types (dry arable, olive groves, and rough grazing lands) has had a dramatically
positive or negative influence on the number of sites discovered. Interpretation of these results is not
straightforward, not just because land use at the time of the Quilici survey may have been significantly
different from the land use that was mapped in 1956, but also because the clustered nature of
7
The total ? 2 of 71 at 14 degrees of freedom is significant at the .001 level.
12 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
archaeological site records makes them notoriously subject to spatial autocorrelation effects8. As an
example of the former, it seems a priori unlikely that the Quilici team could have observed even as many
Figure 3 - Distribution of Quilici sites across 1956 land utilization types. For Legend, see table
1B. Land below 25 m asl is indicated by dashed line.
Spatial autocorrelation refers to the fact that observations made in close proximity to each other tend to have similar
geographical attributes; thus land use observations made at two archaeological sites separated by only 100m have a much higher
chance of turning out to be identical than if the sites were separated by a distance of 1000m.
8
12 -
8
F IELD W O R K
IN THE
S IBARITIDE
as 75 sites under the very adverse visibility conditions typical of grass- and uncultivated land (see Table
1B, cat 19). And indeed it may not be coincidental that the Quilici sites falling within the SIBA’00 survey
transect (see section 7 below) appear to cluster within one such zone of grass- and uncultivated land.
Perhaps these areas had already been brought under the plough by the time the Quilici team surveyed the
area in the late 1960s. Access to the original survey records will play a very important role in assessing the
significance of the land use changes that may have taken place in the 1960s.
Quilici himself adduced an example of the latter effect, interpreting one dense cluster of sites located near
the town of Eianina as part of a large roadside settlement within the territory of the Roman colony of
Interamnium (De Rossi et al. 1969:9-10)9. Figure 3 shows the land use around the cluster to be the
otherwise rare category of mixed vine- and olive yards (Table 1B, cat 9). In combination with the fact that
neither vineyards nor olive groves in monoculture (cats 7 and 8) give rise to anything like the observed
high frequency of archaeological site locations in the mixed unit, the high frequency must be caused by
the spatial clustering of the sites rather than by particularly favourable land use at the time of the survey whatever that may turn out to have been.
The distinct linear clustering of sites, which Quilici himself interpreted as an indication of the existence of
‘villages’along proto- and early historic routes, could also be caused by biases in the Quilici survey itself.
With a survey universe (excluding the coastal plain) of some 350 km2 it is clear that the survey team could
not possibly have systematically covered everything. It would instead have had to focus on those areas
that were most likely (from previous finds) to contain sites and which could be reached within a
reasonable amount of time by road or track. Figure 4 again shows the Quilici data set, this time against
the background of a relief map, and in relation to towns and metalled and semi-metalled roads digitised
off the IGM 25V map series. The proximity of sites to infrastructure seems fairly straightforward in some
areas, especially if we remember that minor tracks were not digitised.
These preliminary studies therefore indicate that the potential of further bias modelling for the Quilici
data set must be explored when the original research archive has been located. The second method for
assessing biases in the Quilici data set is through field studies designed to detect archaeological materials
present in areas, or from periods, that Quilici and his team might have ignored or been unable to detect.
These studies are the subject of the next section.
5
THE SURVEY: APPROA C H E S
The approach taken by us was to select a representative transect through the transitional zone between
the coastal plain and its mountainous hinterland, consisting of marine and fluvial terraces (the 1st and 2nd
geological formations discussed above), and to see how the recorded sizes, locations, interpretations and
dating of the sites mapped by Quilici and his team compare with the current archaeological surface
record. The transect was approximately 6 km long and 1.5 km wide and located between the valley of the
Raganello river and the modern town of Lauropoli (figure 5). The 2nd geological formation, which is
morphologically divided into four distinct levels or ‘terraces’, is today used for the most intensive forms
of agricultural exploitation, consisting of mostly arable fields but with large sections given over to olive
culture; due to the extremely restricted surface visibility in the other three units, our survey took place
almost entirely within this unit. Archaeological sites mapped by Quilici and his team are also exclusive to
this unit.
Quilici identified the modern town of Castrovillari, on the upper reaches of the Coscile river, with Interamnium; but the
distances, of 13 and 8 miles respectively, provided by the Antonine Itineraries and the Tabula Peutingeriana for the stretches of
the Via Popilia between Muranum (modern Murano), Interamnium, and Taurasia (modern Tarsia) make it unlikely that this
identification is correct. They instead indicate a location in the middle Coscile valley.
9
12 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 4 - Association of the Quilici data set (dots) with modern accessibility circa 1956 (paved
roads only). 25m Elevation contour indicated by dashed line.
The SIBA’00 survey was conducted using two teams working independently and studying the same
general area at two different spatial scales. The first team used methods developed in recent years for a
high-intensity survey of all available fields within a 5 by 1 km transect running parallel to the Raganello
river, and covering all four main geomorphologic units (fluvio-marine terraces). In this survey, agricultural
fields were subdivided by pacing into units approximately 50 by 50 m (0.25 hectares) in size, and these
were walked at 10 m intervals, with all manmade materials collected except those that were of obviously
recent date. Further samples were collected if circumstances (such as an increase in finds density)
warranted. In order to assess the effect of differential visibility on the recovery rates, five factors affecting
the recovery of archaeological materials were recorded separately for each collection unit, on a scale of 1
to 5: stoniness, shadiness, vegetation cover, tillage/dust, and amount of recent material on the surface; an
independent estimate of the total visibility, again on a scale of 1 to 5, was also made.
12 -
10
F IELD W O R K
IN THE
S IBARITIDE
Figure 5 - The SIBA2000 fieldwork transect (excluding Monte S Nicola), with extensive
(black) and intensive (white) survey areas. Plus signs: Quilici sites; circles: RPC sites and
scatters.
The second team conducted a much more extensive and site-oriented survey in an area directly
neighbouring the intensive survey transect, and overlapping it in several parts in order to provide material
for comparing the two. The main aim of this survey was to check whether the broad settlement patterns
mapped by Quilici and his team were upheld, by locating and describing any ‘sites’occurring in the area; it
therefore used agricultural fields as its collection units, and a walker interval of 15m. Once a site had been
identified by this team, a more detailed survey was made at 5m intervals in order to map finds density
contours and make diagnostic collections.
Only two forms were used during the survey: a Unit Form to record properties of the collection unit
including bias factors and number, type, and summary contents of the samples (termed ‘Bags’) collected;
and a Bag Form to record the contents of each of these samples. Unique Bag ID’s are formed by subnumbers from unique Unit ID’s; a water proof ticket with the Bag ID was kept with the finds throughout
the processing. This system, developed out of previous administrative systems used at the GIA and
AIVU, is described in more detail elsewhere10. Mapping of spatial collection units was done using a
combination of independent single-receiver GPS measurements and 1:10,000 scaled field maps; again, a
more detailed discussion is forthcoming (Ryan & Van Leusen in press).
10
Van Leusen forthcoming, chapter 8.
12 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
6
FINDS PROCESSING
Rather than making any decisions about the significance of the finds in the field, the intensive survey
team simply collected the finds for processing and interpretation at the survey base. For the purpose of
tracking individual finds bags, a simple system of check-boxes was used on the field forms for each
collection unit. Processing at the survey base began with washing and drying, after which all finds were
classified by Peter Attema according to the system laid out in table 2. Both survey teams together
collected some 210 kg of material (some 6000 finds) from 610 collection units, of which 70 kg were
classified during processing as largely recent or sub-recent tile/brick (subsequently discarded). Table 2
provides the breakdown of these totals by material category. The column headed ‘%arch’gives the
relative weights per category when considering only pre-modern ceramics (totalling 129.2 kg).
Table 2 - Material categories used in summary classifications of the SIBA’00 survey finds, with
preliminary total counts, weights, and percentage by weight of pre-modern ceramics.
Cat
1
1a
1b
1d
2
2a
2b
2c
3
3a
3b
4
4a
4b
4c
5
5a
5b
5c
6
6a
6b
7
8
9
10
13
14
15
16
Description
tile
coarse tile
depurated antique tile
(sub-) recent tile
coarse wares
coarse thick ware (pithos)
coarse medium thick ware
coarse thin ware
Depurated wares
depurated orange ware
depurated pale ware
Fine wares
black gloss / banded ware
terra sigillata
thin painted ware
Indeterminate wares
indeterminate coarse ware
indeterminate depurated ware
indeterminate (sub-) recent ware
Combed wares
hard orange combed ware
hard pale combed ware
Other wares and materials
glazed wares
hard red ware, medium thick
lithics
impasto wares
grumi
glass
waster
slag
totals
Count
230
144
84
2
1208
175
942
91
2134
457
1677
70
62
6
2
2102
300
396
1406
10
2
8
Weight (g)
36300
22955
13220
125
45321
27970
16820
531
23335
4783
18552
718
638
75
5
91712
8532
13567
69613
140
35
105
11
32
10
42
72
5
8
19
5936
225
485
75
730
2625
55
2760
6215
210696
%arch
17.8
10.2
21.6
13.0
0.4
3.7
14.4
0.5
0.1
0.0
6.6
10.5
0.1
0.2
0.4
0.6
100.1
SYSTEM OF CLASSIFICATION
The main aim of the classification of ceramic wares used in the SIBA’00 survey was to assign dates and
functions to the material found. The classification is based on the macroscopic properties of the ceramics
– fabric, inclusions, colour, thickness and morphological characteristics. Since most of the surface
12 -
12
F IELD W O R K
IN THE
S IBARITIDE
material is very fragmented and abraded, a primary division into three clear-cut ware categories was used:
1. impasto, which is the handmade pottery characteristic of the protohistoric periods. It is irregularly fired
and often has a burnished surface;
2. coarse wares, which are turned or finished on a wheel and characteristic of the Archaic to Roman
periods. They are generally orange firing wares made of a clay with fine to coarse inclusions;
3. depurated wares, which are wheel-turned and either pale- or orange firing and have either a hard or a
soft powdery surface. Depending on the clay base and firing characteristics, these wares can be
classified into various periods from the late Iron Age to sub-recent times.
Classifications within this primary division are based on characteristics relating to function (e.g. category 1
“tile” is a functional category; category 2a “coarse thick ware” is functionally interpreted by the addition
‘pithos’). Sherds belonging to these broad functional classes are fairly easily identified even in a muchabraded state. The association of certain types of tiles, pithoi and datable pottery sherds such as black
gloss and terra sigillata may then indicate the existence of a farmstead at a particular time and place.
Categories 10, 13: Impasto wares and ‘grumi’
Impasto wares are hand-formed, with coarse temper and varying types of surface finish. In principle they
can be classified by form, thickness, finish and decorations into various wares, but the SIBA’00 material
allows only a simple division to be made between the well-burnished early to middle Iron Age ware found
near Monte San Nicola and the undiagnostic and often abraded sherds of late Iron Age ware found
elsewhere. Grumi (burnt daub) appears as orange-red low-fired crumbs in ploughed fields and by itself
cannot be dated; in association with impasto or dated archaic wares it may be assigned to one of these
periods.
Categories 1a, 2a-c, 5a: Coarse wares
These are coarse wheel-made wares of red firing fabrics containing visible temper, production of which
began in the Archaic period and continued through the Hellenistic period. Wares can be classified by
form and thickness into tiles, storage, kitchen, and table wares. A further form-based subdivision of
coarse wares is envisaged, but must await closer study of the material (cf. Attema et al. 2000, Appendix 1)
and comparison with collections from regional excavated contexts11. In total, coarse material (cats 1a, 2ac) makes up almost 53% by weight, and the Hellenistic to Roman material (cats 1b, 3, 4, 5a-b) some 46%.
Wheel-turned coarse wares lacking “archaic” characteristics of colour, temper and surface finish may date
from the Hellenistic or Roman periods, but cannot be dated independently.
Categories 1b, 3, 5b: Depurated wares
Depurated tiles and wheel-turned pottery wares are made out of levigated clays, to which temper may be
added to achieve specific properties. These wares range in colour from orange to pale buff; different
wares and forms may be distinguished on the basis of form and thickness. Roof tiles, which usually have
added temper, tend to be preserved in relatively large fragments and can be distinguished by overall shape
and rim profile12. The remaining depurated material mostly consists of orange or pale firing amphora
fragments, and a small amount of ‘fine’wares which can easily be recognised by their surface treatment,
fabric, form, thickness, and decoration. All of these also occur in severely fragmented and/or abraded
form and, being undiagnostic, are then assigned to category 5b.
11
Subsequent closer inspection and cataloguing of the diagnostic material by scatter during the campaign of September/October
2001 has confirmed the existence of sites with an origin in the (later) archaic period, supporting an early date for the associated
coarse wares.
12
These rim profiles will be the subject of a closer study in the near future.
12 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
Categories 4 and 6: Fine and Combed wares
The fine wares encountered during the survey consist for the most part of a pale or orange firing
depurated clay with a glossy surface decoration of black slip, either (archaic/classical) banded ware or
(Hellenistic) black gloss ware. A majority of the sherds had, however, lost most or all of its surface finish;
if these could not be identified as a fine ware by form (e.g. “kylix”) they were classified as category 5b
“indeterminate depurated ware”. Other fine wares occurring in very small numbers are terra sigillata and
an unidentified thin painted ware. Only 10 sherds with a ‘combed’decoration were found during the
survey. The material, which is hard and has either an orange or a pale colour, suggest a late (possibly postantique) date but no parallels have been found yet. It is expected that further study of categories 4 and 6
will result in a finer typo-chronological division of the material.
Category 9: Lithics
No lithics were previously known from the study area, and indeed the SIBA’00 survey collected only a
very few, most of which were judged to be probably of natural origin in view of their irregular negatives
and retouches13. Only three are clearly intentional flakes, one of which is a large white patinated flake
whose distal end was intentionally retouched into a scraper. Although no firm date can be given, the
Levallois-like technique employed suggests an attribution to the Middle Palaeolithic (however, the piece is
atypical). The other two flakes cannot be dated.
7
RESULTS
THE INTENSIVE SURVEY
The intensive survey teams covered some 125 hectares (552 collection units) spread over the four marinefluvial ‘terraces’within the survey transect in a total of 121 person/days (inclusive of administration, GPS
measurements, etc), for an average of 1 hectares per person/day. This is somewhat slower than the
speeds achieved during similar surveys elsewhere (cf. Attema et al. 2001, in press) and was mainly due to
the extra time needed to cope with mapping problems. Table 3 and Figure 5 give an overview of the
fields covered by the intensive survey.
In order to produce distribution maps by period, the raw counts per material category per collection unit
were corrected for unit area, percent coverage, and visibility, using the formula
D = N * (100 / C) * (100 / V) / A
where D = corrected finds density, N = raw finds count, C = estimated percent coverage, V = estimated
percent visibility, and A = Unit area in hectares. It is recognised that this procedure, while producing
normalised finds densities expressed in numbers per hectare, tends to exaggerate small density variations
where low numbers of finds are involved; appropriate care must therefore be taken when interpreting the
distribution maps reproduced here in figure 6.
The distribution of protohistoric material shown in Figure 6A is characterised by small amounts of
material in all surveyed fields, generally associated spatially with the major terrace edges. The material
tends to consist of undiagnostic impasto pottery, except on the higher terraces near the Monte San Nicola
where denser concentrations of more identifiable burnished Bronze Age and Iron Age wares occur. Here
the presence of a large protohistoric site (Peroni & Trucco 1994:819-20 site 31) was confirmed.
Table 3 – Preliminary results of the intensive survey, by field
The nearest Neolithic site is at Favella della Corte, south of the Crati river. The lithics collected by the SIBA2000 survey were
studied in April 2001 by Marcel Niekus of ARC (Centrum voor Archeologische Research & Consultancy).
13
12 -
14
F IELD W O R K
Units
1001 –
1031
Toponymic
Azienda La
Silva West
1032
1058
–
Azienda La
Silva North
1059
1087
1088
1117
1118
1142
1143
1208
2001
2036
–
-
–
Aloisi
Olive grove
Sferracavallo
-
2037
2055
2055
2064
2065
2089
2090
2105
2106
2125
–
2126
2148
–
2149
2161
2162
2177
–
2178
2184
2185
2234
–
2236
2268
2270
2275
2276
2289
2290
2330
–
2331
2344
–
14
–
–
–
Aloisi
North
Rubbish
dump
Elongated ploughed field with unploughed sections; contains a very diffuse scatter of coarse and
thin wares in the topmost units (scatter 17, units 2079-81 and 2084-86).
Ploughed field along Raganello valley; no remarks.
–
Irregular ploughed field along Raganello valley with adjoining units along lower terrace slopes;
found Quilici site 135 (‘scatter of Hellenistic-Roman material, including some large sherds’; is
scatter 18, units 2106-8 and 2112-5), some impasto sherds, and off-site material provenient from
the next higher terrace.
Group of irregular ploughed fields along lower edge of terrace; three independent grids were used
(for units 2126-35, 2136-7, and 2138-48); finds were concentrated along the north-eastern edge
facing the Raganello valley (units 2126-7, 2126/8, 2136), but no distinct concentrations could be
defined.
Irregular ploughed field; no remarks.
–
–
Notes
An old olive grove with stands of oak, cut along the slope by a wide concrete-lined V-shaped ditch
and a U-shaped aqueduct (both probably dating to the 1930s). Quilici site 62 (‘scatter of
14
Hellenistic-Roman sherds, including BG ’) was found again but appears ‘smeared out’ across a
large part of the field; two out of several finds concentrations are sufficiently distinct for
individuation (scatter 20 in units 1020-23 and 1030-31, scatter 21 in units 1001-04 and 1009-11).
Quilici site 63 (‘scatter of Hellenistic-Roman material, including grey ware’), supposedly present at
the western edge of this field, could not be found despite specific searching.
Ploughed field. Part of Quilici site 62 was found nearest the Azienda buildings (scatter 20, unit
1045); the rest of field contains one other major (scatter 8, units 1048-9) and two minor finds
concentrations (scatter 11, unit 1052; scatter 12, unit 1050). Concentration of recent material in
unit 1037.
Ploughed field; one small site was located (scatter 1, unit 1060)
Large ploughed field just east of Lauropoli; steep slopes into valley on two sides; contains four
sites (scatter 9, unit 1168; scatter 7, units 1186-7; scatter 5, unit 1202; scatter 6, unit 1207).
2001 – 2022 irregular ploughed field with evidence of deep ploughing; mapped two concentrations
of thin wares (scatter 15, unit 2001; scatter 16, units 2005-6), the latter continuing across a recent
path into units 2025-7, and further disturbed by works relating to the driveway and garden of the
house belonging to Dr Aloisi. 2023 – 2036 were surveyed in a young olive grove before harrowing.
Irregular ploughed field; units 2038-9 contains a finds concentration (scatter 19) which might be a
kiln but also contains recent material; 2049 contains a sub-recent tile kiln.
Irregular ploughed field; no remarks.
–
–
S IBARITIDE
Young olive grove, just harrowed; used concrete posts to set out grid; units 1108-9, 1088-9, 10947, 1106-7 and 1113-4 were surveyed by team 2 before harrowing; no sites present.
Irregular field; found one small sherd concentration with impasto (scatter 10, unit 1133).
–
–
IN THE
Ploughed field, sloping steeply toward the Raganello valley, plus units adjoining path along top of
the slope; found only a sub-recent finds concentration approximately where Quilici site 134 (‘diffuse
scatter of Hellenistic-Roman material’) should have been (unit 2176).
Some small ploughed fields on slope facing the Raganello valley, with unploughed section in
between; no remarks.
Irregular ploughed fields with steep terrace and gully slopes on their eastern side; no remarks.
Rebecco
(Masseria
Frascino)
Four ploughed fields on steep slope facing the Raganello valley; located a possibly Bronze Age /
Iron Age ‘grumi’site on fairly level bottom section of slope (scatter 13, units 2263-4 and 2268).
Part of a ploughed field, surveyed rather too quickly because of failing light (= low quality). No finds
concentrations.
Two irregular ploughed fields on terrace along Raganello valley; no remarks.
-
–
–
Monte
Nicola
S
Three irregular ploughed fields on terrace along Raganello valley; steep slopes on southern edges;
located finds concentrations in two fields (scatter 4, unit 2306; scatter 14, units 2312-3 and 23167).
Three small irregular ploughed fields just below the unnamed hilltop on which stands the radio
tower; containing one previously unknown Hellenistic site (scatter 2, unit 2344). Found part of
Peroni Bronze Age / Iron Age site 31 on the saddle (scatter 3, units 2332 and 2334) and on the
south-east slope of Monte S Nicola (grab sample).
BG = Black Glaze pottery (also termed Black Gloss, the black decorative surface coating is in fact a clay slip).
12 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
Coarse (unburnished) protohistoric impasto fabrics are very difficult to find in arable fields under the
conditions prevalent in the Sibaritide foothills. Their recovery was made extremely difficult because most
fields were full of bits of ploughed-up conglomerate rock of a colour and shape often mimicking
potsherds. It is noteworthy that this material tended to be found only when surveyors were sitting down
or were looking especially for it; we can therefore not be certain that the map shown in figure 6A is an
accurate reflection of the distribution of this material.
For both the coarse (figure 6B) and the Hellenistic-Roman material (figure 6C), the corrected finds
distributions show a quite intensive use of all terrace edges (including minor ‘internal’ones), except
perhaps those that are very small or inaccessible. The two distribution patterns themselves are very
similar, any differences being quantitative rather than qualitative: no concentrations of coarse wares occur
away from concentrations of Hellenistic-Roman material. It therefore remains to be determined whether
the coarse wares should be interpreted as evidence for an early (6th – 4th century BC) settlement expansion
followed by a Hellenistic phase of continuity, or as just another category of Hellenistic wares. In the latter
case, little direct evidence for Archaic-Classical rural settlement remains, and the first substantial rural
expansion in the foothills could have occurred as late as the late fourth – third century BC. Among the
fine wares, one small group of ‘banded ware’kylix fragments has already been tentatively assigned to the
Archaic period on the basis of comparable material present in the museum at Amendolara, but this needs
further study.
It may be noted further that, contra Quilici, the Hellenistic to Roman material in the survey transect
consistently points to dates in the range 325 BC – AD 100, and only rarely into the later Imperial period.
This contradicts received wisdom regarding the historic settlement dynamics in Calabria, referred to in
section 9.
THE EXTENSIVE SURVEY
The extensive survey team covered some 50 agricultural fields (315 hectares) in an estimated 38
person/days, for an average of 8.3 hectares per person/day. Table 4 lists and describes the sites located
by this team and the sites mapped by Quilici in the same area.
12 -
16
F IELD W O R K
IN THE
S IBARITIDE
Figure 6 - Corrected finds density distributions resulting from the intensive survey. Grid spacing:
1 km. A: protohistoric period (cats 10, 13). B: “Archaic” material (cats 1a, 2a-c). C:
Hellenistic to Roman material (cats 1b, 3, 4, 5a-b).
12 -
17
V A N L EUSEN : P A T T E R N T O P ROCESS
Table 4 – Site list for the extensive survey.
RPC
ID
3002
3003
Quil
ID
126?
Max
2
N/m
50
30
5x5
5x4
3008
127
10
10 x 15
3016
125
35
50 x 30
3018
128
35
50 x 75
3025
New
5
15 x 15
3026
131?
-
75 x 125
3029
3032
New
New
2
25
2x2
30 x 20
3039
New
5 - 10
10 x 20
3040
New
-
10 x 10
3041
New
-
40 x 40
3042
New
-
20 x 10
3047
new
5 - 10
10 x 20
3048
130
-
90 x 30
3055
2-5
40 x 20
3056
-
10 x 10
15
TS = Terra Sigillata.
12 -
18
Area (m)
Notes
These two concentrations possibly constitute a new site, but are more probably related
to Quilici 126 (a ‘Hellenistic-Roman scatter containing tile frs’). They contain both
“archaic” coarse ware and BG ring bases, pithos sherds, and tiles, including one half of
a terracotta mould for votive feet. Site ‘halo’ extends into fields 3004/3010, including
Hellenistic imbrex, but mostly consisting of tiles with many quartz inclusions. Nearby in
this same field is a large dump of stones, presumably ploughed up locally.
Site contains “archaic” coarse tile and pottery, Hellenistic and Republican roof tile, and
th rd
4 -3 c BC pithos. Separate from this in field 3007, but concentrated along the modern
th
road bed, is some off-site material: rim and handle of a late 4 c BC skyphos, and a
loom weight. Logged by Quilici as a ‘diffuse scatter of Hellenistic-Roman material’.
Scatter of coarse “archaic” ware and Hellenistic material located on terrace edge
overlooking the Franceschiello valley, partly obscured by modern path and rubbish
rd
nd
dump: BG pottery, pithos, foot of 3 c BC kantharos, early 2 c plate. Site ‘halo’,
extending over 100m into field 3015, contains some concentrations of cobbles:
th
Hellenistic-Roman material, material of late 4 c BC BG cup and Corinthian B
nd
rd
15
amphora, 2 –3 c AD ‘African’ TS , and a coarse ware loom weight. Logged by
Quilici as a ‘large dense Hellenistic-Roman scatter including BG’.
Large scatter containing tile, storage pots, coarse and fine wares. Located on an
elevated portion of the terrain cut by a modern field boundary. It is very homogeneous
st
in fine wares and contains a/o several 1 c AD Roman TS frs (CFTS form 21.3), some
with plastic decoration, and a column fragment. Diffuse material occurs off-site in fields
3017 and 3019; dump of stones, tile and dolium frs in eastern tip of 3019. Logged by
Quilici as a ‘Hellenistic-Roman scatter and villa with floor in opus spicatum (information
from farmers)’.
Scatter of Hellenistic material, dolium, BG, loom weight, all material of similar fabric;
also a number of river cobbles present.
Large area of diffuse finds in northern corner of field; mostly amphora, but also
th
rd
containing 4 – 3 c BC Corinthian material, and one loom weight. Note: not
intensively walked (perhaps off-site). Identification with Quilici 131 (a ‘diffuse
Hellenistic-Roman scatter containing an amphora handle’), which could not be found in
the neighbouring field 3027, rests on the assumption that the latter may have been
inaccurately mapped; the core of site 3026 is probably located in the neighbouring field
to the north (not surveyed).
Small scatter of Hellenistic tile, coarse and depurated wares.
nd
Vitrified kiln material, wasters, a 2 c BC amphora spike and neck of a jar in pasta
nd
st
grigia, and 2 – 1 c BC Roman roof tiles.
Scatter located on edge of terrace overlooking the Franceschiello valley. Material of
“archaic”coarse ware, Hellenistic cooking pots, Roman roof tiles, possibly TS.
Site contains a great many river cobbles; coarse “archaic” tiles and thick-walled pottery
present but not diagnostic, hence no clear dating evidence. May be related to site 3041
nearby.
Site is strategically placed, with a good view of the terrace. Contains Hellenistic tile, BG
ceramics, Corinthian A/B, dolium, and recent fabrics. Note: visibility low because of
bushes.
Very diffuse scatter of shapeless impasto sherds on terrace edge overlooking the
Franceschiello valley; this material occurs lower down the valley slope in field 3043 as
well.
Scatter of possibly Hellenistic material, contains no clear diagnostics. Depurated
orange firing roof tiles with some inclusions; various medium and thin coarse wares; a
few depurated orange firing sherds with some inclusions.
Sizeable scatter of Hellenistic material found on both sides of the modern road bed:
coarse “archaic” roof tiles and medium thick wares, 1 x BG, one coarse loom weight.
Note: very low visibility because of olive trees. Logged by Quilici as a ‘diffuse
Hellenistic-Roman scatter’.
Scatter of Hellenistic material collected after intensive survey (= scatter 6): coarse tiles,
th
medium thick and thin wares. Site contains relatively many luxury items: late 4 c BC
Corinthian A/B amphora frs, BG including foot of cup, and other depurated pale and
orange firing sherds of table wares.
Low-density scatter of coarse material, probably including LIA impasto, collected after
intensive survey (= probably scatter 5). Coarse tiles, medium and thin wares, including
one rim with comb motives on the lip.
F IELD W O R K
IN THE
S IBARITIDE
The extensive survey recorded a small concentration of impasto material on site 3042 (extending into
field 3043), and another one of coarse impasto-like material in field 3056. The large majority of the
material and sites were dated to the Hellenistic period. Sites 3002/3003 appeared to consist of recently
ploughed-up high-quality material, possibly belonging with Quilici site 126 (located some 100 m upslope
but of which almost no remains could be found); confirmed 4th-2nd c BC sites include 3008, 3016, 3026,
3041, and 3055; generic Hellenistic material occurs at sites 3025, 3029, 3039, 3047 and 3048; early Roman
(2nd-1st century BC) at sites 3032, 3039; and Roman Imperial wares (1st-3rd century AD) are found at sites
3016 and 3018.
8
S E T TLEMENT AND INFRASTR UCTURE
Whilst the results of the SIBA’00 survey should first and foremost be compared to those of the Quilici
survey which they were intended to test (De Rossi et al. 1969:147-155), the general lack of rural
settlement data pertaining to the coastal and alluvial plains of the Sibaritide means that we must also view
them in the less specific geographical and historical context of Roman Calabria, as provided by Accardo’s
recent study (2000). It bears repetition that most of the surface material found north of the Crati-Coscile
was ascribed by Quilici either to the late Hellenistic (late Republican) period or to the Roman Imperial
period. Even sites and scatters not expressly dated by him were to be ascribed to these three centuries
(150 BC – AD 150). Most of these appeared to represent small farms continuing into the Imperial period,
and tending to cluster into ‘villages’located on the lowest parts of the foothills; there were also a few
small villages lying along the water-rich valleys south-west of Cassano allo Ionio, and a few widely
scattered larger farms (fattorie padronali) which may be recognised by architectural remains. The territory as
a whole was therefore littered with settlements (villas and villages) in the Roman period, clusters of which
‘define a finely branched network of routes’, until at least the second century AD. However, the material
itself was described by Quilici as ‘quite unpretentious, even poor, indicating local production and use’recalling Strabo’s (VI, 253) judgement of Sybaris in his time as ‘barbaric’.
At the regional scale, Quilici contrasted the pattern of autarchic ‘villages’to the north of the Crati-Coscile
with that of single farms to the south and west, and wondered if this might not also express a difference
of social, even ethnic, units rather than being purely a consequence of the differing geomorphologic
structures of these areas. He suggested (1969:149) that the difference may be related to the presence of
‘colonial’landscapes of single villas emanating from the Roman colonies of Interamnium (Eianina and the
upper Coscile) and Copia Thurii (the hills south of the Coscile), while the area of larger villages in the
foothills north of the Coscile represents a survival from the pre-Roman Hellenistic society. If true, the
Hellenistic settlement pattern recovered by the SIBA’00 survey should consist of sporadic villages along
major thoroughfares. The location of roads connecting the coast to inland regions would have been
dictated by the major valleys and passes in the uplands and mountains (1969:151); a system of fortlets and
watchtowers safeguarded Hellenistic Thurioi and may have done the same for Archaic Sybaris, but so far
no evidence for this has been found.
Although actual finds predating the late Hellenistic period were rare at the time of his survey, Quilici also
attempted to reconstruct protohistoric patterns of settlement and, especially, infrastructure (viabilità) in
the Sibaritide (1969:97, 152-5). Several of his routes either cross or come near to the SIBA’00 survey area.
Using an 18th century parallel, he postulates the existence in the early Iron Age of a ‘coastal’route
following the foothills, connecting the large protohistoric centres, and avoiding the (marshy) plain16. A
second route is postulated parallel to this along the lower lip of the lowest ‘terrace’; a third route follows
the ‘radial’ morphology of the terrain and the line of a known Roman road from Castrovillari to
Frascineto and further on to Civita and, presumably, Thurioi/Copia in the plain; a fourth route
The use of historic parallels in the reconstruction of prehistoric routes is justified on the assumption these routes are the ‘most
natural’and by the supposed conservatism of their use; compare Small et al. 1998:338 for a similar argument concerning
transhumance routes in Apulia.
16
12 -
19
V A N L EUSEN : P A T T E R N T O P ROCESS
connecting Torre Mordillo, Cassano, and Civita forms part of a postulated protohistoric long-distance
‘mountain route’. Quilici derived this protohistoric infrastructure from the locations of known sites and
the terrain morphology, interpreting rivers as barriers, ridges as routes, and valley floors as difficult terrain
because of water and vegetation. The peaks and gorges are orientation points, sulphurous springs meeting
places. The formation of routes under ‘free’conditions, he maintains, would have been similar to that of
the Middle Ages which is preserved today in ancient mule paths. Some of the latter (between Francavilla
Marittima, San Lorenzo Bellizi, and Alessandria la Carreta) were explored during the SIBA’00 campaign
and were found to have protohistoric sites along them, but it is not yet proven that such sites cluster on
these routes. Figure 2 shows the two main protohistoric routes postulated by Quilici.
9
CONCLUSIONS
The results of the SIBA’00 survey can successfully be used to re-interpret the geographical and
chronological patterns of settlement suggested in the late 1960’s by Lorenzo Quilici. Both the intensive
and the extensive surveys have recovered about twice the number of sites mapped by Quilici in the same
area, despite the deterioration of the soil archive in the intervening decades. The resultant site distribution
suggests, but does not prove, that Quilici’s ‘villages’are artefacts of research and visibility biases. A
definite answer to this question will require targeting a wider area for extensive survey in the future, and
obtaining access to the original research. Although routes may well have existed where he posits them,
sites also occur well away from these along all ‘minor’terrace edges. With respect to dating, the materials
collected in the SIBA’00 survey also suggest consistently earlier dates (Archaic to early Roman) than those
suggested by Quilici (late Hellenistic to early Imperial17); and these results may tentatively be extrapolated
across all of the Sibaritide.
PROTOHISTORIC TO ARCHAIC
The adverse conditions for retrieval of protohistoric ceramics on the intensively worked terraces must be
taken into account when interpreting the archaeological record of relatively well preserved upland
protohistoric sites. There is currently no evidence that protohistoric (BA-EIA) settlement occurred
anywhere in the survey transect but at the very highest elevations (the Monte S Nicola at 500 m ASL
stands some 150 m above the highest terrace). This can be interpreted as confirmation of existing thought
about protohistoric settlement patterns, but it leaves unexplained why the most important hilltop
settlements should be located on hills overlooking the Sibaritide plain. Could the needs of pastoralism –
winter grazing in the plain, which at this time would have been a heavily wooded and seasonally flooded
marginal area with soils too stony or clayey for palaeo-technic agricultural use – have been so vital as to
be the determining element in settlement type and location choice? Alternatively, the postulated ‘ring’of
protohistoric centralised settlements around the plain may, in part, prove to be chimerical, the result of
biases in academic research and interpretation. One indication that this may have been the case is the
apparent concentration of protohistoric research on the hilltops of the fascia collinara (see above, note 4).
Vanzetti recently again stressed the tendency of the Roman school of protohistoric studies to assume the
presence of a settlement where the evidence permits rather than suggests this (Vanzetti 2000,
forthcoming: 7, 23). For example, Broglio di Trebisacce is seen as one single large settlement rather than
as the two nuclei separated by 250 m for which archaeological evidence exists. Equally, settlements are
assumed to be as large as the geomorphologic units (plateaux) they occupy, and ‘missing’periods are
assumed to be present even in the absence of evidence. Taken together, these assumptions may mask
more complex realities in the protohistoric evidence. We feel the possible continuation of the prehistoric
settlement pattern into the hinterland must be studied before further regional interpretations can be
made.
Hellenistic ceramics were dated during the survey by G-J. Burgers, using ceramic typologies developed by D. Yntema at the
Free University of Amsterdam.
17
12 -
20
F IELD W O R K
IN THE
S IBARITIDE
For the Later Iron Age and Archaic periods, the evidence from the survey is equivocal. Despite the
historical evidence for the establishment of Sybaris around 720 BC and the archaeological evidence for
continued use of the sanctuary and necropolis at neighbouring Timpone Motta into the late 6th century
BC (Attema et al. 2000), no securely datable materials from this period have been found in our survey.
Given our experience with the very low visibility of coarse impasto wares among the naturally occurring
stones in the survey area, we must conclude that neither our own surveyors nor the Quilici team were
able to identify such material with any degree of reliability; we cannot therefore infer anything from its
absence. Further fieldwork will be needed to produce a more reliable distribution map of this material.
For the Archaic, much will depend on a closer dating of the coarse wares, which make up more than half
of the finds by weight, through association with datable fine wares or through typological comparison
with excavated material within the region.
HELLENISTIC TO ROMAN IMPERIAL
No such visibility problem occurred with the classical Hellenistic and Roman ceramic types. It appears
from the results of our field work that the large-scale spatial patterns mapped by Quilici in the 1960’s are,
at least for the transitional zone between the plain and the hinterland, partially correct. Large and small
sites of the Hellenistic/Roman period do occur in elongated clusters along ‘major’ terrace edges.
However, such sites also appear to cluster locally along the edges of small valleys cut into the terraces; this
is at least the case along the eastern edge of the Vallone Organata /di Franceschiello which runs just to the
east of Lauropoli. The possibility that Quilici’s site clusters are caused by, rather than merely correlated to,
modern accessibility and land use must remain a hypothesis at this stage; it may eventually be confirmed
or rejected when we obtain access to his survey archive.
Among the ‘new’small classical sites identified by the intensive survey, several are located within the
clusters initially identified by Quilici while others are scattered all over. It was observed both by the
Quilici team and in the SIBA’00 survey18, that many of these Hellenistic-Roman ceramic scatters are
relatively small and poor, with a ‘standard’assemblage consisting of some coarse and depurated storage
and kitchen ware, a few bits of fine ware including Black Gloss, and one small pyramidal loom weight. To
confirm this impression, a comparative study of the assemblages taken from several such sites both in the
Sibaritide and in the Brindisino is being planned.
It now remains to put these results into the general context of Roman Calabria (after Accardo 2000).
Despite the settlement expansion apparently ongoing from the late 4th century BC, the Greek colonies of
southern Italy were already under pressure from the equally expansive indigenous Bruttii by the early 3rd
century BC, and the Romans became involved when they were called in against a coalition formed by the
Bruttii, the colony of Taras, and Pyrrhus of Epirus. We are told that, when this coalition was finally
defeated in 272 BC, much land in Calabria was already confiscated from the Bruttii; resistance flared up
again during the second Punic war and by the end of the 3rd century the remaining Bruttii are enslaved.
Both these wars and the advent of malarial disease caused a severe decline of the Greek city-states during
that same period.
Following a period of military occupation and administration, Roman colonies were established on ager
publicus (and other towns revived or refounded) in Calabria in the course of the 2nd century BC – Copia
Thurii being one of the first. From this time until the 2nd century AD, medium to large-sized agricultural
villas spread in Calabria mainly in the territories of these coloniae. The Via Popilia, extending the Via
Appia from Capua to Rhegium and probably built in 132-1 BC, ran through the Sibaritide from Muranum
in the north to Consentia in the south, probably following the valleys of the middle Coscile and Crati
rivers19 and with secondary roads to Copia. These roads and rivers must have been extremely important
Quilici remarks that “black gloss is encountered only rarely and generally the assemblages are poor with no evidence for
substantial constructions; there are no cisterns and graves are very unassuming as well” [my translation].
18
19 Accardo (2000:30-39). Thought to have been navigable, the former for a distance of 18 km until the confluence with the Esaro;
the latter for 40 km, at least halfway to Consentia.
12 -
21
V A N L EUSEN : P A T T E R N T O P ROCESS
in enabling the marketing of agricultural produce, and archaeological indicators of colonisation and
Romanisation are therefore likely to cluster on them20. In the centuries AD, the last traces of the
institutions of Magna Graecia were supplanted by those of the Roman state, and areas of extensive
(latifundia) and intensive (villa) cultivation developed.
However, besides the land designated for the colonies, large tracts of land were also sold to rich Romans
who developed slave estates from about 100 BC. High land prices and the capital outlay needed for
market-oriented production of wine and olive oil induced many small farmers to sell their land. A slavebased capitalist villa economy developed in Calabria, as elsewhere, into the early Empire. From the
beginning of the 4th century AD Calabria suffered from the general crisis of the Roman Empire, and
some 40% of villas were abandoned or subsumed into latifundiae founded on a combination of cereal
farming and pastoralism. By the end of that century these had developed into independent and fortified
power bases for the potentiores, foreshadowing the feudal estates of the middle ages. However, literary
references indicate that the region in general continued to be productive into the 7th century AD.
Only one clearly Roman villa, possibly of the platform type, was (re-) located (from Quilici site 123)
during our survey, confirming the results of the 1995 survey on the terraces between Francavilla and
Broglio di Trebisacce and the pattern of widely spaced villas with no off-site distribution as observed by
the Quilicis. While it is possible that a denser pattern of villas was located in the plain, its depositional
history currently precludes us from finding proof of this; a study of Medieval and sub-recent historical
documents may be helpful in back-tracing Roman settlement patterns. The same goes for the Byzantine
and medieval periods, for which no physical evidence was recovered at all during the survey although
several chapels and a monastery are known to have existed in the region21. Future research should be
targeted at those local geomorphologic units in the plain for which there is evidence or at least a
possibility that antique and later remains might be observable.
ACKNOWLEDGEMENTS
Aspects of the research reported here have previously been the subject of student projects at the GIA.
We would especially like to thank Kathelijne Kruidhof for her initial analysis of the Quilici data set,
Erasmus exchange student Lia Pala for helping prepare the Quilici maps for digitisation, Wieke de Neef
for her excellent MA thesis on the Iron Age to Archaic land use of two neighbouring Greek colonies, and
Michiel Rooke, Jasper Huis in ‘t Veld, and Tymon de Haas for their preliminary analyses of the survey
data collected since 1995. Further thanks must go to the mayor and council of Francavilla Marittima for
furnishing us with excellent accommodations at the Timpone Motta site museum. Finally, we would like
to thank the Soprintendente, Silvana Luppino, for the interest she has shown in our work and for her
guidance in obtaining the necessary permissions.
REFERENCES
Accardo, S 2000
Villae Romanae nell’Ager Bruttius. Il paesaggio rurale Calabrese durante il dominio Romano (Studia
Archeologica 107). Roma: Bretschneider.
Attema, P, G-J Burgers, M Kleibrink & D Yntema 1998
Case studies in indigenous developments in early Italian centralisation and urbanisation: a Dutch
perspective. European Journal of Archaeology 1:326-81.
Attema, P, G-J Burgers & M van Leusen (in press)
The Ostuni Field Survey Campaign 1999. Studi di Antichitá (2002).
Attema, P, J Delvigne, E Drost & M Kleibrink 2000
Habitation on plateau I of the hill Timpone della Motta (Francavilla Marittima, Italy): a preliminary
report based on surveys, test pits and test trenches, Palaeohistoria 39/40 (1997/1998):375-411.
Attema, P, E van Joolen, & M van Leusen 2001
According to Accardo (2000:56), villas in the territorium of Thurii lie mostly in these valleys, are medium-sized and specialise
in the production of wine and oil. Almost all date from the 2nd c BC to the 2nd c AD; some continue into the 4th and 5th c AD.
20
21
One Byzantine chapel was located on top of the Timpone della Motta at Francavilla Marittima.
12 -
22
F IELD W O R K
IN THE
S IBARITIDE
A Marginal Landscape: Field work on the beach ridge complex near Fogliano (South Lazio).
PaleoHistoria 40/41:149-162.
Bergonzi, G, V Buffa, A Cardarelli, C Giardino, R Peroni & L Vagnetti 1982
Ricerche sulla protostoria della sibaritide. Cahiers du Centre Jean Bérard VII-VIII. Naples: Institut
Français.
Bintliff, JL (ed) 1991
The Annales School and Archaeology. Leicester [etc]: Leicester University Press.
Cambi, F. & N Terrenato 1994
Introduzione all'archeologia dei paesaggi. Studi Superiori 203. Roma: La Nuova Italia Scientifica.
CGC 1969
Carta Geologica della Calabria. Foglio 221-II NE della Carta d’Italia 1:25.000 dell’IGM. Cassa per
opere straordinarie di pubblico interesse nell’Italia meridionale.
CNR 1956a
Carta dellútilizzazione del suoli, 1:200.000, Foglio 19. Consiglio Nazionale delle Ricerche, direzione
generale del catasto and Touring Club Italiano.
CNR 1956b
Carta dellútilizzazione del suoli, 1:200.000, Foglio 20. Consiglio Nazionale delle Ricerche, direzione
generale del catasto and Touring Club Italiano.
D’Angelo, S & F Ch Oräzie Vallino 1994
La Sibaritide. Lineamenti geografico-ambientali ed insediamento umano, in Peroni, R & F Trucco
(eds) 1994:785-829.
De Haas, T 2001
Over scherven, samples en sites. Beoordeling en analyse van vondstpatronen van diverse surveys in de
Sibaritide. Unpublished report: University of Groningen.
De Neef, W 1998
Patronen van Verandering. Landschapsgebruik en menselijke activiteit in de Metapontino (Basilicata)
en de Vlakte van Paestum (Campanie) in de Ijzertijd en de Griekse kolonisatieperiode (1000-400
v.Chr.). Unpublished M.A. thesis, University of Groningen.
De Rossi, GM, L Pala, L Quilici & S Quilici-Gigli 1969
Carta Archeologica della piana di Sibari, extract from Atti e Memorie della Societa’ Magna Grecia,
Nuova Serie IX-X (1968-1969): 91-155. Roma: Societa Magna Grecia.
Feiken, R & J Weterings 1998
Monte Sellaro ‘98: een survey van de Monte Sellaro. Unpublished report: University of Groningen.
Greco, E 1996
City and Countryside, in Pugliese Carratelli (ed) 1996:233-242.
Haagsma, BJ 1996
Survey in de Sibaritide, Calabrië. Een preliminair verslag van drie campagnes. In TMA 17:47-52.
Lerici, CM 1960
I nuovi metodi di prospezione archeologica alla scoperta delle civiltà sepolte. Milano: Lerici.
Peroni, R 1994
Le comunità enotrie della Sibaritide ed i loro rapporti con i navigatori egei, in Peroni, R & F Trucco
(eds) 1994:831-79.
Peroni, R & F Trucco (eds) 1994
Enotri e Micenei nella Sibaritide. Tarento.
Pugliese Carratelli, G (ed) 1996
The Western Greeks. Bompiani.
Rainey, FG 1969
The location of Archaic Greek Sybaris, AJA 73 (1969):261-73.
Rainey, FG & CM Lerici 1967
The search for Sybaris: 1960-1965. Roma: Lerici.
Ryan, N & M van Leusen, in press
Educating the Digital Fieldwork Assistant, in Burenhult, G (ed), Proceedings of the CAA2001
conference (BAR Int Ser). Oxford: Archaeopress.
Vanzetti, A 2000
Costruzione e problemi dei “paesaggi di potere” nella Sibaritide (Calabria) dall’età del bronzo alla
prima età del ferro, in Camassa, G et al. (eds), Paesaggi di potere: problemi e prospettive (Quaderni di
Eutopia 2):239-83.
Vanzetti, A forthcoming
12 -
23
V A N L EUSEN : P A T T E R N T O P ROCESS
Results and problems of some current approaches to protohistoric centralisation and urbanisation in
Italy, in Attema, PAJ et al. (eds), Regional Pathways to Complexity. Proceedings of a conference held
in Groningen (Netherlands) in April 2000.
Zanotti Bianco, U (ed) 1969
Atti e memorie della Società Magna Grecia, nuova serie IX-X, Rome.
12 -
24
CHAPTER
13
A COMPARISON OF
ARCHAEOLOGICAL DATA SETS
F O R T H E P O N T I N E R E G I ON
1
INTRODUCTION
In chapter 2 of this thesis I discussed the possibilities for supra-regional comparison and high-level
explanation of the settlement dynamics in the three study regions, concluding that comparisons made on
the basis of the available high-level narrative are not satisfactory and should instead be based directly on
an evaluation of the archaeological record. Here, I intend to explore the potential for this type of
comparison from two angles, using data and examples from the Pontine region: firstly, the construction
of a relational database capable of representing the diversity and variable quality of regional archaeological
records and the interpretative constructs based on them; and secondly, the practical comparison of
multiple data sets deriving from a single, or several adjacent, landscape units within the general region.
Given the ultimate goal of spatio-temporal comparison, within a GIS environment, of the Late Bronze
Age to Roman settlement history of the three regions studied by the RPC project, a compendium of
archaeological data available for these regions and periods must be created. In fact, two largely separate
sets of questions have to be confronted
• How to design and build a regional database that can hold all existing traditional site-based
observations and interpretations, as well as new area-based survey data, and allow the addition of
extra layers of RPC interpretations?
The first part of this question has of course been asked many times before, whenever archaeologists have
attempted to create regional-to-national scale archaeological databases. A seminal publication in this
regard has been Larsen’s (1992) volume on national archaeological records which, according to its
editorial preface, was ‘a balanced status of attempts… to computerize the archaeological heritage’.
Although the volume may be said to summarise the experience gained since the mid-1980s by the
national digital records builders of (western) Europe, unfortunately the only high level data models
presented in it are those for the Danish and Dutch national archaeological records (Christoffersen
1992:15, Roorda & Wiemer 1992:119). It seems that documents detailing these and other high level data
models have, since that time, been circulating as internal reports, deemed to be of insufficient outside
interest to warrant publication. The second and third parts of the question are only now beginning to be
tackled with the inception of regional analytical (as opposed to management oriented) archaeological
databases, in the context of research projects such as (within Italy) the RPC and the Tiber Valley Project,
and the project set up in the mid- 1990s at the University of Lecce for collating the indigenous settlement
of southern Italy (Puglia and Basilicata) in a GIS environment. According to Francesco D’Andria
(n.d.:105) the latter project, for the first time, brought together the detailed data needed to effectively
study indigenous settlement dynamics, without which regional models such as those created at the
Accordia Research Institute at the University of London are of doubtful value, but again no detailed
13 -
1
V A N L EUSE N : P A T T E R N T O P ROCESS
publication of the data model has been forthcoming1. Section 2 of this chapter can be regarded as a case
study in the design of wide-area GIS integrated archaeological databases. It introduces and details the
design considerations for the RPC interregional archaeological database.
• How to compare the various data sets which exist within each RPC region, and how to interpret the
results of such a comparison?
The first part of this question, regarding the methodology of comparison, has already been posed in
chapter 2 of this thesis, along with the subsidiary questions of: What grounds do we have for believing
that inter- and infraregional comparisons can be made at all, and: Which things would we want to
compare? The tentative reasons for the comparability of regions given there rested on geographical, ethnoarchaeological, historical (Iliad & Odyssea, classical authors), and archaeological (distribution of elite
goods) arguments; the proposed comparanda included demographic developments, cultural developments,
and processes such as centralisation, urbanisation and colonisation. It was tentatively concluded that our
poor understanding of the problems and potential involved in interregional comparisons would benefit
from approaching the more modest goal of intra-regional comparison first. In section 3 below I therefore
describe, assess and attempt to compare the data sets that have so far been available to the RPC project in
one of its regions. General conclusions regarding the comparison of the core processes within and among
all three RPC study regions are drawn in a final section.
2
TOWARDS AN INTERRE GIONA L DATABASE
2.1
AIMS
In a broad sense, then, the purpose of the RPC database is to collect together all of the available data on
archaeological remains within the three regions of interest, and to use this data in an interpreted form for
archaeological landscape analysis. More specifically, we want to create a database in order to:
1.
2.
3.
4.
collect available site-oriented information from sources
assess and document that information; generate metadata
reorganise it if necessary (eg, terminology, splitting or merging of observations)
interpret it in terms of meaningful entities
We also want it to:
5. hold non-site data types (eg, off-site survey data such as those generated by the RPC project’s recent
fieldwork, and negative observations)
6. be able to answer specific archaeological questions about spatial patterning
The database must hold the unaltered data as presented by the available sources, and keep them separate
(and separable) from any additional or transformed data and interpretations. A particularly important role
will be set aside for metadata (point 2 above), because of the need to prevent low quality data from
inadvertently ‘polluting’our analyses:
a) We want to be able to specify a particular spatial and/or temporal scale within database queries
b) We want to be able to apply interpretative ranking criteria (point 4 above) to queries
c) We want to be able to apply criteria of data precision, accuracy, and probability to queries.
DESIGN PRINCIPLES FOR THE RPC PROJECT RDBMS
Point 4 above requires that we should be able to organise observations (~ ‘events’) into recursive
hierarchies of interpretative entities2, and that we be able to distinguish interpretations by phase or period.
1
Possibly, Semeraro (1997) has some description of database design and methodology of data creation.
13 -
2
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
For example, in the case where a particular period is ‘missing’at a site, we should be able to choose
between regarding the suite as unoccupied during that period, or regarding it as a ‘possible’or even
‘probable’site. The importance of making such distinctions was recently highlighted again in a paper by
Alessandro Vanzetti (forthcoming), when he noted the tendency of the Roman school of protohistory to
assume continuity in cases where there is a geographical or chronological hiatus in the archaeological
evidence.
Given the problems discussed above in interpreting existing archaeological records, it is essential that the
database framework should keep the ‘information trail’intact so as to allow the researcher to access and
check all data transforming steps occurring in between raw data input and high-level interpretation. This
means that compiled data should be entered into the database unaltered in any way3 and that all subsequent
interpretative steps are to be fully formalised and documented.
The database should also be scale independent, that is, it should be able to hold data irrespective of its
level of detail (size, duration). This principle has long been championed by Arroyo-Bishop and Zarzosa
(1992, 1995) in their object-oriented database system called ArchaeoDATA. Their database design
permits features, artefacts, ecofacts, and any other object to be linked into interpretative entities at any
scale.
The landscape archaeological approach of the RPC project calls for a reconsideration of the ways in
which basic information about archaeology should be structured. The current approach to regional
archaeological database design, including that of the RPC project, is to structure the database around the
archaeological entities (typically, sites). The database structure mirrors the processes by which
archaeological information has been produced, must be transformed, and will be analysed: field
observations are turned into ‘records’(for publication or archiving) through various processes by a
‘source’, who usually (but not always) groups and interprets them as ‘sites’before publishing them; these
constructs must then be deconstructed again by us into their constituent ‘observations’,and reinterpreted
in an iterative process as complexes of observations (interpretative hierarchy).
An alternative structuring principle presents itself if a close integration of the RDBMS with a GIS is
desired. Rather than the 'antiquarian' approach in which an ever increasing stock of knowledge is built,
the 'landscape' approach takes the presence of a limited amount of geographical space containing a
limited amount of archaeological resources as its starting point. This geographical space is represented by
a limitless number of themes (or map layers) considered relevant by the user; each theme is partitioned as
and when 'events' relevant to that theme occur. For example, the event of a survey taking place will be
recorded in a 'research activity' theme, while the discovery of a Roman villa during that survey will be
recorded in a 'Roman period' theme. Each 'event' not only has a specific spatial extent, but also a
temporal extent (or duration). Basing the database design upon this principle is preferable from a
landscape archaeological viewpoint as well, which also looks at cultural remains in the context of the
landscape as a continuous whole.
2.2
DATA
The first step in creating a regional archaeological database is to compile existing published or archived
material. For the RPC regions there are three main types of sources: published volumes of the Forma
Italiae series of map sheet surveys, publications in local (Italian) journals, edited volumes and monographs,
and the publications and records kept by GIA and AIVU themselves. This represents a mixture of
primary, secondary, and even tertiary sources so in some cases the link with an actual field observation
In the Dutch ARCHIS system (version 1), observations are aggregated into meaningful sets termed ‘complexes’(Roorda &
Wiemer 1992:118-120); ARCHIS 2 is intended to include a set of rules for performing this aggregation. A ‘recursive’hierarchy is
one which allows such complexes to be aggregated into ever higher level interpretative entities.
2
3
Translations, if subject to proper quality control, are allowable.
13 -
3
V A N L EUSE N : P A T T E R N T O P ROCESS
Table 1: Sources for the RPC site database, Pontine Region
RPC Map ID
10,101 – 10,152
10,401 – 10,500
10,501 – 10,530
10,551 – 10,594
10,601 – 10,700
10,701 - 10,854
10,855 – 10,900
10,901 – 10,951
10,952 – 10,963
Project / Source Reference
Lanuvium Survey (Attema forthcoming)
Gierow 1964a, b
Ninfa Survey 98/99 (Van Leusen 1999)
Fogliano Survey 98/99 (Attema et al. 2000)
Chiarucci 1978 (only 1 site entered)
Segni Survey (Attema 1997, Carpino 1997, Enei 1990)
Attema 1993
Sezze Survey (Attema, Zaccheo & Pasquali)
Norba Survey (King 1995)
Cisterna Survey (Attema 1993)
Olmobello Survey (Attema)
No of sites
52
67
24
38
11,001 - 11,200 /
11,401 – 11,445
11,201 - 11,396
11,501 - 11,600
11,601 - 11,801
11,802 – 11,934
11,951 – 12,000
12,001 – 12,171
Lugli, Forma Italiae: Terracina
Picarreta 1977, Forma Italiae: Astura
Lugli, Forma Italiae: Circeo
Vittucci 1968, Forma Italiae: Cori
Drost (unpublished)
Mazzolani 1969
Morselli & Tortorici 1982, Forma Italiae: Ardea
196
~76
201
133
51
171
13,001 – 14,000
Agro Pontino Survey (Voorrips et al. Unpublished)
381
153
46
51
12
297
Figure 1: Survey areas of the Pontine Region. White: forma italiae (1 Ardea, 2 Cori, 3
Astura, 4 Circeo, 5 Terracina), Black: Agro Pontino Survey areas 1981-1989,
Orange: Pontine Region project & RPC project survey areas (A Lanuvium, B Segni,
C Ninfa, D Norba, E Olmobello, F Sezze, G Fogliano)
13 -
4
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
will be quite tenuous. The complexity of this dataset may be illustrated by the listing of sources for just
one of the three study regions, the Pontine region (see table 1 and figure 1). Blocks of 5-digit RPC map
codes are set aside for each major publication. Table 1 identifies these sources and the conversion used to
obtain RPC Map ID’s from the site ID’s used by the source. A complete concordance between source
ID’s and RPC Map ID’s can be found in the RPC sites database.
At current count, the database for this region contains about 1880 site observations. This extant
archaeological record cannot be taken at face value, but must be interpreted on at least three different
levels during compilation. Firstly, the meaning of the terms used may have changed over time or may be
idiosyncratic as records are usually produced by many different people; secondly, interpretation and
observation are usually mixed together and sometimes a description of the original observation fails
completely; and lastly, elements of the record may contain errors, inaccuracies, and imprecisions (cf.
Scollar 1992). All of these require careful interpretation if the compilation is to become a useful, and
usable, archaeological database.
QUALITY
If we are to interpret what the sources tell us in terms of the classification system presented above, we
also need to find ways of dealing with the ‘fuzziness’of the source data. This can take any of several
guises: lack of clarity, or overlap, in the definition and scope of descriptive terms; lack of distinction
between observations and interpretations; and measurement errors and uncertainties.
What does it mean (terminology)?
Many of the terms used to describe archaeological field observations, and constructs based upon them,
do not have precise definitions, have been used differently by different authors or by the same author
over time, and/or have been used to describe overlapping sets of archaeological observations. The lack of
formal definitions to accompany the terms used precludes a precise comparison of results; one example
where different terms have been used to describe broadly similar phenomena is Perkins’use of ‘village’in
preference to ‘nucleated settlement’(cf. Section 2.4 below). The latter term is employed by Burgers (1998)
who distinguishes the following four levels of settlement: dispersed settlement, nucleated settlement,
hamlet/village, town. Attema (1993) also distinguishes four levels of settlement, but used different terms:
isolated farm, hamlet, proto-urban settlement, urban settlement. Neither gives precise criteria by which to
distinguish the levels, but they do seem to attempt to describe similar phenomena.
How much of it is interpretation rather than observation?
In bringing together information provided by many sources and over a long period of time, we cannot
assume that all the interpretations made by these sources were either correct at the time, or have remained
so ever since. In an ideal world we would be able to separate these interpretations from the observations
that they were based on, but this is unrealistic for two reasons. Most importantly, no such clear distinction
exists in reality between ‘observations’(a term suggesting value-free data) and ‘interpretations’(a term
which acknowledges the changeable nature of what we consider to be the significance of archaeological
remains). Hence it may be better to regard all observations as interpretations as well (eg, Scollar 1992:98).
Secondly, many sources record little if any of the descriptive information on which their interpretations
are based, and in practise we will therefore not be able ever to evaluate the nature of most historical
primary (or even secondary) records. What we can attempt to do is to separate such records into
descriptive elements and interpretative elements; this would allow us then to ignore previous
interpretations and attach our own interpretations to the compiled descriptive information instead.
What is the precision and accuracy of measurements and estimates?
All measurements and estimates given by the source (for example, of the number, size and density of
finds in a scatter, or of its location) carry the implicit properties of precision and accuracy. Some of this
fuzziness is indicated by the source in an often informal and inconsistent manner, through the use of
qualifications such as ‘probably a site’, ‘nearby’, ‘late Republican or early Imperial’; but more often such
13 -
5
V A N L EUSE N : P A T T E R N T O P ROCESS
indicators remain implicit4. Within a GIS, when precise characteristics of large numbers of site locations
are derived from their recorded locations (co-ordinates), even relatively small errors can have large
consequences. For example, let us assume that a site’s location has been stated unambiguously by a pair
of co-ordinates rounded to the nearest 10m grid point; its GIS-derived slope might then be an equally
unambiguous 3% and its aspect NE. However, if we know that the co-ordinates provided by the source
have a measurement error of 50 meters, then the true location of the site is no longer unambiguous – it
may be anywhere within a 50m radius of the co-ordinates, though it has a higher probability of being
close than of being far away. In this latter case the slope of the site could be derived by the GIS as ‘3%
with a probability of .67, 2% with a probability of .33’, and its aspect as ‘.50 NE, .25 E, .25 none’. It is
therefore very important to include assessments of the precision and accuracy of such measurements
where-ever possible. There are at least two ways in which such assessments can be made:
1. by codifying our degree of confidence in the general accuracy of the information provided by a
source, and of its judgement in individual cases;
2. by deducing measurement error from the available information about the methods employed by the
source, and applying this error to its observations.
A standard locational error can be deduced if the measurement method is known. Thus, locations
determined on the basis of topographical maps contain a standard error due to the minimum size at
which map features can be drawn, equal to 1/2000th of the map scale denominator; for a 1:25000 scale
topographic map, this error is 12.5m. This represents the minimum error, assuming that the original map
features were drawn at maximum precision and that no other degrading factors (e.g., map reprojections)
were involved.
A similar standard error is associated with measurements taken using survey equipment based on GPS
(see chapter 7). The size of the error depends on the type of equipment (survey quality equipment having
a much higher accuracy than ‘navigational’GPS) and even the location, date and time of the measurement
(the notional horizontal accuracy of ‘navigational’GPS has gone up from about 70m to about 10m after
the removal of signal degradation on 1 May 2001; however, in all cases the accuracy of measurement
varies throughout the day with changing satellite configurations).
Errors and omissions
There may well be errors in the source records; these may be incidental (as in the mistaken identification of
a Roman terrace wall as a road revetment) or pervasive (as in the misidentification of African red slip ware
before Hayes published his typological study of it in the 1970s). The database framework must allow for
both types of error to be corrected, and for the correction process itself to be formalised. If the
inaccuracy cannot be corrected, then the database framework must allow it to be described and flagged so
that the inaccurate record can be avoided during queries. Archaeological records are also typically
incomplete, ie they do not contain the full complement of data elements. Since missing data may or may not
be fatal to the interpretation, the database framework must contain a formal procedure for deciding how
the absence of any data element affects any query (ie, queries must contain a set of rules about the
minimum requirements to each data element).
METADATA FOR SOURCES AND OBSERVATIONS
In all this, it is clear that a very important role in the interpretative process will be reserved for our
assessment of our sources of information. Can they be trusted? What are their limitations? How confident
are we that the recorded information is accurate? The implication for the data model to be constructed is
that data fields derived from external sources (table A) must be linked to a table holding both descriptive
and evaluative information about these sources (table B).
Attema (pers. comm.) notes that sites mapped on old IGM 1:25,000 map sheets in his 1987 transect survey have a particularly
fuzzy location because of the regular absence of physical controls (field boundaries, etc) in the neighbourhood of the sites.
4
13 -
6
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
Table A: OBSERVATIONS
ID
1
2
Descriptive fields
… .
…
Short ref
Picarreta 1977
Picarreta 1977
Ref_spec1
pp 30-34
pp 35
Ref_spec2
Site 16
Site 17
Table B: SOURCES
ID
1
Short ref
Picarreta
1977
2
…
Long ref
Picarreta, F 1977
Forma Italiae regio I, volumen XIII: Astura.
Firenze.
..
Metadata
Topographic
…
survey;
…
Table C: OBSERV_META
Obs_ID
1
2
Obs_date
1969/08
…
Obs_method
Topo survey
…
Obs_LULC
arable
…
Obs_circ
Not specified
…
Other aspects of data quality do not concern the source itself but rather the circumstances under which
he or she collected information. These require that metadata fields regarding the circumstances of
discovery and description be added. Three sets of such metadata can be defined (Table C): the research
method, ranging from archive study to excavation and aerial reconnaissance; the land cover / land use
(LULC, see also chapter 14); and the specific find circumstances, eg ‘chance find’, ‘in eroding bank along
canal’. Since these metadata apply only to the time frame in which the observation was made, a fourth
metadata field is needed to record the observation date.
2.3
UNUSUAL DATA TYP ES
Until recently, archaeological records everywhere contained only ‘sites’, essentially point observations
ranging from single stray finds to large complexes of earthworks, which were documented in a series of
text/number fields, occasionally referenced to archived materials such as notes, sketches and
photographs, and indicated as a point, line or area on fairly small-scale (usually 1:25,000) topographic map
sheets. Initially, the advent of GIS did nothing to change this situation. No provision was made within
archaeological databases for the inclusion of non-vector topological data types such as continuous off-site
ceramic densities and toponymics, of probabilistic data types, or of linked or embedded documents
(including graphical documents). A number of potential approaches exist to allow us to include such
unusual data type in the RPC RDBMS.
Firstly, the quality of regional archaeological databases could be significantly enhanced if the database
framework would allow the inclusion of unusual document types (‘objects’) such as word processor files,
annotated aerial photographs, graphics, etc. The mechanism for this already exists in the current
generation of office database software (‘Object Linking and Embedding’or OLE). A primary benefit of
this would be that it allows us to include all of the original source data in the database, and not just the
descriptive text.
SURVEY DATA AND ‘NONSITE OBSERVATIONS’
The cartographic heritage of traditional archaeological records is still with us in the current generation of
archaeological databases, which are geared to recording the presence of archaeological entities and
assemblages (‘sites’), and completely ignore the requirements of modern landscape archaeology which are
to record information for all parts of the landscape (including what has been termed ‘off-site’and even
‘non-site’observations5). In practise, this has meant that information from landscape survey must be
‘degraded’into a set of ‘sites’before it could be included in the database, and that an observation of site
5
Contra Scollar (1992:98), who believes that recording absence data only results in needless clutter.
13 -
7
V A N L EUSE N : P A T T E R N T O P ROCESS
absence (eg, during a watching brief) could not be recorded at all. Since the RPC database is to contain a
significant amount of data resulting from landscape surveys, its database framework must be able to deal
with both these data types.
In addition to knowing where archaeological remains were found, it would be very important for a student
of a regional archaeological record to know where they were looked for. In other words, the history of
research and discovery within the region has to be documented within the database as well. A workable
format for this consists in the mapping of the ‘activity areas’of those individuals who have contributed
multiple observations to the database, either in the form of formal study areas (eg, Forma Italiae sheets,
municipal inventories) or of informal activity areas (active amateur archaeologists, cf. Fokkens 1991), and
in the mapping of interventions (eg, infrastructural works) observed by an archaeologist. The database
framework should allow each observation to be related to such an ‘observation window’, which itself
could be recorded in a vector format in the associated GIS. This format also allows efficient recording of
‘nonsite’, that is, site absence observations. As a first approach, the outline of all study areas could be
recorded as simple polygons; the possibility of going into more detail (recording of individual agricultural
field boundaries, grid units, or even transects surveyed) should be studied.
The integration of a GIS with the database part of an archaeological record now provides the means by
which the ‘landscape’approach can become the leading principle. Information about the absence of finds
and the absence of observations can now be assigned with ease to a variety of spatial objects, and these
can be endowed with ‘fuzzy’spatial and temporal properties. The analytical use of such an approach lies
primarily in the possibility to reason about bits of landscape rather than about the archaeological remains
that happen to have been recorded in it. In a very pragmatic sense, once we have data layers recording the
history of land use and archaeological research and discovery within a region of interest, we can start to
implement formal bias models which are able to distinguish between absence of evidence and evidence of
absence.
2.4
CLASSIFICATIONS
Classifications of objects and assemblages of objects (eg, sites), invariably informal in the compiled
sources, have to be formalised in order to ensure that similar phenomena are interpreted in a similar
manner in the database. The first step in this is to create authority lists which formalise terminology and
limit the freedom of classification; such lists can range from simple option lists (providing essentially a
limited vocabulary with which to describe the objects and assemblages) to full thesauri (which describe
the conceptual domain of the classification and can provide it with a polyhierarchical structure). Many
international examples of such authority lists can be found in Larsen (1992); the full and detailed authority
lists for the Netherlands central archaeological register (Brandt et al. 1992) run to more than 300 pages.
An authority list becomes a classification by the addition of criteria. The simplest type of classification is
one where the classes are mutually exclusive and the class boundaries can be established in an
unambiguous manner, such that all instances can be assigned to a single class. Site typologies are a good
example of this and will be looked at in more detail below.
A COMPARISON OF SITE TYPOLOGIES
The vast majority of classifications used by archaeologists to distinguish site types on the basis of surface
survey data are based on either ‘historical’or ‘empirical’approaches. Two recent exceptions to the
unfortunate fact that such classifications are usually not properly formulated and published are presented
here by way of illustration. The historical approach, making use of historically identified Roman site types,
was pursued by Arthur for his survey in northern Campania (1991:19-21). The criteria for his
classification, and the interpretations based on it, are presented here in Table 2. The alternative, empirical,
approach is based on sets of qualitative and quantitative criteria established in a more or less pragmatic
and ‘ad hoc’manner by the researchers involved. Among the criteria used in the past in central Italy to
classify sites have been the presence of architectural elements (Cosa survey, Dyson 1978:275), size
(Molise, Lloyd & Barker 1981:296; Agro Pontino survey, Koot 1991:130), and location and assemblage
13 -
8
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
(Liri Valley survey, Wightman 1981:281-5). The ‘polythetic’classification of surface scatters devised by
the Albegna Valley / Ager Cosanus survey team on the basis of overall criteria of size, shape and density,
and specific finds or find types provides a good example here (Table 3; Perkins 1999a:106-7, 1999b:1826).
The differences and similarities between these examples are instructive. Arthur (1991:19) rejects the
simple mechanistic criteria applied by Perkins, such as scatter size or the presence of particular structural
features, in favour of more complex combinations of criteria that parallel classical terminology. While
these, in turn, have the advantage of being readily interpreted in terms of their place in the landscape and
function in society, their classical focus and strong (compulsory) criteria fail to address the (undiagnostic,
pre- or post-Roman) nature of the majority of actual sites recovered by modern surveys. The latter tend
to fall in Arthur’s undifferentiated class ‘pottery scatter’, and here Perkins’classification based on the
application of relatively simple size criteria appears more practical6. Perkins accepts that, without
excavation, not even Roman sites will consistently provide the type of evidence needed for an
unambiguous classification into historical types.
Table 2: Historical classification of Roman site types for the northern Campania survey
(after Arthur 1991)
Class
Town
(Colonia,
Municipium)
Forum
Vicus
Pagus
Sanctuary
Villa
Maritime
villa
Farm
Cemetery
Pottery
scatter
Criteria
1: material indicates varied and distinct areas of
activity beyond the purely agrarian, eg exchange,
manufacturing, ritual
2: site is sufficiently large as to indicate the habitation
of various family nuclei
3: presence of distinct public buildings provided these
are not purely religious [otherwise class as village]
1 & 2: as 1 & 2 above, but manufacturing evidence
may be absent or outweighed by agrarian evidence
Identical to Forum, but in a setting that indicates
spontaneous rather than planned development
Not readily recognisable except through inscriptions
etc
1: Concentration of votive material
1: Stone and/or tile built rural structure with clear and
differentiated functional areas for agricultural and
residential use
2: residential areas differentiated by size and quality,
some with, eg, bath structures and interior decoration
1: as 1 above but can also be suburban
2: as 2 above but with clear evidence of luxury, eg
mosaics, wall paintings, architectural marbles
3: proximity to the sea
1: Stone and/or tile built rural structure with evidence
of domestic occupation and areas of agricultural
activity
[2: evidence for a degree of comfort]
3: no evidence for the presence of internal social
hierarchy (cf Villa)
1: distinct clusters of specific artefact types associated
with tomb construction or grave goods, within a welldefined area
[2: presence of multiple frs of the same object,
presence of human bone]
3: absence of heterogeneous fragmentary artefacts
indicative of rubbish accumulation
1: not identifiable as any of the above
2: definable margins
Interpretation
Basic Roman political and administrative unit;
diversified economic base; productive, marketing,
and mercantile facilities; population consisting of
multiple families; developed street system
Public administrative centre; small nucleated
settlement with productive and market facilities
Smallest legally recognised unit of nucleated
settlement, with basic market facilities; hamlet
Territorial division which can contain one or more
vici
Site of religious congregation, usually also
political and (controlled) market function
Agricultural estate centre with resident slave
familia as rural workforce
Vacational residence with sea view; eventual
productive functions are secondary
Agricultural family establishment; production may
be partly based on hired labour or tenants, but
not on slaves
Group of burials
Small single family settlement site; outbuilding;
temporary activity area; cemetery
Table 3: Polythetic classification of surface scatters for the Albegna Valley / Ager Cosanus survey
(after Perkins 1999a,b)
So far as I have been able to ascertain, no theoretical context is given for setting size criteria for site type classifications; it is just
a way of formalising subjective impressions of ‘small’,‘medium’,and ‘large’scatters.
6
13 -
9
V A N L EUSE N : P A T T E R N T O P ROCESS
Class
House
House or
tomb
House or
necropolis
House 2
Villa
Fortified hilltop
Village
Minor centre
City
Kiln
Sporadic
criteria
2
1: scatter smaller than 100 m and containing a loom weight,
slag, a grinding stone, or pithos when associated with roof tile
2: consistent and continuous scatter of roof tile and/or pottery
2
2
and/or building stone larger than 25 m but smaller than 100 m
2
1: thin scatter smaller than 200 m with no other distinguishing
characteristics
2
1: thin scatter larger than 200 m with no other distinguishing
characteristics
1: identical to ‘house’but with additional evidence for structures
in concrete, cocciopesto, or floor tiles
2: scatters between 0.15 and 0.25 ha in size, with large
quantities of building materials, but no structural or architectural
evidence
1: presence of standing structure, eg cryptoporticus
2: presence of architectural evidence, column drums, bases or
capitals, or painted wall plaster
3: scatter larger than 0.25 ha with dense concentration of esp.
building materials
4: bibliographic accounts
1: presence of defences around a hill top smaller than 4 ha
1: identical to ‘house’but extends over more than 0.1 ha and
less than 4 ha
2: a series of distinct scatters classifiable as ‘house’, each larger
than 0.01 ha but extending over less than 4 ha as a group
1: identical to ‘house’but extends over more than 4 ha
2: presence of defences enclosing more than 4 ha
Temple
1: identical to ‘house’but extends over more than 30 ha
1: presence of ceramic wasters
1: thin or diffuse scatter or stray find, not classifiable in any other
class
1: architectural terracottas of temple type
Road
1: alignment of stones or a cutting in rock
comments
- each artefact class is indicative of
food preparation or domestic crafts
- pithoi without roof tiles may indicate
tombs
- in the absence of decisive finds,
stone or tile may also indicate a tomb
- the only difference with ‘house or
tomb’is the size
- Roman period only
- Roman period only
- poorly understood but in generally
strategic location
Aka ‘nucleated settlement’
- distinguished from ‘village’and
‘fortified hill top’on the basis of size
- may include modern settlements
with no other type of evidence
- may also occur at domestic sites
- often used for small quantities of
earlier material found at Roman site
- in practise, these have occurred
only in relation to known temples
-
All of the above approaches can be contrasted to the statistically derived site type classifications
developed and employed within the ARCHEOMEDES programme (Favory and Raynaud 2000), which
are based on a hierarchical cluster analysis of a large number of archaeological and environmental
variables measured for 934 sites in the Hérault region of southern France. Since such statistical
hierarchisation of site types is beyond the scope of the RPC project given the current state of its site
database, and Arthur’s classification is of very limited use to the diachronic and undiagnostic survey data,
it was decided early on to develop an empirical classification very much in the manner of Perkins. The
table below gives the proposed interpretative classification for sites in the RPC database, which will only
become operational by the addition of sets of criteria by which to distinguish the classes7.
IMPLICATIONS
A good illustration of the wider implications that may be attached to the proposed terms and criteria is
the difference between a ‘hut’and a ‘villa rustica’or rural villa. Villas are defined as Roman buildings with
(amongst other criteria) tiled roofs which indicates that they were part of a wider distributary system,
whereas huts are buildings for which no such durable materials have been used. Yet in Lazio Archaic
farm buildings often have a (partly) tiled roof – should such buildings be called huts or villas, or should a
third class of building be added? Clearly, a farming settlement for a single family unit will move from one
class to another over time, as society changes and the physical building style with it.
Table 4: proposed site type classification for the RPC database.
7
The final synthetic volume on the RPC project (in prep.) will include this work.
13 -
10
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
label
rural settlement site
Order 1
A
label
isolated hut / capanna
isolated farmstead / house
fattoria
small fattoria
large fattoria
small cluster
large cluster
hamlet / village
villa rustica
outbuilding
cult place
B
sanctuary
mountain sanctuary
cave sanctuary
source sanctuary
temple
shrine / altar
votive deposit
town / urban site
C
production site
D
colonia
quarry
kiln
pottery kiln
amphora production site
latifundium
metal kiln
oil / wine production site
infrastructure
E
road
bridge
mansio
statio / mutatio
ford
centuriation ditch / bank
drainage / irrigation
F
burial
G
cuniculus
isolated grave / tomb
grave / tomb group / cemetery
unknown
uncertain
other
X
Y
Z
hoard
maritime villa / resort
(watch) tower
Order 2
A00
A02
A03
A04
A05
A06
A07
A08
A09
A10
A11
B00
B01
B02
B03
B04
B05
B06
B07
C00
C01
D00
D01
D02
D02a
D02b
D03
D04
D05
E00
E01
E02
E03
E04
E05
E06
F00
F01
G00
G01
G02
X00
Y00
Z00
Z01
Z02
Z03
It should be kept in mind that classifications, just like authority lists, are interpretative constructs even
though they may not have been intended as such. Thus, whilst the classification of field observations may
proceed on the basis of objectively applied criteria, the terms used to describe the classes represent
interpretative concepts and put the student into a specific theoretical context. In the case of the site type
classifications presented above, this context is predominantly that of the formalised and structured
classical landscape. Observations relating to the indigenous, and informal classical, landscapes are deemphasised by the process of classification itself.
2.5
FUZZINESS
Archaeological records typically include a great deal of uncertainty of various sorts. These range from
stated uncertainties in the records themselves, through implicit incertainties in its elements (eg, a
location), to inferred uncertainties arising from our assessment of the source (see also next section). The
13 -
11
V A N L EUSE N : P A T T E R N T O P ROCESS
database framework must be able to record the type and degree of uncertainty associated with any
element. Fuzziness is mainly applicable to the spatial, chronological, and interpretation variables.
If a data element within an archaeological record has an inherently probabilistic nature (eg, the
identification of a skeleton as ‘male’or ‘female’,the assignation of a flint tool to a particular period), then
the database framework should accommodate this by using a ‘fuzzy’data type (Crescioli et al. 2000) and
by incorporating rules about how to deal with varying degrees of fuzziness. Probably the main use for
such a data type would be in the description of the location of an object or assemblage, since any stated
location can be thought of as being the central value of a Gaussian probability curve at that location. One
can envision that sites whose locations are not recorded with a 95% confidence to within 50 m of the
stated location will be excluded from certain types of sensitive queries.
SPATIAL FUZZINESS
Spatial fuzziness, or the probabilistic nature of recorded geographical locations of archaeological
observations, and its formal representation have been a concern especially since the advent of GIS. Harris
and Lock (1992:118-120), for example, provide an early discussion of spatial error and fuzziness in the
context of the representation of archaeological records in a GIS.
For the purposes of a regional compilation, all source observations have locations which are known with
a greater or lesser precision and accuracy. The relevant database table could contain fields on whether the
observation concerns a point, line, or area, on the original coordinates, projection and datum under which
the element was mapped, and on the scale on and method by which the element was mapped. From this
information could then be derived other database fields containing the most probable location of the
observation in the current coordinate system, and a measure of the uncertainty of that location.
Archaeological co-ordinates may be available in any of several projections (among them geographic and
UTM) and co-ordinate systems (for the RPC project: ED1940, ED1950, and WGS1984). The original coordinates in this table, provided by the source, may therefore have to be transformed to some standard
projection and datum – for the RPC project, this standard is the old Italian national grid system under the
Gauss-Boaga projection, datum ED1940. In addition to the spatial location, observations may have one
or more administrative locations. Inasmuch as these are relatively stable, they can be derived from
appropriate GIS map layers (regione, province, comune, sheet numbers and names of the IGM 25V and 50
map series). However, the location of older observations will often be related to a local toponymic such
as a field or house name, or to a property (“on the land of … ”); this information must be stored in the
table as well because it may not be possible to convert it into a spatial location in a reliable manner.
Finally, metadata fields must be added to the table in order to record information relevant to the spatial
accuracy of the observation. Foremost among these is the mapping scale (referring to the scale of the
available input maps; other scales may exist but be inaccessible to us), but indications regarding the
mapping method may also be available from the source.
A single pair of co-ordinates indicates a site centroid; the site may have the further spatial attributes of
size (radius, diameter), spatial precision, and spatial fuzziness. Alternatively, archaeological remains may
be represented cartographically by vectors in the form of lines (eg, stretches of road, wall, or drainage
ditch) or polygons. In the case of a line vector, metadata may be needed to record the width of the feature
being represented; in the case of a polygon metadata are needed to record the nature of the feature (eg,
whether it represents a site core, a site with ‘halo’,an agricultural field, or a survey grid unit). In both cases
metadata on precision and fuzziness remain necessary.
RPC interpretations at all levels of aggregation must have spatial characteristics as well, in order to be
used and analysed in a GIS environment. These spatial characteristics must be constructed (either
manually or automatically) out of the spatial characteristics of the constituent observations8. Since spatial
8
And their fuzzy properties must also be inherited!
13 -
12
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
nearness is the single most important reason for aggregating observations into higher-order
interpretations, the process of interpretation must include an assessment of the stated spatial location of
an observation, and this may lead to a new metadata field containing the spatial accuracy of the
observation. This will usually be related to the mapping method used by the source – for example,
location sites by eye on 1:25,000 maps in areas where few topographic controls such as buildings and field
boundaries are available can result in errors of up to several 100 meters.
TEMPORAL FUZZINESS
The chronological attributes of source observations come in a huge variety of terms and classifications,
many of which need historical interpretation. Again, source observations must be deconstructed into
elements potentially belonging to different periods within the classification used by the source. The dating
evidence presented by the source must be recorded as well, for subsequent interpretation.
Table D: OBSERV_PHASES
Obs_ID
1
2
Period
Hellenistic
Bronze Age
Evidence
Presence of BG
Guess, based on description of fabric as thick, red and with sandy temper
Interpreting these data requires us to transform the variety of dating systems and terms used into a single
dating standard, as well as to assess how reliable the data in themselves are. Since typo-chronologies tend
to diverge as the distance between the observations increases, and a single period term may have different
chronological significance depending on where an observation is made, period terms used by the RPC
project are valid only at the regional scale. Table 5 below presents the two chronological systems used in
central and southern Italy against an absolute time-line. In order to effect supra-regional chronological
description, we are forced to use either very broad periods or absolute dating. However, we do not want to
lose the fuzzy nature of period terms – in terms of absolute date, ‘Hellenistic’carries a probability
function that allows overlap with the subsequent Roman period – and must therefore devise a system of
fuzzy dates.
Table 5: comparison of chronological systems in use in south Lazio and southern Italy.
Abs date
Salento
S Lazio
-11
-10
-9
BF / EIA
BF
EIA
-8
-7
IA
MIA
LIA
-6
Arc
Arc
-5
-4
Clas
p-Arc
-3
-2
-1
eHel LHel/lRep
Republican
+1
+2
Imperial
Imperial
+3
Interpreting the dating evidence presented by a source is often extremely difficult, and so various
metadata fields are needed to record the nature and amount of uncertainty associated with any period
assignment. The nature of the dating evidence can be stored in one such field. A default ‘source reliability’
might be used to indicate our confidence in the general quality of the periods assigned by the source; in
specific (older) cases we may be able to adjust source terminology to modern usage; and targeted
fieldwork may enable us to assign probabilistic periods to source observation (see especially the reinterpretation of the Quilici dates for Hellenistic-Roman sites in the Sibaritide, chapter 12).
2.6
CONCLUSIONS AND FURTHER WORK
The power (and weakness) of an RDBMS lies in the web of relations between the tables described above.
One ‘site’can have many phases, each of which can have many features; conversely, a feature can belong
to several phases. Usually one source will contribute many observations, but there may also be several
sources contributing information about the same site9. In other cases, it may not be certain that these
9
Not about the same observation, which is equivalent to a ‘visit’and represents a unique episode.
13 -
13
V A N L EUSE N : P A T T E R N T O P ROCESS
sources are about the same site. Several observations by the same or multiple sources may become part of
one interpretative entity, but they may also become part of several ‘competing’interpretations.
Where-ever possible, the content of the database fields described above should be entered in the form of
agreed term lists or ‘authorities’, which are not part of the web of relations. The table structure outlined
above is itself documented within the database as well, by means of another unrelated table containing all
table names and full table descriptions in Word OLE objects. The relations between the tables are
specified in the ‘Relationships’window of MS-Access, and are themselves also documented in a table.
Much has been written above about the variable, and often lacking, quality of the data that together form
the archaeological record for any region. Broadly, there are two ways of avoiding the issue: either records
containing vague, inaccurate, or incomplete data are excluded from all analysis, or all such records are
included in all analysis without any regard for the implications of data quality concerns. The former
option is attractive because it would considerably simplify the database design and the interpretation of
source data, leaving only relatively high quality data to work with; but the down side is that it may result in
an unacceptably high percentage of records being rejected. The latter option is also attractive because
ignoring problems with data quality simplifies the database design, data entry, and querying, but at the
same time it might lead to an unacceptable ‘pollution’of the results of such queries. The alternative
pursued by the RPC project is therefore to include all records and to implement appropriate means of
dealing with data quality issues.
FURTHER WORK ON THE RPC DATABASE
For the RPC database to become operative, the database design has to be finalised and implemented, the
process of data entry and interpretation must be completed, and the project’s archaeological questions
have to be translated into database queries. This work, once finished for all three RPC study regions, will
form the basis for much of the final project synthesis which is currently in preparation. Initially, many of
the design principles discussed above will be implemented partially or not at all because a full
implementation would require more time than is warranted for the limited goals of the RPC project. The
final database design (tables and relations) will therefore be much simpler than was suggested above.
Much work still needs to be done to establish and test authorities for the chronological and typological
classification of source data and interpretative constructs, especially on implementing their ‘fuzzy’
properties such as the date range associated with the various diagnostic materials. Once stable
classification criteria have been worked out, the process of data entry and interpretation is expected to
progress fairly rapidly. However, no fuzzy GIS operations have been defined as yet to make use of the
fuzzy data types that will be part of the database. More generally, the work of constructing queries that
make use of the metadata information that will be generated for the database has not yet started, and this
is where, ultimately, the effort of creating the database will have to pay off.
3
COMPARING DATA SET S OF THE PONTINE REG I O N
3.1
INTRODUCTION
Tobler’s (1970) first law of geography reads “everything is related to everything else, but near things are
more related than distant things.” It follows that the problems and potential of interregional comparison
can perhaps best be approached by tackling the lesser task of intra-regional comparison first. Beginning at
the largest possible scale, I will first attempt to compare several datasets resulting from surveys within the
same landscape unit (the colluvial slopes in the northern section of the Lepine margin, see section 3.2
below). This is followed by a comparison of data across a much wider area and multiple landscape units,
covering most of the Pontine region (section 3.3 below). Ultimately, these comparisons have three
objectives: firstly, to explore approaches to the formal comparison of archaeological data sets; secondly, to
13 -
14
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
review the settlement and land use history of the Pontine region in the light of all available data; and
thirdly, to review the effect that research methods have had on the results and interpretations of these
surveys. The following quantitative and qualitative properties will be used in my exploration of formal
methods for comparing archaeological data sets at a regional level:
n
n
n
n
n
Total number & density of sites, number & density per period, ratio of count & density
per period
Absolute and relative composition of assemblages, either within sites or within other
geographical units
Site size and rank-size distribution
Site continuity/discontinuity by period
Site location characteristics
Given the variability in archaeological recording techniques, a highest common denominator approach
implies that we must base our comparison on ‘sites’,perhaps comparing their sizes, densities per km2, and
rank-size distributions across regions and the development in their size (demography) and features
(function) over time. All of these comparanda must be sufficiently well-defined for us to feel confident
that we are comparing like with like. However, recent surveys yield more and more ‘off-site’data, and it
may also be argued that we cannot afford to ignore such data, particularly since these intensive surveys
often provide the information we need for ‘source criticism’of the less intensive survey data. In these
cases we might attempt to construct and compare ‘off-site’density histograms summarising the intensity
of ‘landscape use’per period as inferred from off-site finds densities. However, in view of the widely
acknowledged variability in the amount of material recovered in any single survey pass, such an approach
might reduce us to studying and comparing the relative abundance of the various material groups rather
than their absolute number or density. Such an approach, although it appears feasible, lies outside the
scope of this chapter.
DATA SETS
Let us now have a more detailed look at the available data sets for the Pontine region, introduced in
section 2.2 above. A first coarse classification of the available survey data into four qualitatively distinct
classes can be made on the basis of whether the survey was or was not site-oriented, and whether data
collection was or was not systematic (see figure 2).
Large sections of the Pontine region have been surveyed in an unsystematic site-oriented manner. The
topographic surveys of the Forma Italiae series are good examples of this; five sheets have so far been
published for the Pontine Region (see figure 1). In what follows, the topographic survey covering the all
of the Cora sheet plus the southern half of the adjoining Artena sheet of the IGM 25V map series (a total
area of 146 square km; Vittucci 1968) will be used as an exemplar because of its overlap with the RPC
intensive survey area near Ninfa.
Site oriented
yes
no
Site survey, eg Norba,
Lanuvium
Off-site surveys, eg Fogliano, Sezze
Sampling surveys, eg Agro Pontino project
no
Systematic
yes
Topographic surveys,
eg ‘forma italiae’
Incidental finds
Figure 2: classification of survey types
Two types of systematic, none-site oriented surveys have been conducted in the Pontine region: extensive
sampling surveys and intensive off-site surveys. The Agro Pontino Project surveys, by the University of
Amsterdam, ran between 1982 and 1989. In three of these seasons (1984, 1986, and 1988) the surveys
were conducted according to a probabilistic sampling design. Only preliminary analyses of the APP data
are available, and the ceramics in particular were not recorded consistently across all campaigns (Voorrips
13 -
15
V A N L EUSE N : P A T T E R N T O P ROCESS
et al. 1991). In the field, only very dense and localised artefact scatters were recorded and sampled as a
unit; subsequently, further ‘sites’were defined by administrative grouping of adjacent agricultural fields
but the details of this procedure are unclear (Koot 1991:124-5). The distribution maps presented by Koot
must therefore be interpreted with caution, and are best understood in the context of the five vegetational
zones into which the Pontine plain was subdivided (see figure 3).
Intensive and systematic off-site surveys were conducted in the period 1994-1999 by teams from the
University of Groningen under the direction of Prof. Attema in the vicinity of Sezze (Lepine footslopes)
and Fogliano (coastal landscape, see chapter 10 for details). The Sezze area had been chosen as an
example of a landscape that had undergone 4th century BC rural colonisation, while the Fogliano area had
been selected as a typically 'marginal' area. Finally, systematic intensive site-oriented rural surveys were
carried out by other teams directed by Attema near Lanuvium in the Alban hills (1995), and Norba (1995)
and Ninfa (1998; see chapter 9 for details) again in the Lepine footslopes. Attema chose the volcanic
ridges east of Lanuvium as an example of a developed Latin landscape being Romanised in the middle
Republican period, and the Norba and Ninfa areas as examples for similar Roman colonisation of a
'marginal' landscape of Lepine footslopes. These surveys provide the first quantifiable data on the preRoman landscapes of south Lazio.
3.2
COMPARISON WITHI N THE SAME LANDSCAPE UNIT: THE NORTHERN COLLUVIAL
SLOPES
The Lepine footslopes between Cori and Norma are part of the northern colluvial landscape unit, and
were included in the southeastern part of the area surveyed by Vittucci sometime before 1968. In the
1930s a major canal (formerly known as the Canale Mussolini but now named Canale delle Acque Alte)
was built parallel to the Lepine slopes in order to collect waters from it and transport them to the
Tyrrhenian sea near Torre Astura. The works related to this, and the wide ‘footprint’of the canal itself,
have served as a convenient boundary to the nearby Ninfa98 and Norba95 surveys. Both took place on
the footslopes of the Lepine mountains directly northwest of the Late Iron Age /Archaic proto-urban
centre of Caracupa/Valvisciolo, but the two survey areas are physically separated by a 400 m wide loop in
the Canale della Acque Alte where this has followed the elevation contours to abut directly on the
limestone slopes of the Lepine scarp. Here, the higher ground formed by the two fluvio-colluvial fans
emanating from the direction of Cori (“Vigne Vecchie”) and the mouth of the Fosso della Valle must have
originally given way to the lower-lying land of the Pontine plain proper. Little or no colluvial soil covering
has built up, which may have been the main reason why a subrecent limestone quarry was situated here as
well. The whole area is dominated by the Norba promontory, and in the later medieval period was the site
of the town of Ninfa.
The Norba survey (King 1995), intended to study the Roman ‘colonial’villa landscape near the colony of
Norba, covers part of the Lepine footslopes between Caracupa and Norba itself. The 74 hectares covered
by this survey were augmented in 1999 by a further 9 ha in the area known as ‘Pellicio’, extending the
survey just across the Canale Mussolini. The Ninfa survey had also originally been intended as a simple
extension of the Norba survey in a northwestern direction along the Lepine footslopes, but quickly
became more oriented toward intensive off-site recording methods as it became clear that pre-Roman
material occurred in scatters in between the Roman villa sites, as well as on them. The total area of 27
agricultural fields surveyed intensively in 1998 was 91 ha. In 1999 a number of these fields and sites were
re-visited in order to collect additional data.
SETTLEMENT DATA
The Norba95 survey resulted in the discovery of 11 sites (King 1995). Only one of these (a Late Iron Age
hut) dates to the protohistoric period. Archaic (6th century) material was found on four sites, each of
which was also occupied in the Roman Republican period (no evidence from the intervening postArchaic period was found). Seven more sites beginning in the Roman Republican period were also found,
bringing the total to eleven, of which three were ‘platform’villas. Only three of the 11 Republican sites
show evidence of continued occupation into the early Imperial period.
13 -
16
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
The Ninfa98 survey resulted in the mapping of approximately 13 new sites in addition to the 12 already
described by Vittucci10. These new sites range in date from the late Iron Age to the early Imperial period,
so that not only was a complete pre-Roman landscape added to the classical landscape known to the
topographers, but also the density of Roman settlement was proved to be nearly double that recorded in
the topographic survey.
Table 3a: site counts (#), percentage (%), and average density per century/km2 (d)
per period for the Cora, Norba, and Ninfa surveys. Duration in centuries: EIA: 2, LIA: 1, ARCH: 1,
PARCH: 1.5, REP: 3.25, EIMP: 1.25, IMP: 2.
Cora 1968
%
d
6
31
2.0
1
5
0.9
12
64
19
100
#
Early Iron Age
Late Iron Age
Archaic
Post-Archaic
Roman Republican
Early Imperial
Full Imperial
Indet. Roman
totals
#
Ninfa 1998-9
%
d
2
4
1.1
17
34
9.3
12
10
7
1
1
50
24
20
14
2
2
100
8.8
3.4
6.2
0.5
-
Norba 1995
%
2
10
4
20
11
55
3
15
20
100
#
d
2.4
4.8
4.1
2.9
-
Table 3b: site counts (#) and percentage (%) per site type for the Cora, Ninfa, and Norba surveys.
hamlet
scatter
Building scatter
villa
Platform villa
Infrastructural
terracing
Graves/tombs
totals
Cora 1968
#
%
5
31
1
6
3
19
6
38
1
6
16
100
Ninfa 1998-9
#
%
1
5
10
50
4
20
3
15
2
10
20
100
Norba 1995
#
%
3
25
3
25
3
25
3
25
11
100
The relative absence of Late Iron Age and Archaic sites in the Norba95 survey area, as opposed to the
Ninfa98 area, may be conceived to be an expression of the fact that settlement in this period was
centralised at Caracupa/Valvisciolo (Attema 1993a:122 suggests an identical explanation for the volcanic
landscape between Cori and Cisterna di Latina); but on the other hand it may be caused by the fact that
the survey only aimed to locate Republican (platform) villas in the area. The one Late Iron Age hut site
found contains evidence for both weaving and storage but was not continued in the Archaic. Indications
that an Archaic sanctuary may have been present at the edge of the fluvio-colluvial fan in the Pellicio area
can be interpreted in the context of the liminality models suggested for Etruria (see chapter 15 for more
on this).
Some additional observations for this colluvial landscape unit can be gleaned from Attema (1993,
forthcoming). In an extensive site-oriented survey conducted in 1987, one of Attema's transects ran
through the volcanic and colluvial landscape between Cisterna di Latina and Cori, revealing continuous
occupation from the early Iron Age onwards. However, he only discovered two sites in the Lepine
footslopes: site 5 ‘Torretta” was a large scatter dating from the Archaic into the early Imperial period, and
site 6 “Grotticelle” was an Augustan villa with prior rural use since the post-Archaic period (Attema
1993a:117-22, 269-71). A second transect ran almost north-south and covered several fields in the
alluvium/colluvium below Valvisciolo/Caracupa. High Archaic finds densities here were related primarily
to the settlement and necropolis of Caracupa, while material on the lower part of the alluvial fan was
difficult to interpret because of post-depositional sedimentation and earth movement (Attema 1993a:122-
10 A precise number cannot be given because the identification of sites with those mapped by Vittucci was uncertain in several
cases, and some Vittucci sites could not be relocated. For full details, see chapter 9.
13 -
17
V A N L EUSE N : P A T T E R N T O P ROCESS
33 and catalog 1b). Finally, intensive surveys were also conducted in 1988 by Attema on the ‘proto-urban’
site of Caracupa/Valvisciolo itself and, slightly the east on the plateau of Contrada Casali near Sermoneta
(Attema 1991a, 1991b, 1993a:157-80, 1993b). All of this supports the notion that a continuous developed
Archaic settlement system was present along the rim of the Lepine mountains in the northern colluvial
landscape.
The lack, in the Norba95 area, of any evidence from the post-Archaic period (500-350 BC) at sites
occupied in both the Archaic and Republican periods may be interpreted as the outcome of – literally –
‘unsettled’circumstances due to the Roman-Volscan conflict. Until as late as 330 BC, the Norba territory
was raided by Privernates, and it is quite possible that defenceless farm sites were abandoned early on.
However, post-Archaic ceramics in south Lazio are generally of a very undiagnostic nature, hampering
consistent recognition in surveys, so they could not be used to establish the presence of post-Archaic sites
in 1995. Fabric studies were begun to address this problem, and by the time of the Ninfa survey postArchaic fabrics were tentatively identified (Attema & Oortmerssen 2000). Rather than settlement
discontinuity, the Ninfa data demonstrate a gradual decrease both in the number and the density of sites
from the full Archaic into the Republican period, which could be interpreted as a gradual consolidation
into a smaller number of more substantial farms.
Rebuilding in recognisably Roman architectural style and materials, if not re-occupation, may have
occurred sometime in the 3rd century BC, although the datable finds generally point to a 2nd century date
for the Roman villa sites. The evidence appears to support the presence of a late Republican site hierarchy
in two levels as first proposed by Attema (Van Leusen 1998). A few platform villas, located in clearly
higher elevations and on steeper slopes away from agricultural land, constitute an upper rank of
settlement; the larger number of other (platform or non-platform) villas constitutes the ‘standard’rank.
The apparent decline in settlement density setting in in the early Imperial period fits in well with general
trends across the Pontine region (eg, Attema et al. 2001), and may be connected to the progressive loss of
Roman markets for grain and olive oil.
CONCLUSION
Comparing the results of all the available surveys, it becomes clear that less intensive survey results in the
discovery of a predominantly classical landscape, because sites from this period are the most obtrusive
(large and dense scatters containing both tile and ceramics, standing architecture, many diagnostic wares).
Nearly all Cora68 and Norba95 sites were clearly visible for these reasons. A further hurdle to the
comparison of all three data sets is presented by the apparent use of incongruent chronological and
typological classifications, which studies of the ceramic fabrics involved are now going part way to
resolve.
3.3
COMPARISON ACROS S LANDSCAPE UNITS: T HE PONTINE REGION
We will now extend the comparison of data sets to include both landscape units directly adjacent to the
northern Lepine footslopes, and those further afield in the Pontine region.
THE VOLCANIC LANDSCAPE OF THE ALBAN HILLS AND THE LEPINE SLOPES NEAR CORA
Most of the Cora volume of the Forma Italiae series covers the easternmost part of the landscape of low
ridges and valleys of volcanic origin typical of the Alban hills, the alluvial / colluvial fan at the apex of
which sits the town of Cora itself. To a lesser extent, it also covers accessible parts of the Lepine
mountains and its colluvial footslopes. To the west of this topographic survey area, Attema's intensive
site-oriented Lanuvium survey covered a few square kilometers of several very similar volcanic hill
13 -
18
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
systems. Of the 201 sites recorded in the Cora68 area, 134 were dated and 142 were given a site type
interpretation (see table 4)11.
Table 4a: Cora68 and Lanuvio95 sites by period
Number
discontinued
new
Number
continued
Density per
century / km2
Iron Age
Archaic
Post-Archaic
Republican
Imperial
Medieval
3
4
68
21
1
1
2
61
21
3
2
66
12
1
2
2
9
0
0.02
0.02
0.14
0.04
-
Roman
54
3
53
1
-
Density per
century / km2
Count
Cora 1968
*counts based on total ‘stringsquare’samples; 17 out of 40 IA sites are based on less than 5 sherds.
Lanuvio
1995*
40 (incl eArch)
3.4
46
5.4
45
4.1
-
-
Table 4b: Cora68 site count by type (additional uncertain attributions in parentheses)
Type
Habitation
Infrastructure
Water
management
Other
Subtype
Building
Farm
Farm / Villa
Villa
Large villa
Castle
Road
Bridge
Cuniculus
Fountain
Cisterna
Tomb, necropolis
Quarry
Terracing
count
3
9
2
40 (7)
7
1
23 (2)
2
9 (2)
1
14 (1)
2 (2)
1
14
The first thing that one notices is the almost completely classical nature of the landscape of roads and
villas mapped and dated for the most part to the period from the late 2nd century BC to the late 1st
century AD (Vittucci 1968:17). 133 out of the 134 dated sites are Roman; only one site (of Archaic to
post-Archaic date) has no Roman successor. Clearly, this must result from a combination of the high
obtrusiveness of, and high interest in, remains of the classical period - very much in the tradition of the
earlier Forma Italiae. The assignment of dates on the basis of sometimes very summary evidence is
hazardous, and certainly introduces a bias toward the classical periods. The protohistoric periods and the
post-Archaic go almost unrecognised – only 5 sites were recorded - but a glimpse of the presence of the
latter period on many later Roman sites is provided by Vittucci’s (1968:17) mentioning of ‘ceramica
d’impasto rossiccio’. Closer dating of sites within the Roman period on the basis of the architectural
evidence remains problematic, since only some of the opus (walling types) have clear chronological
implications – and this applies a fortiori to the four 'polygonal' styles of terracing. 53 out of the 133 Roman
sites (40 %) could therefore not be assigned to any particular part of the Roman period. Of the sites that
could be assigned to a sub-period, 68 were assigned to the Republican period and 21 to the Imperial
period, but continuity from the previous period was only attested in 9 of the latter. The low figures for
continuity across all periods strongly suggest that sites were regarded by the researcher as single-period
entities, and little notice was taken of evidence of the presence of previous or later periods.
Depending on the type and amount of evidence, farm buildings are classified as ‘building’,‘farm’,‘farm /
villa’, ‘villa’, and ‘large villa’. Altogether, there are 68 such sites. An overall predilection for structural
remains, coinciding with the classical preference already mentioned above, is one of the main
11
After re-interpretation of the Vittucci data by Drost (1996) and Attema.
13 -
19
V A N L EUSE N : P A T T E R N T O P ROCESS
characteristics of this data set. In addition, the survey was conducted according to topographic tradition
with a view to the reconstruction of the viabilità antica, and mainly concentrated along a limited number of
roads and paths (Vittucci 1968:19); undoubtedly, this must have contributed to the creation of spurious
spatial site patterns in the form of linear clusters. Besides a large number of sections of Roman road
revetment, other structural remains of note relate to water management (cuniculi and cisterns).
In the 1995 Lanuvium survey, Attema abandoned the total collection methods developed in earlier
surveys in favour of more efficient artefact counting in the field and collection of diagnostics only. This
has had the unfortunate effect of making it impossible to check his results by subsequent re-analysis of
raw data. In preliminary analysis (Attema & Van Leusen, forthcoming), a total of 52 sites were defined in
an area of 339 ha for a total site density of 13.0 / km²; table 4a shows the chronological distribution of
these sites. Attema concluded that the area “showed continuity in ceramic production and supply from
the Iron Age to the Roman period” and dates the incipient rural infill of the Alban hills area near
Lanuvium to the 8th and 7th centuries BC, with an intensification occurring towards the end of the 7th
century (Orientalising period; Attema 2000:424). This rather diffuse infill is transformed (‘crystallises’)
into a fairly dense pattern of Archaic farmstead scatters in the 6th century.
It should be noted here that the fabric provenance data supplied by Attema (2000:422 and fig. 8) indicate
that the survey methods employed in the Lanuvium survey were not suitable for the recognition of
diffuse scatters, and also that red firing (early) fabrics were much more likely to occur in such diffuse
scatters than were the later orange or pale firing fabrics. In other words, the site counts per period
significantly underrepresent the number of 7th and 6th century BC sites. We again see that both the
settlement chronology of the Lanuvium area as a whole and the density of settlement in individual
periods, differ so radically from those of the adjacent Cora68 survey area, that a comparison only serves
to underscore how much remains unrecorded even in areas that were investigated by professional
archaeologists.
THE LANDSCAPES OF THE PONTINE PLAIN
Extending our comparison to landscape types further away, and often quite different from, what has been
considered the traditional homeland of the Latial tribes, we now turn our attention to the Pontine plain
proper. The geological formation of the Pontine plain is determined by an active horst and graben system.
The geomorphology of the former is sculpted by the recession, in several stages each leaving coastlines
and lagoons at diminishing elevations, of the sea; the gradual sinking of the latter has resulted in a net
depositional environment characterised by alluvial and colluvial deposits emanating from the Alban and
Lepine hinterland, and an accumulation of clays and peats in the remaining parts. On the basis of the
different soil types formed in these deposits, the plain has been divided into five vegetational zones by
members of the Agro Pontino project. These are convenient landscape units for current purposes as well,
against the background of which the archaeological data sets for this region can be compared. Figure 3
shows the distribution of findspots by period; table 6 presents the findspot count by period. The term
findspot is used advisedly because, as our source (Koot 1991:124-5) is careful to mention, the collection
and recording methods employed by the Agro Pontino Survey were geared to highly diffuse and extensive
lithic distributions and do not allow the definition of meaningful ceramic sites.
13 -
20
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
Figure 3: Vegetational zones and distribution of APP findspots in the Agro Pontino (after Loving et
al. 1991, fig 2 and Koot 1991, fig 1-5). Top left: vegetation zones. A-aeolian sands, B-lagoonal
clays, C-peats and humic clays, D - colluvium, E-beach ridge-lagoon mosaic of sands and clays. Top
right: distribution of protohistoric ceramics. Triangle - Neolithic, diamond - Bronze age, circle - Iron
age. Bottom left: distribution of Archaic findspots. Grey - early Archaic, red - late Archaic. Bottom
right: distribution of Roman findspots. White – Campanian, red – Samian.
INTENSIVE SURVEYS NEAR SEZZE AND FOGLIANO
In the Sezze survey, initially some 20 mainly Roman sites were recorded in an area of 83 ha; in subsequent
analysis of the off-site ceramic distributions (Feiken 2000) several more sites and a much greater time
depth to the existing sites were defined. With some difficulty (recorded sites that fall outside the
boundaries of the surveyed fields had to be removed first), we can calculate an overall site density of 24.1
per km2 for Sezze. Two previous extensive surveys in the Sezze area (Zaccheo & Pasquali 1972, PRP94)
also recorded an almost exclusively Roman (post-Archaic to Imperial) landscape before analysis of the
1995 survey data showed the protohistoric roots and Archaic predecessors of what had appeared to be a
virtually pristine landscape colonised by the Romans in the 4th century BC (Feiken 2000).
13 -
21
V A N L EUSE N : P A T T E R N T O P ROCESS
Table 6: Number of APS findspots per period (from Koot 1991).
Count
1
6
8
48
Veget. zone
B
A, B, E
D, E
All (few on B)
Remarks
Ceramics recognisable only by their association with lithic scatters
3 of these occur in the Sezze colluvium
Especially dense in the northern colluvium
49
all
Post-Archaic
Republican
37
125
on
C
Early Imperial
36
all (very few
B, many on
and D)
All (very few
B, many on
and D)
No change in the spatial distribution; 12 findspots are continued from
previous period
Based on the presence of ‘local ware’; no distribution map given
th
nd
Based on the presence of Campanian ware, mostly late 4 to 2 c.
BC; number of findspots based on estimate, not a count
on
C
Based on the presence of Samian ware; distribution similar to that of
previous period
Neolithic
Bronze Age
Early Iron Age
Late Iron Age /
Archaic
Late Archaic
Traces of human occupation of the Fogliano survey area were present in an unbroken sequence from the
Middle Palaeolithic onwards. The earliest ceramics, which are of a friable reddish-brown fabric with a
sandy temper, probably date to the late Bronze and early Iron Ages, and by advanced Iron Age (c. 800
BC) ceramic scatters occur in all major geomorphological subunits of the landscape. By the beginning of
the Archaic site density had doubled and essentially all of the coastal beach ridge and lagoonal landscape
was in use, excepting only the clayey hinterland. In the relatively restricted spaces afforded by the long
and thin beach ridges, no major changes in settlement and land use patterns are apparent throughout the
post-Archaic and Republican periods, but the picture was dramatically different for the larger and more
fertile landscape of aeolian sands to the southeast. Here the number of settlement sites doubled in the
post-Archaic, and again more than doubled in the late Republic. The growth of a rural village here has
been linked to the economic pull by a small number of large maritime villas which controlled industrial
fish farming in the lagoons and along the coast. As in the Sezze area, there is little or no evidence to
indicate that any of these settlements continued into the middle Empire (full details available in chapter
10).
Table 7: site counts (#), percentage (%), and density per km2 (d) per period for the Sezze and
Fogliano surveys. *: includes late Republican sites
#
Late Bronze Age
Early Iron Age
Late Iron Age
Archaic
Post-Archaic
Republican
Early Imperial
Full Imperial
totals
5
6
~20
Sezze 1994
%
d
9
6.0
10
7.2
~34
~24.1
~20
~34
~24.1
7
58
12
99
8.4
#
Fogliano 1998
%
d
2
3
0.7
9
12
14
12
24 *
4
77
12
16
18
16
31
5
101
3.3
4.4
5.2
4.4
8.9*
1.5
Table 7 appears to show that there is not much difference in the relative site counts by period between
the Sezze and Fogliano areas, until one realises that there is a difference of two centuries between the
respective dates for Romanisation – around 300 BC in the Sezze area, around 100 BC in the Fogliano
area12. If one looks at the densities (counts per km2) per period, the much higher intensity (roughly six-
The late dating of Republican material from Fogliano raises a problem in the interpretation of the post-Archaic material in the
Fogliano area. If the post-Archaic date is correct - if it really is 5th and 4th century - then a middle Republican settlement hiatus
opens up; if the post-Archaic material could in fact date throughout the 5th to 2nd centuries BC, then the site density for this
period drops by a factor of more than 2. Moreover, we must then face the implications of similarly wide dates for post-Archaic
material in other surveys.
12
13 -
22
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
fold) of settlement and land use in all periods except the late Republican / early Imperial in the Sezze area
becomes apparent.
INTERPRETING THE DIFFERENCES AND SIMILARITIES BETWEEN THE DATA SETS
If we now compare the data sets introduced and discussed above, we can to some extent use the later,
more intensive, surveys to re-interpret the earlier topographic and extensive survey data. For example, in
the light of the protohistoric finds made in the Ninfa and Fogliano surveys we can now interpret the very
low number of Bronze Age and Iron Age findspots reported by the APP survey to mean that a lowintensity rural landscape had already formed throughout the habitable Pontine area, excluding perhaps
only the wet and peaty sections of the Pontine graben. In the Iron Age, but certainly by the Archaic
period, a hierarchically organised social and settlement system must have existed along stretches of the
colluvial landscape along the Lepine margin. Only the relative dearth of good farming land there and in
the coastal landscape may have prevented these hierarchies from developing at the same pace as in the
nearby Alban hills.
Similarly, the spread of early and late Archaic findspots from the APP survey confirms that simple farms
spread across the landscape in this period not just in the traditional ‘core’areas, or near the known protourban settlements, but all along the ‘marginal’Pontine coast up to about 10km inland as well. It is quite
possible that larger settlements, up to proto-urban rank, remain to be discovered here; in any case it now
seesm legitimate to assume that demographic expansion took place here as much as in the Alban area.
Although Koot (1991) did not publish a distribution map of ‘local’wares for the APP survey, he does
note that its distribution indicates that the Volscan wars did not lead to a depopulation of the plain – a
conclusion now supported by the post-Archaic site densities reported by the Ninfa and Fogliano surveys.
However, dating evidence is still weak for this period, and small shifts in the attribution of some fabric
and ware groups could still cause significant changes in the settlement history of the Pontine region for
this period.
It is only with the introduction of identifiable Roman ceramics, especially when these occur in
combination with architectural remains, that we can begin to compare the older topographic surveys with
the ones already mentioned above. Here, the choice of researchers to study areas centering on historically
known urban sites is unfortunate since it has served only to create an impression that landscapes
colonised culturally, if not physically, by the Romans surrounded, and centred on, these towns. Whilst this
impression could conceivably be correct, it is equally likely that it is based on a spurious pattern caused by
research and discovery biases – e.g., the tendency of topographers to cling to existing infrastructure
automatically leads to higher site densities near the nodal points13 (for more on such biases, see chapter
4).
For the early Republican period, high density rural settlement appears to concentrate on the Lepine side
of the Pontine plain which, from about 350 BC, appears to have been colonised in the true sense, with
drainage and infrastructural works enabling the colonists to farm previously marginal areas, and local and
regional markets accessible along the Via Appia. The more than three-fold increase in site count reported
by the AP survey for the Republican period relative to the post-Archaic must be ascribed to this
colonisation of marginal land rather than to the colluvial part of the Ager Setinensis. The Fogliano survey,
whilst confirming the lack of early Republican (4th – 2nd century BC) findspots reported by the Agro
Pontino Project survey for the coastal landscape, resulted in the mapping of a dense late Republican (2nd
– 1st century BC) patterns of farms in an area where the latter survey recorded only a single findspot. It
may be hypothesised that these later farms would have occurred in clusters, tied economically and
perhaps socially to the presence of large maritime villa estates bordering on the coastal lagoons. In this
context Vittucci’s dating of ‘most’Roman rural villa sites in the Cora map sheet to the late Republican
The clustering of rural villa sites around the colonies of Cora and Setia reported by Attema (1993, fig 148) should, for example,
be re-examined in this light.
13
13 -
23
V A N L EUSE N : P A T T E R N T O P ROCESS
period is remarkable. One surmises that she dated Roman sites by a limited number of late diagnostic
features such as the use of opus reticulatum in walls, ignoring potential indicators for earlier (middle
Republican) phases. Nevertheless, if the late dates are at least partially correct, they should perhaps be
interpreted as a sign that Romanisation progresses only slowly in a society which already had a strong
native Latial structure.
All types of survey agree that a great (approximately three-fold) drop in site density occurred in the
Pontine Region by the end of the 1st century BC. This dramatic drop continues into the middle Empire,
by which time many areas contain no or very few sites at all. Differences in surveying coverage and
intensity again assume great importance by the end of this period and into subsequent late Antiquity and
early Middle Ages, as the chances of hitting scarce sites, and scarce diagnostic material within sites,
diminish.
SETTLEMENT DYNAMICS AND CORE PROCESSES
It now remains to connect these interpretations to the processes occurring in the Braudelian longe durée in
the Pontine region - centralisation, urbanisation, and Romanisation / colonisation.
If, as at the core of the Alban Hills, settlement centralisation in the Pontine region were to begin already
in the Final Bronze Age, then surveys in low-lying areas may not be appropriate to detect this and surveys
may have to be redirected to up- and highland areas. However, a better understanding of differential
circumstances of preservation and discovery of protohistoric fabrics must be developed before we can
discount the occurrence of lowland Final Bronze Age/Early Iron Age centralisation altogether14.
The study and comparison of the data from rural surveys in the Pontine region has little direct bearing on
the issue of (proto-) urbanisation, except insofar as this is put in perspective by the discovery of ‘hamlets’
(roughly 1 – 2 ha-sized scatters) in both the Lanuvium and Ninfa survey areas. Although the criterion on
which this identification is based - scatter size - is rather weak, a basic rank/size hierarchy of settlements
can now be hypothesised to develop during the Archaic not just in the Alban hills, but also in the alluvial
/ colluvial landscape along the Lepine mountains as far south as Sezze and quite possibly in other parts of
the region as well. Whilst copious evidence for the development of urban polities in the Archaic has come
from the historically known centres, lesser or 'truncated' polities will have centred on places such as
Valvisciolo/Caracupa and Cisterna di Latina. At this point it would be advisable as well to remind
ourselves of the probable existence of several ‘lost’towns in this same area, as reported by Livy (eg,
Longula, Polusca, and Corioli; Ab Urbe Condita II 33).
A future regional study of the composition of Archaic rural site assemblages might be able to establish the
degree of penetration of non-locally produced ceramics such as roof tiles, from which deductions might
be made regarding the existence of core-periphery relations between the Alban hills and adjacent Lepine
margin colluvium on the one hand, and the Pontine plain and coastal margin on the other. The
standardisation of Archaic ceramic fabrics indicates that craft specialisation had already begun in this
period; by the end of the 4th century pottery production had become centralised to such an extent that no
local ‘archaic’wares were produced anymore.
As has been made clear, the impact of the expanding Roman Republic within the Pontine region varied
both in time and in space, depending both on the claims and resilience of the local inhabitants and on the
accessibility and perceived value of the land. The intriguing fact that data sets for both the northern
colluvium (Norba95 and Ninfa98), the coastal landscape (Fogliano98-9) and, possibly, the Alban hills
(Cora68), appear to indicate that visible Romanisation in the form of simple rural villas (with or without
platforms) took place only by the late Republican landscape calls for further investigation. There is now
Attema’s recent research on what appears to be a series of middle to late bronze age industrial sites on the Tyrrhenian coast
between Anzio and Astura, possibly involving the exploitation of fish or salt, provides an additional and strong indication that
our understanding of protohistoric societies in south Lazio is far from complete.
14
13 -
24
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
little doubt that most Roman rural villas anywhere in the Pontine region, excepting the newly drained
marshlands, were built on the site of Archaic to post-Archaic predecessors, but still no evidence either for
or against the notion that native farmers might have been dispossessed in favour of Roman colonists. In
this context the correct dating of post-Archaic ceramics which may straddle a period of local settlement
discontinuity, assumes great importance. Certainly there are now few areas left where Roman rural
expansion was either contemporary with, or centred on, the colonial towns on the Lepine scarp.
3.4
FURTHER WORK
My attempts at comparison of archaeological data sets have tended to reveal more about the
methodological differences than about the relative settlement dynamics within the landscape units of the
Pontine region. A successful comparison of results, even from two surveys conducted by the same team
using the same general methodology, requires extensive standardisation of recording methods and
creation of metadata. It is not clear that there are good reasons for the extreme lack of standardisation of
survey methodology (an issue elaborated in Van Leusen et al., in prep.). Nevertheless, general trends now
appear to have been identified by independent surveys conducted over a period of 20 years, and these in
turn allow a re-interpretation of even older survey data:
• Late Iron Age/Archaic rural infill occurs not just in the Alban hills but in the northern colluvium (as
demonstrated by both the Sezze and Ninfa surveys and the APP) and – to a lesser extent – in the
sandy coastal landscape of the Pontine region as well (as demonstrated by the Fogliano survey). In
the latter area the overall density of Archaic farmstead sites appears to be rather lower, though, than
in the Lanuvium or Ninfa areas, where distances of two or three hundred meters between sites are
not unusual.
• Substantial rural infill appears in the middle Republican period in both the colluvial and the peaty
landscape of the Pontine graben (APP and Sezze surveys), strengthening the historical and air
photographic evidence for a Roman colonisation in this area. Other landscapes remain largely
unaffected until the late Republican period, though apparently for different reasons (Lanuvium and
Ninfa surveys, Fogliano survey and APP).
• A drastic (three- to six-fold) reduction in the number of inhabited rural settlements is seen in all data
sets in the early Imperial period.
Future research in the Pontine region should take these case studies into account both in its choice of
landscape unit and in its methodology. Weaknesses in the core classifications employed for regional
archaeological interpretations must be addressed first:
• Local ceramic chronotypology must be further developed through fabric classification (cf. Attema
2000, Attema et al. forthcoming) and seriation of survey assemblages, especially for the post-Archaic
period.
• Site type classifications, particularly for the pre-Roman periods, should be improved through a
programme of field tests involving surveys, trial trenches, and excavations at a representative sample
of site types.
• Surveys should address the current lack of data for what has traditionally been considered the
marginal parts of the ancient landscape; especially the up- and highlands.
• Loving and colleagues (1991) note that the original APP sampling design had been too optimistic in
its estimates of the number of fields to be surveyed for conclusions to be drawn from the sampled
area with the desired confidence level. Despite this and other problems with the interpretation of the
results of the APP surveys, the intensive surveys conducted in the various landscape units since the
early 1990s now provide a solid basis to revive the idea of a regional stratified sampling design. The
13 -
25
V A N L EUSE N : P A T T E R N T O P ROCESS
conduct of future survey campaigns within such an overarching strategy can result in a more efficient
use of limited resources and should generate more easily comparable data for the region as a whole.
4
CONCLUSION
The discussion of regional database design presented in section 2, and the case studies in intra-regional
comparison of survey data sets presented in section 3, reveal that there is an altogether surprising degree
of variation in the way archaeological data were collected, interpreted and analysed. These differences
occur not just between the different research traditions employed historically and, to some extent,
nationally, but also between data sets produced by the same team within the same region, and are likely to
be present in equal measure in the other RPC study regions. Hence, it is equally difficult to effect formal
comparisons between survey data sets.
At the current state of research, the discussion is necessarily concerned mainly with technical obstacles to
the comparison of the data sets presented above. Issues include cross-project differences in typochronological phasing, the use of different ‘guide fossils’and site typology, and the lack of definitions and
standards (cf. Van Leusen et al., in prep.). The single most important observation that may be made here
is, that the collection of data in none of the surveys discussed in this chapter was aimed at enabling
comparisons beyond the immediate survey area.
The general lack of standardisation and formal definitions is most apparent in the fact that, although
almost all surveys until very recently have reported their results in the form of site maps and catalogues, a
wide variety of informal criteria for the definition of sites were applied so that we are simply not
comparing like with like. Nor can finds densities be taken to be a better measure of land use intensity: as
re-surveying experiments have tended to show (see chapter 8), the obtrusiveness and visibility of surface
remains are extremely variable depending on material category and conditions of observation.
Consequently, site and find numbers of pre-classical (and presumably post-classical) landscapes tend to be
strongly de-emphasised relative to the classical landscape.
It appears that, if we want to make fruitful comparisons at any scale above the purely local, we must
conduct our surveys according to specified procedural and data standards, which include a prominent role
for quality control and source criticism and are codified in an appropriate database design along the lines
sketched in section 2 of this chapter. As has been made clear there, implementing such a regional
archaeological database is not a trivial matter, not just because of the wide range of data types that may
need to be included and the extreme variability in data quality and definition, but also because seemingly
uncontroversial decisions about its structure and authority lists may deeply affect the kinds of results
obtainable through GIS or database queries.
Supra-regional comparisons based on relatively obtrusive remains, if they properly account for differences
in research intensity, are quite likely the only type of comparison that can successfully be made at this
stage. Bintliff’s comparative study of classical ‘take-off’in Greece is a good example, but the case study
presented here in section 3 demonstrates that the appearance of a demographic ‘take-off’is not born out
by closer investigation. Many landscapes had already been cultivated, and were in the process of
developing regional hierarchical structures, as early as the Archaic period (6th century) before they came
under significant influence from outside powers, and we should be very careful drawing conclusions from
diachronic variations in either site or finds densities. A Republican rural villa may easily leave 100 times as
many sherds for us to find as does its Archaic predecessor, but it cannot be concluded that it housed a
significantly larger number of people.
Paradoxically, our case study also indicates that a comparison of regional pathways to complexity would be
tantamount to ignoring many significant intra-regional differences in the history of settlement and land
use. The Pontine region and the Salento Isthmus appear to be both too large to have developed a single
homogeneous trajectory, and too small to encompass the clearly relevant adjacent up- and highlands
13 -
26
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
whose populations drove early centralisation and urbanisation processes, and who resisted the
demographic and military encroachment of the lowland powers.
REFERENCES
Arroyo-Bishop, D & L Zarzosa 1992
The ArchaeoDATA system: A method for structuring a European archaeological information system (AIS),
in Larsen, CU (ed) 1992: 133-156.
Arroyo-Bishop, D & L Zarzosa 1995
To be or not to be: Will an object-space-time GIS/AIS become a reality or end up an archaeological entity?,
in Lock, GR & Z Stancic (eds), 1995:43-54.
Arthur, PR 1991
Romans in Northern Campania. Settlement and land-use around the Massico and the Garigliano Basin
(Archaeological monographs of the British School at Rome 1). London : British School at Rome.
Attema, PAJ 1993
An Archaeological Survey in the Pontine Region. A contribution to the early settlement history of south
Lazio 900 - 100 BC. 2 Vols. PhD thesis, Groningen: Archeologisch Centrum Groningen.
Attema, PAJ 2000
Ceramics of the first millennium BC from a survey at Lanuvium in the Alban Hills, Central Italy: Methods,
aims and first results of regional fabric classification, Palaeohistoria 39/40 (1997/1998): 413-439.
Attema, PAJ & PM van Leusen forthcoming
The Early Roman Colonization of South Lazio: a survey of three landscapes, JMA.
Attema, P, A Beijer, M Kleibrink, A Nijboer & G Van Oortmerssen forthcoming
Ceramics research at Satricum and Latium Vetus, Paleohistoria 43/44 (2001/2002).
Attema, PAJ, E van Joolen & PM van Leusen 2001
A Marginal Landscape: Field work on the beach ridge complex near Fogliano (South Lazio), PaleoHistoria
40/41 (1998/1999):149-162.
Attema, PAJ & PM van Leusen 2000
Kern en periferie in het RPC-project (1); de Doganella di Ninfa-survey in de Pontijnse regio (MiddenItalië), in PaleoAktueel 10: 25-30.
Brandt, RW et al. 1992
ARCHIS Archeologisch Basis Register, Versie 1.0. Amersfoort: ROB.
Christoffersen, J 1992
The Danish National Record of Sites and Monuments, DKC, in Larsen, CU (ed) 1992:7-21.
D’Andria, F n.d. (1999)
Ricerche recenti sugli insediamenti indigeni di Puglia e Basilicata, in La forma della città e del territorio,
esperienzemetodologiche e risultati a confronto (Atti dell’Incontro di studio – S Maria Capua Vetere 27-25
novembre 1998):103-18.
Drost, E 1996
De rol van de rivier de Astura bij de archeologische invulling van het Pontijnse landschap in Zuid-Latium.
Unpublished MA thesis, University of Groningen.
Drost, E 1997
Nederzettingen en landschap in het stroomgebied van de Astura, Zuid-Latium, Italië. PaleoAktueel 8
(1996):79-83.
Dyson, SL 1978
Settlement patterns in the Ager Cosanus: The Wesleyan University survey, 1974-76, JFA 5:251-268.
Favory, F & C Raynaud 2000
Définition ou hiérarchisation des sites? Approche intégrée en Gaule Méditerranéenne, in Pasquinucci, M &
F Trément (eds), Non-Destructive Techniques Applied to Landscape Archaeology (The Archaeology of
Mediterranean Landscapes 4): 223-232.
Feiken, H 2000
Survey en Survival. De Romeinse kolonisatie van het Fogliano en Sezze gebied in de Pontijnse regio.
Unpublished report, Groningen Institute of Archaeology.
Gierow, PG 1964
The Iron Age culture of Latium I. Classifications and analysis. Publications of the Swedish Institute in
Rome 4, XXIV(1). Lund.
Gierow, PG 1964
The Iron Age culture of Latium II. Excavation and finds 1. The Alban hills. Publications of the Swedish
Institute in Rome 4, XXIV(2). Lund.
13 -
27
V A N L EUSE N : P A T T E R N T O P ROCESS
Harris, TM & GR Lock 1992
Toward a Regional GIS Site Information Retrieval System: The Oxfordshire Site and Monuments Record
(SMR) Prototype, in Larsen, CU (ed) 1992:185-199.
King, C 1995
An Archaeological Survey in the Pontine Region, Italy (June 11th - 24th 1995). Unpublished field work
report, University of Groningen.
Koot, CW 1991
The analysis of the ceramics of the Agro Pontino Survey, in Voorrips et al. (eds) 1991:117-131.
Larsen, CU (ed) 1992
Sites & Monuments: national archaeological records. Copenhagen: National Museum of Denmark.
Loving, SH, H Kamermans & A Voorrips 1991
Randomizing our walks: the Agro Pontino survey sampling design, in Voorrips et al. (eds) 1991: 61-78.
Lloyd, J & G Barker 1981
Rural settlement in Roman Molise: Problems of archaeological survey, in Barker, G & R Hodges (eds),
Archaeology and Italian Society, BAR International Series 102: 289-304.
Mazzolani, M 1969
Forma Italiae regio I, volumen sextum: Anagnia. Istituto di Topografia Antica dell’Università di Roma.
Mills, N 1981
Luni: settlement and landscape in the Ager Lunensis, in Barker, G & R Hodges (eds), Archaeology and
Italian Society, BAR International Series 102: 261-8.
Morselli, C & E Tortorici 1982
Ardea (Forma Italiae, region 1, vol XVI). Firenze.
Perkins, Ph 1999a
Reconstructing the Population History of the Albegna Valley and Ager Cosanus, Tuscany, in Gillings, M,
D Mattingly & J van Dalen (eds), Geographical Information Systems and Landscape Archaeology (The
Archaeology of Mediterranean Landscapes 3):103-115. Oxford: Oxbow books.
Perkins, Ph 1999b
Etruscan Settlement, Society and Material Culture in Central Coastal Etruria. BAR International Series
788. Oxford: Archaeopress.
Picarreta, F 1977
Forma Italiae regio I, volumen XIII: Astura. Firenze.
Roorda, I & R Wiemer 1992
The ARCHIS Project: Towards a New National Archaeological Record in the Netherlands, in Larsen, CU
(ed) 1992:117-122.
Scollar, I 1992
The Bonn Archaeological Database, in Larsen, CU (ed) 1992:97-114.
Tobler, WR 1970
A computer movie simulating urban growth in the Detroit region, Economic Geography 46
(Supplement):234-40.
Van Leusen, PM 1998
Archaic Settlement and Early Roman Colonisation of the Lepine Foothills, in Assemblage 4 (1998),
http://www.shef.ac.uk/~assem/4/.
Van Leusen, PM & PAJ Attema 2000
Kern en periferie in het RPC-project (2); de Fogliano-survey in de Pontijnse regio (Midden-Italië), in
PaleoAktueel 10: 31-35.
Van Leusen, PM, RE Witcher & M Gkiasta in prep.
Survey data normalisation.
Vanzetti, A forthcoming
Results and problems of some current approaches to protohistoric centralization and urbanization in Italy,
RPC Conference proceedings.
Vittucci Brandizzi, P 1968
Cora, Forma Italiae regio I, Vol 5, Roma.
Voorrips, A, SH Loving & H Kamermans (eds) 1991
The Agro Pontino Survey Project, Studies in Prae- and Protohistorie 6. Amsterdam: Instituut voor Pre- en
Protohistorische Archeologie.
Wightman, EM 1981
The lower Liri valley: Problems, trends and peculiarities, in Barker, G & R Hodges (eds), Archaeology and
Italian Society, BAR International Series 102: 275-287.
Zaccheo, L & F Pasquali 1972
13 -
28
C O M P A R I S O N O F D A T A S E T S (P O N T I N E R EGION )
Sezze della preistoria all’etá romana (Historica Setina selecta). Sezze.
13 -
29
C HAPTER 14
L A N D U S E / L A N D C OV E R B I A S I N
T H E W R OX E T E R H I N T E R L A N D
1
I N T RO D U C T I O N
1.1
AIM
The aim of this case study, and of LULC mapping in the Wroxeter Hinterland Project, is to assess, model,
and compensate for the distorting effects of land use and land cover in the discovery of archaeological
sites – so-called bias modelling. These biases can be said to operate at two distinct scales of research: a)
the local scale, of an area investigated by field walking survey, and b) the regional scale, of an area
investigated by site-based desktop study or by aerial photography.
1.2
BACKGROUND
From 1994 to 1997 a 40 by 30 km area around the Roman civitas capital of Viriconium Cornoviorum
(modern-day Wroxeter in the county of Shropshire, UK) was the object of a regional GIS-supported
study funded by the Leverhulme Trust. A general description of the project can be found in Gaffney &
Van Leusen 1996; among its many aims was the study of methods for dealing with the main biases
present in the archaeological record of any area, and the Wroxeter hinterland in particular (Van Leusen
1996; chapter on biases). The current case study concentrates on just one of those biases – the varying
discovery rates of archaeological remains under different types of land use and land cover (‘LULC’), and
the effect that changes in LULC have on the nature of that record. For a more detailed discussion of
biases in regional archaeological data sets, see Chapter 4. The first specific aim of this case study is to
demonstrate that such effects can be measured, modeled, and – at least in part – corrected within a GIS
environment.
Whereas recent LULC is argued to be an important factor in creating patterning in the archaeological
record (see Chapter 2), it is of course recognised that social actions within the landscape, and the
archaeological remains resulting from those actions, are themselves also non-randomly distributed. A
subsidiary aim of this case study is therefore to review the historical, toponymic, and archaeological
evidence for a reconstruction of ancient (Roman) LULC in the WHP area.
Historic land use and land cover data for the Wroxeter Hinterland study area were obtained and digitised
from a number of sources, not all of which proved to be of use for the current case study. The three sets
of LULC data that will be presented and discussed here were obtained by the digitisation of land use
maps created by the Ordnance Survey of Great Britain (OSGB) around 1928 (i.e., before the post-war
industrialisation of agriculture in Britain)i; by georeferencing and supervised classification of a Landsat
TM image of the area dating to February of 1992ii; and by field work conducted by local volunteers in the
14 -
1
VAN LEUSEN: PATTERN TO PROCESS
course of the WHP parish survey (1996; partial coverage only)iii. Changes in LULC over time can be
traced by comparing these maps. These data will be compared with the distributions and characteristics of
the archaeological site data obtained from the Shropshire county Sites and Monuments Records.
1.3
TWO APPROACHES TO THE USE OF LULC HISTORY IN LOCATIONAL MODELLING
In this case study, two distinct approaches to the use of LULC history in the regional locational modelling
of sites in the Wroxeter Hinterland will be explored. Firstly (section 2), the quantitative approach that
regards LULC as just one of a range of environmental variables; and secondly (section 3), the historical
approach that attempts to reconstruct LULC for the periods being studied. Similarities and divergences
between the two approaches will be explored in a concluding section (section 4).
At the local scale, the history of land use is likely to vary field by agricultural field, and each of these can
be said to have its own ‘cultural biography’ which must be taken into account when, for example, the
results of field walking surveys are being analysed. The same effect occurs as well at the regional scale,
and with the site-oriented archaeological data typically available at that scale. Land use and land cover, and
changes in them over time, deeply affect not just the archaeological remains themselves (post-depositional
processes), but more specifically the archaeological record – i.e., the type, amount, and location of finds
and sites coming to the attention of professional archaeologists.
The inclusion of historical LULC data in locational models therefore has a very great impact on results.
There is evidence that LULC, through its differential effect on the chances of discovery of archaeological
remains, is one of the most important variables in ‘predicting’ site location (cf. Van Leusen 1993:114115). In their simplest form, LULC maps may be included in predictive locational models as just one
more ‘environmental’ variable (cf. discussion in Gaffney & Van Leusen 1995), to be correlated with the
locations of known archaeological finds. Alternatively, LULC data may be used to ‘correct’ traditional
predictive models by deriving the latter for each LULC category separately, through a masking step. That
this results in important changes to site characteristics, will be shown below.
2
A Q UA N T I TA T I V E A P P ROA C H
2.1
PROPERTIES OF THE LULC MAPS
Table 1 gives percentages of land use for 1100 km2 (88%) of the WHP study area, as derived from the
1928 mapping and the 1992 satellite image. It can be seen that some 15% of the total surface area was
converted from grasslands and rough pastures to other uses; unfortunately the satellite image
classification is not of sufficient quality to put much trust in the 1992 LULC percentages. There has also
been a large increase in the area of built-up between 1928 and 1992, but this can not be demonstrated due
to the restricted accuracy of the data used.
Higher quality data are available for a sample area of 148 km2 which was surveyed by local WHP
volunteers. Table 2 gives percentage coverage of each land use type within this area for the 1928 and 1996
mappings. These data clearly show the increased intensity of land use in the sample area, with over one
quarter of the land converted from extensive (hay and grazing) to intensive (arable) use between 1928 and
1996. Exactly when this change took place is not clear, but it is likely to have been progressive from the
post-war mechanisation at about 1950 onward.
14 -
2
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
Table 1 – Changes in land use over the period 1928-1992, Wroxeter hinterland.
land use category
1928
1992
change
woodland
6.23
8.94
+ 2.71
arable land
22.18
28.23
+ 6.05
meadowland and permanent grass
59.35
50.99
- 8.36
6.00
-
- 6.00
heath, moorland, commons and rough pasture
houses with large gardens, built-up areas
6.13
6.12
- 0.01
water bodies
0.10
1.28
+ 1.18
-
4.43
+ 4.43
100.00
100.00
0.00
Unknown / no data
totals
Table 2 – Changes in land use over the period 1928-1996, sample area.
land use category
1928
woodland
1996
change
6.03
8.75
+ 2.73
arable land
25.34
51.56
+ 26.22
meadowland and permanent grass
62.70
34.04
- 28.66
heath, moorland, commons and rough pasture
1.65
0.22
- 1.43
houses with large gardens, built-up areas
4.09
4.42
+ 0.33
water bodies
0.19
1.02
+ 0.83
100.00
100.00
0.00
totals
2.2
PROPERTIES OF THE SHROPSHIRE SMR DATA
At the time this study was conducted (1994-6), the Shropshire County Sites and Monuments Record
(SMR) held 936 site records for the 1178 km2 (38 by 31 km) study area. For the purposes of this case
study not all of the many attributes of the SMR records are relevant; the two properties we will discuss
here are site discovery mode and site type.
DISCOVERY MODE
In order to trace the circumstances under which archaeological remains were discovered, an obvious first
step is to use the information held in Shropshire SMR field 90 (Form). This field has been filled out for
810 records, and contains four relevant categories: Aerial photographic mark, Finds, Subsurface deposit,
and Earthwork. The category Finds is further subdivided into Finds Only and Finds Also. Table 3 gives the
counts and percentages of sites for these categories. Out of the 810 records, 28 must be rejected because
information about the original discovery is lacking. ‘Subsurface deposits’ is a category almost exclusively
applied to excavated deposits, which were presumably preceded by discovery in some manner other than
excavation. Likewise, it is unclear whether the category ‘Finds also’ covers finds made following discovery
by another method. In what follows, these categories will therefore play no role. Further information
regarding site discovery was sought from SMR fields 160 (Land use on site) and 220 (Description), but it was
found that the content of the former field was generally recorded post hoc from air photographs (severing
the potential link between LULC and discovery), while the log of events provided by the latter proved to
be not very informative about the circumstances of first discovery. Neither field was therefore used, and
the following analysis is solely based on the content of SMR field 90, and hence the total number of sites
available for analysis is 782. Of these, well over half (58%) were discovered by aerial reconnaissance,
another 29% consists of chance finds of (mostly) single objects, and the remaining 13% are relatively
obtrusive earthworks such as barrows and hillfort defenses.
14 -
3
VAN LEUSEN: PATTERN TO PROCESS
Table 3 – Recorded discovery modes of pre-Conquest sites in the Shropshire SMR.
Discovery Mode
Count
% of all records having a
discovery mode (n=810)
% of records with
primary discovery mode
(n=782)
AP mark
457
56
58
Earthwork
101
12
13
14
2
-
Finds
238
29
-
- Finds only
224
28
29
Subsurface deposit
- Finds also
totals
11
1
-
810
99
100
SITE TYPE
A total of 855 out of the original 936 pre-Conquest sites in the Shropshire SMR data set have a site type
attribute, as set out in Table 4. The spatial distributions of these sites are mapped against the background
of the 1928 LULC data in Figure 1. Since the majority of the non-stray and non-obtrusive records in the
SMR derives from aerial photographic reconnaissance, Table 5 lists site types for this discovery group as
well.
Table 4 – Breakdown of pre-Conquest site types recorded in the Shropshire SMR for the Wroxeter
hinterland.
Count
% of
total
-
18
2
Villas (Roman)
-
22
3
Enclosures
-
281
33
Pit Alignments
-
38
4
Burial Mounds
Ring ditches
73
9
Barrows
44
5
Field Systems
-
88
10
Trackways
-
33
4
Stray finds
a.o. Flint 50, Coins 30, Early Roman 73, Late
Roman 29
258
30
855
100
Site type
Subtype
Hillforts
totals
Table 5 – Breakdown of pre-Conquest site types discovered by aerial reconnaissance in the Wroxeter
hinterland.
AP mark type
Subtype
Enclosure
Curvilinear 22, Rectilinear 88, Hybrid 18
4
% of total
280
60
Pit alignment
-
38
8
Linear feature
-
43
9
Field system
-
44
9
Ring ditch
-
62
13
467
99
totals
14 -
Count
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
2.3
UNIVARIATE ANALYSIS
The first step in locational modeling generally involves checking (‘exploratory data analysis’) of the
correlations between the locations of site types and the values of independent variables. Univariate
locational preferences form the basis for deciding which variables are likely to be useful predictors. To
begin with, distribution maps were produced of the three major discovery classes (figure 2), and onesample univariate correlations with the three LULC maps were investigated (table 6).
Table 6a - Site List: smr290APmark (457 sites)
Layer: Land Use map 1930s (reclassified)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) water bodies
128285
13
( 1) woodland
109538
6.2
27.7
12
8.889
1
( 2) arable land
390001
22.2
98.6
216 139.854
1
( 3) meadowland and permanent 1043602
59.4
263.8
201
14.948
1
( 4) heath and moorland
105519
6.0
26.7
5
17.610
1
( 5) other built-up areas
107855
6.1
27.3
10
10.931
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
444.0
444 192.231
4
Table 6b - Site List: smr290Only (224 sites)
Layer: Land Use map 1930s (reclassified)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) water bodies
128285
15
( 1) woodland
109538
6.2
13.0
13
0.000
1
( 2) arable land
390001
22.2
46.4
29
6.528
1
( 3) meadowland and permanent 1043602
59.4
124.2
110
1.618
1
( 4) heath and moorland
105519
6.0
12.6
14
0.166
1
( 5) other built-up areas
107855
6.1
12.8
43
70.913
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
209.0
209
79.225
4
Table 6c. Site List: smr290Earthwork (101 sites) Layer: Land Use map 1930s (reclassified)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) water bodies
128285
11
( 1) woodland
109538
6.2
5.6
13
9.724
1
( 2) arable land
390001
22.2
20.0
7
8.435
1
( 3) meadowland and permanent 1043602
59.4
53.5
20
20.952
1
( 4) heath and moorland
105519
6.0
5.4
47 319.984
1
( 5) other built-up areas
107855
6.1
5.5
3
1.155
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
90.0
90 360.250
4
It is already evident from figure 1 that the existing recorded archaeological sites have a marked
‘preference’ for arable land in the 1920s mapping. If we break the record down into groups according to
discovery mode, a detailed picture emerges in which arable land use is an excellent predictor for the
presence of AP marks (table 6a, cat 2), stray finds ‘prefer’ intensively visited and worked areas and ‘avoid’
the only relatively inaccessible arable (table 6b, cats 2, 5 & 6), and earthworks are preserved from the
plough mostly on uncultivated heath- and woodland and, to a lesser extent, on permanent grassland
(table6c, cats 1, 3 & 4)iv.
14 -
5
VAN LEUSEN: PATTERN TO PROCESS
Figure 1 - top: SMR discovery type against 1928 LULC
(N=810). Red box: AP mark. Yellow diamond: earthwork. Blue
diamond: stray find. Bottom: SMR site types against 1928
LULC (N=855). White box: hillforts. Orange diamond: villa
(closed=certain). Yellow star closed: enclosure; open: field system.
Green box closed: barrow; open: ring ditch. Magenta box: pit
alignment. White x: stray find.
14 -
6
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
Figure 2 - Major site discovery classes for 855 pre-Conquest sites in the Wroxeter
Hinterland (source: Shropshire SMR). Red dots: AP marks. Yellow boxes:
chance finds. Blue dots: earthworks.
The overall preference for arable appears to be largely caused by a subset of enclosures and field systems
that were identified on aerial photographs, a prospection technique which is known to introduce a bias in
favour of arable land – either freshly ploughed or under a young or mature crop. Rather than interpreting
the patterning we have discovered in our data as a reflection of the original patterned distribution of this
type of site, we would suspect it to be caused perhaps by especially disruptive agricultural practices in
these areas, or by a heightened soil and crop response which makes certain types of archaeological
features show up better in aerial photography. Either way, recent and modern land use is heavily
implicated in the formation of the pattern.
Univariate preferences / avoidances for enclosures and fieldsystems show that, already in the 1920s, these
sites are characterised by avoidance of woodland, grass- and meadowland, and low density builtup (the
latter presumably caused by the fact that these areas were taken out of agricultural use) and the converse
strong preference for arable. This situation continues in 1992, but with a less strong avoidance of
grassland, which however in 1996 returns (by which time other cats are below statistical threshold).
14 -
7
VAN LEUSEN: PATTERN TO PROCESS
2.4
MULTIVARIATE ANALYSIS
LULC and the locations of archaeological sites are of course correlated not just through the process of
discovery, but also through the land use qualities of the soils and geomorphology of the study area. This
three-way correlation can be quantified by statistical measures (if the scale of the variables is at least
ordinal), but it is not possible to extract causal relationships by this means. Two examples of this
phenomenon are given below.
Figure 3 - Enclosures discovered by aerial reconnaissance (n=280) vs. soil groups of the
Wroxeter Hinterland.
14 -
8
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
Table 7 (a-e) – Chi2 result tables for site types on simplified 1928 LULC. A: enclosures (n=280), B:
ring ditches (n=62), C: field systems (n=44), D: linear features (n=43), E: pit alignments (n=38).
A)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) no data
128285
9
( 1) uncultivated land
215057
12.2
33.2
8
19.108
1
( 2) arable land
390001
22.2
60.2
129
78.735
1
( 3) meadowland and permanent 1043602
59.4
161.0
130
5.972
1
( 4) built-up areas, etc.
107855
6.1
16.6
4
9.602
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
271.0
271 113.417
3
B)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) no data
128285
5
( 1) uncultivated land
215057
12.2
7.0
1
5.122
1
( 2) arable land
390001
22.2
12.7
36
43.060
1
( 3) meadowland and permanent 1043602
59.4
33.9
16
9.425
1
( 4) built-up areas, etc.
107855
6.1
3.5
4
0.071
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
57.0
57
57.678
3
C)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) no data
128285
0
( 1) uncultivated land
215057
12.2
5.4
1
3.573
1
( 2) arable land
390001
22.2
9.8
28
34.020
1
( 3) meadowland and permanent 1043602
59.4
26.1
15
4.749
1
( 4) built-up areas, etc.
107855
6.1
2.7
0
2.702
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
44.0
44
45.043
3
D)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) no data
128285
0
( 1) uncultivated land
215057
12.2
5.3
2
2.024
1
( 2) arable land
390001
22.2
9.5
17
5.818
1
( 3) meadowland and permanent 1043602
59.4
25.5
23
0.254
1
( 4) built-up areas, etc.
107855
6.1
2.6
1
1.019
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
43.0
43
9.115
3
E)
degrees
cells
%
expected actual
chi
of
Site Characteristics
cover
cover
sites
sites square freedom
-------------------------- ------- -------- ------ -------- ------( 0) no data
128285
0
( 1) uncultivated land
215057
12.2
4.7
2
1.512
1
( 2) arable land
390001
22.2
8.4
24
28.706
1
( 3) meadowland and permanent 1043602
59.4
22.6
12
4.955
1
( 4) built-up areas, etc.
107855
6.1
2.3
0
2.333
1
------- ------- -------- ------ -------- ------Totals
1756515
100.0
38.0
38
37.507
3
14 -
9
VAN LEUSEN: PATTERN TO PROCESS
Comparing discovery method with soil type, we find that enclosure sites discovered from the air avoid
peat, ground-water gley, lithomorphic soils and podzols, preferring brown earths and surface-water gley
soils. These associations become even stronger when we exclude areas that were not extensively studied
from the air - brown earths are now even more preferred (Chi2 of 19), while surface-water gleys are less
favoured, and ground-water gleys are more clearly avoided. We know that this must partly be caused by
visibility biases - ground water gleys being much less subject to drying out, and podzolic soils tending to
be poor and therefore covered with woodland, scrub or heath (cf. Jones & Evans 1975). Of the two
major soil classes in the area, surface-water gleys are slightly less workable than brown earths, and tend to
occur farther away from streams and roads.
Figure 4 - Distribution of barrows (white) and ring ditches (red) vs. soil groups in the
Wroxeter Hinterland. For legend, see figure 3.
Table 8 - Chi2 table of AP mark enclosures (n=280) vs soil groups.
Site Characteristics
-------------------( 0) no data
( 3) lithomorphic soils
( 5) brown calcareous earth
( 6) podzolic soils
( 7) surface-water gley soi
( 8) ground-water gley soil
( 9) man-made soils
( 10) peat soils
Totals
14 -
10
cells
%
cover
cover
------- ------31332
495
0.1
184050
41.8
35748
8.1
194997
44.3
9758
2.2
7058
1.6
7762
1.8
------- ------439868
100.0
degrees
expected actual
chi
of
sites
sites square freedom
-------- ------ -------- ------4
0.3
0
0.311
1
115.5
163
19.550
1
22.4
4
15.144
1
122.4
108
1.684
1
6.1
1
4.286
1
4.4
0
4.429
1
4.9
0
4.870
1
-------- ------ -------- ------276.0
276
50.273
6
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
A second example illustrating the effects of visibility on the distribution of archaeological site types are
barrows and ring ditches. Set against soil types, ring ditches display a preference for brown earths
although they also occur on surface-water gley soils, while barrows have a very strong preference for
podzolic soils (Chi2 of 92) and occur to a much lesser extent on either brown earths or surface-water
gleys. Yet we know that these two data samples are drawn from one parent population - mainly Bronze
Age barrows, which have been agriculturally degraded on some soils while being preserved on others.
Table 9 - A: Chi2 table of barrows (n=44) vs soil groups. B: Chi2 table of ring ditches (n=73) vs soil
groups. Units with expected sites < 3 have been omitted.
A)
Site Characteristics
-------------------( 0) no data
( 5) brown calcareous earth
( 6) podzolic soils
( 7) surface-water gley soil
Totals
B)
Site Characteristics
-------------------( 0) no data
( 5) brown calcareous earth
( 6) podzolic soils
( 7) surface-water gley soil
Totals
degrees
cells
%
expected actual
chi
of
cover
cover
sites
sites square freedom
------- ------- -------- ------ -------- ------225620
5
736200
44.4
17.3
8
5.003
1
142992
8.6
3.4
21
92.568
1
779988
47.0
18.3
10
3.788
1
------- ------- -------- ------ -------- ------1659180
100.0
39.0
39 101.359
2
degrees
cells
%
expected actual
chi
of
cover
cover
sites
sites square freedom
------- ------- -------- ------ -------- ------225620
7
736200
44.4
29.3
47
10.716
1
142992
8.6
5.7
2
2.391
1
779988
47.0
31.0
17
6.341
1
------- ------- -------- ------ -------- ------1659180
100.0
66.0
66
19.449
2
There are several ways in which we can approach the three-way correlation of natural environment –
LULC – archaeological site location. Among the specialist multivariate tools at our disposal are 3-way
contingency table analysis and log-linear analysis, but the simpler approach chosen below reduces the
multivariate problem to a series of bivariate problems. First, we study locational preferences within each
discovery class (discovery mode invariant); next, we study them within each LULC class (LULC class
invariant).
DISCOVERY MODE INVARIANT
One way of studying the potential significance of such biases in archaeological records is to assume that
the distribution of sites first recorded by one particular method is entirely due to bias, whereas any
differences in distribution of site types within this group may be due to locational factors. The largest
group of known sites by discovery mode are the AP marks, and these therefore give most scope for this
type of analysis. As we saw, the Shropshire SMR contains 457 records of type AP mark within our study
area. We have examined the distributions of the largest subgroups by morphology - Enclosure, Ring
ditch, Field system, Pit alignment, and Linear feature (see figure 3 and table 7), and found that, whereas
the distribution of enclosures and linear features are not significantly different from that of all AP marks,
the distributions of ring ditches and field systems seem to be more restricted to the central lowland. Most
remarkably, pit alignments display a clearly different distribution from that of all AP marks - they are
located mainly in the Tern and upper Severn watershed areas! This presents us with a subject for further
study: do these alignments bear any relation with topography, with the division of the landscape between
farmsteads, were they wind breaks?v
14 -
11
VAN LEUSEN: PATTERN TO PROCESS
LULC CLASS INVARIANT
The inverse of the above operations assumes that LULC bias can be avoided by separately modelling site
location preferences for each LULC type. Thus, if the discovery of AP marks is contingent on the
presence of arable, then the distributions of site types discovered by aerial reconnaissance within the arable
may be used to model site location preferences that do not suffer from these biases. For this, we cannot
use the 1928 LULC map by itself, because it represents only one early stage in the history of modern land
use in the study area. We must instead construct a map that shows the probability that any particular area
will have been in arable use in the latter half of the 20th century.
I have approached such a map by adding up all arable in the three base maps (i.e., dated 1928, 1992, and
1996) and applying distance buffer zones of 100, 250, 500, and 1000 meters to that. To obtain the
cleanest possible results with the given data, I further restricted the study area to that in which the aerial
reconnaissance unit of the RCHME operates (west of the line x=363000). The results, presented here in
figure 5 and table 10, show that AP marks are extremely strongly correlated with arable land use. All but
seven out of 422 AP marks are either on, or within 250 meters of, known arable land (Chi2 of 155 at 1
degree of freedom).
Table 10 – Correlation of 422 AP marks in the Shropshire SMR to combined arable. Analysis region
bounded by north: 325000, west: 336000, east: 363000, south: 294000 NGR.
Site Characteristics
-------------------( 0)
( 1) Arable
( 2) 250 meter buffer
Totals
degrees
cells
%
expected actual
chi
of
cover
cover
sites
sites square freedom
------- ------- -------- ------ -------- ------163042
7
540100
45.9
190.6
317
83.876
1
636058
54.1
224.4
98
71.222
1
------- ------- -------- ------ -------- ------1176158
100.0
415.0
415 155.098
1
It is probable that this correlation could be strengthened even more by taking further variables into
account, such as the areas effectively excluded from observation because of flying restrictions, ancient
woodland and water bodies, and pre-war built-up areas. However, the point of the exercise was to
demonstrate how bias factors can be incorporated into locational models.
CORRECTION
Having identified land use as a factor contributing bias to our recorded site distributions, it is now
possible to use GIS to, firstly, quantify that bias and, secondly, compensate for it. Various GIS techniques
for doing this have been proposed (Terrenato & Ammerman 1996, Van Leusen 1996), which are
generally referred to as weighting or evaluation schemes.
Quantifying the bias in a distribution of archaeological sites could be done by directly assigning weights or
values to each of the land use categories on the basis of its presumed effect on the visibility of sites. For
example, high weights (ie, low visibility) are assigned to built-up and forested areas and water bodies, and
low weights (high visibility) to arable land. Weights or values could also be derived automatically by
assuming that the known distribution of archaeological sites directly reflects differential visibility – in this
case, one could for instance use the Chi squared or P values for random distribution of sites as the
weight. Problematic in these approaches is, that the weight variable has never been independently
measured. It is important to remember, however, that these are just the extremes of a whole range of
possible methods for creating a map quantifying a particular bias.
14 -
12
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
Figure 5 – Distribution of AP marks (black diamonds, n=422) with respect to distance to arable
land. Light grey: arable in 1928, 1992, or 1996. Dark grey: 250m buffer around arable. Grid
line spacing: 10 kms.
3
M O D E L I N G A N C I E N T L U L C : A H I S TO R I C A L A P P ROA C H
Whilst modern LULC has the unfortunate effect of biasing our recovery of the archaeological record in
various ways, causing spurious patterning, that of course does not mean that no ‘real’ patterning exists in
the archaeological record. A study of the ancient (in this case Roman) LULC might clarify why we find
the material residue of various social actions in particular parts of the landscape. In the absence of
14 -
13
VAN LEUSEN: PATTERN TO PROCESS
any direct historical or archaeological (eg, paleo-environmental) evidence, we have investigated if later
historical, archaeological, and place-name evidence might be extrapolated to the remoter past, and thus
bring us closer to Roman LULC. The following discussion concentrates on the evidence of most
immediate relevance to LULC, while a more extensive discussion of literary sources may be found in
White & Barker (1998:130-6).
STABILITY IN THE LONGUE DURÉE
3.1
A reconstruction of Roman LULC in the central Shropshire area must start with the observation that
historical evidence indicates that land use in the study area appears to have been extremely stable over the
centuries. Thirsk (1987:29) mapped the farming regions of England for the period 1500-1650, showing
the WHP study area as open pasture used for rearing and fattening. The situation appears little changed
during the period 1640-1750, with farming land use down to cattle and sheep rearing, sometimes with
dairying (on fells and moorland), stock-fattening with horse breeding, and fishing and fowling (in
fenland).
Shropshire and Cheshire, the core Cornovian lands, are today still some of the best cattle land in the
country. Given this continuity, we may ask if perhaps it is possible to trace this emphasis on animal
farming back to even earlier times. The validity of such extrapolation is supported by authors such as
Williamson (1988) who projects Anglo-Saxon land management systems in the Norfolk area back into
late Roman times:
‘By the end of the Roman period there were around 1.3 settlements per square kilometre. (...) settlements
tended to cluster near to the margins of the lighter soils of the valleys, or on the floors of the valleys themselves.
Plateaus were more sparsely, but apparently quite evenly, populated, but the settlement pattern was more mobile
and the settlements themselves more short-lived. (…) Settlement had been re-established in all parts of the area
studied by the end of the Saxon period, and its pattern exhibits a number of similarities with that of the Roman
period. In particular, the farmsteads and hamlets away from the major valleys were of lower status than those
located adjacent to them; the former were tenurially dependent on the latter, and were therefore usually unnamed in
the Domesday survey.’ (Williamson 1988:162-164)
Such remarks are relevant to at least the southern parts of the WHP study area, where a similar
topography could have resulted in a similar settlement system. Thus, documentary evidence seems to
argue that the longe durée LULC in Shropshire is one of animal husbandry. White further supports this with
his reconstruction of the economic wealth of late pre-Roman Iron Age Cornovian society and certain
striking features of Roman Viroconium itself – the probable forum boarium and extensive tanneries in the
northern part of the town (White & Barker 1998: 79, 92).
3.2
PLACE-NAME ETYMOLOGY
However, a further and more specific type of evidence can be adduced by studying Anglo-Saxon charters
and place-name evidence. To begin with the latter, Gelling (1992) discusses the available place-name
evidence for the presence of ancient woodland in Shropshire (see Figure 6). She mapped place-names of
‘leah’ (modern -ley) type and found
‘…a very dense concentration of symbols in the eastern half [of Shropshire]. Two groupings can be
discerned, one running from Leegomery (south of the Weald Moors) to Meadowley (west of Bridgnorth), and
another occupying the south-eastern corner of the county. The River Severn was here flowing through dense
woodland, some of which survived to form the basis of the early industrial activity in Coalbrookdale… Some ()
names have first elements (such as ‘cat’ in Ketley, ‘burdock’ in Clotley, ‘fern’ in Farley, ‘burnt’ in
14 -
14
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
Barnsley) which suggest a relatively underdeveloped landscape. It seems probable that in eastern Shropshire the -ley
names give a fair impression of the whole extent of the Anglo-Saxon woodland, rather than just marking the core
of it. If this woodland had been continuous it would have exceeded that of north Warwickshire in extent, but the
place-names give evidence of an open belt running from Upton Cresset to Eardington…’
‘There is another belt of -ley names running east/west across the centre of Shropshire. Two outliers,
Bradley and Farley, at the eastern end of the belt are minor names in Much Wenlock parish. West of these are
Harley, Hughley, Kenley, Langley, Ruckley, Frodesley, Lydley Heys and Leebotwood. These are all ‘major’
names, the last two referring to places in Botwood where Haughmond Abbey was making assarts (the technical
term for woodland clearings) in the second half of the twelfth century. Since leah is not likely to have been used in
the sense of ‘forest clearing’ as late as that it is likely that some small settlements here had the name Lege from an
earlier date, and that these provided bases from which the abbey developed more land (…) No other groups of -ley
names in Shropshire are as extensive as these belts in the east and centre, and there are large areas in which the
word does not occur at all. There was some ancient woodland north of Shrewsbury (referred to in Pimley,
Albrightlee, Astley), and Lee Brockhurst, Marchamley and some minor names attest to another patch further
north, between the Roden and the Tern.’ (Gelling 1992:15-17)
We argue that if ‘–ley’ place names represent Anglo-Saxon clearings within (partly) wooded areas, then
these areas were likely to have been (managed) woodland in late Roman times and possibly earlier. Such
woodland occurs as a belt along the lower slopes of the Long Mynd and other high relief terrain in the
south of the study area, and up through the Telford area; another belt of wooded land lies across the
Severn Valley where Shrewsbury is today. Both of these woodland belts seem to be related to areas of
lower accessibility, mostly related to the peculiar geology of the study area, which causes the topography
to run at right angles to the main Severn corridor. Elevated areas are therefore generally less accessible
and have lower soil quality. This is most clearly demonstrated by the latter woodland belt, which is aligned
on the geological ridge connecting Lyth Hill, Bayston Hill, and Sharpstones Hill to the south of the
Severn with Haughmond Hill to its north.
Overlaying the -leah map with the locations of enclosure sites (see figure 6) shows a strong correlation
between farmstead sites and non-wooded areas. The central woodland belt separates two clusters of
enclosures, with the core of the northwestern cluster somewhere near Great Ness, some 20 kms from
Wroxeter. Unfortunately, the ‘gap’ between the clusters is occupied by modern-day Shrewsbury, with the
ensuing lack of data. However, it is certain that the woodland and the high relief would together have
formed a clear physical as well as a psychological barrier, even though there might well have been a wide
corridor on both sides of the Severn which linked both clusters.
It is noteworthy that the northwestern cluster occupies the largest continuous area of good soils within
the Severn valley and probably had good communications to Wroxeter in the Roman period. We are
justified in expecting the presence of a major secondary centre (a la Rutinium), or perhaps a lesser one near
Montford Bridge (a la Uxacona).
Similar reasoning applies all the more strongly to the larger south-eastern woodland belt. The relief in this
area is much more dramatic, and communications with the central Severn valley much more restricted.
We find that the density of enclosures increases again to the south and east of this woodland, indicating
that we are here entering the territorium of another town, perhaps located at the Severn crossing at
Bridgenorth.
14 -
15
VAN LEUSEN: PATTERN TO PROCESS
Figure 6 - comparison of the distribution of –ley names (black), enclosure sites (red), and charter
boundaries (dashed line). Green: reconstructed woodlands. Blue: historic wetlands. (-leah place
names in Shropshire after Gelling 1992, fig. 6; reconstruction of Wroxeter’s territory in the
late/sub-Roman period after Bassett 1989)
3.3
DOCUMENTARY SOURCES
Further and independent evidence confirming this configuration comes from documentary sources
compiled by Bassett (1989). Combining certain oddities in diocesan boundaries near Wroxeter with
mention of the area of ‘Tren’ and ‘the Ercalls’ and a group of people known as ‘Wreoconsaetna’, Bassett
postulates a British territory extending from the Long Mynd in the south-west of the study area to the
Staffordshire borders in the north-east (see figure 6, dashed line). This putative territory can be shown to
have had some residual meaning in the 9th century at the latest, with Welsh poems referring back to the
7th, which would take this territory right back to a period when Viroconium was still in existence
14 -
16
LULC B I A S I N T H E W R O X E T E R H I N T E R L A N D
(cf. White & Barker 1998: 132-6). If true, this constitutes powerful confirmation of the existence of a
territorial boundary following the line of the three hills mentioned above in section 3.2; to the south and
southeast, Wroxeter’s territory would encompass the valley and slopes facing the Severn; to the north of
the Severn boundaries are less well defined but it may well be that the important wetland ecozone to the
northeast was included together with most of the valleys of the Tern and the Roden (the latter if the
Ercalls are to be included); the least well-defined boundary is to the north-west.
4
CONCLUSIONS
The study of LULC for regional archaeological research can be said to have a methodological and a
historical purpose. The former is perhaps best approached by the use of GIS to store and compare
historical cartographic data about land use. As in the case of the WHP, such data may be derived from
archival records made for military, legal or taxation reasons; from studies of agricultural productivity; and,
more recently, from historic aerial photography and satellite imagery. LULC maps can then be compared
and correlated to visibility and discovery mode aspects of the archaeological record. The case study
presented here found that recorded discovery mode in the Shropshire SMR does indeed correlate with
historic LULC, and two examples were given. It was also shown how analysis of the archaeological record
within a particular discovery mode can flag up significant deviations.
The second, historical, use of LULC studies is here demonstrated by reconstructing a regional pattern of
arable vs woodland for the late Roman period, which is supported by archival studies of diocesan
boundaries and literary topographic references. Such studies are needed if we are to distinguish between
archaeological patterns relating to ancient LULC, and those relating to modern LULC.
The scope of the current case study has been limited by the available time and by the relatively low quality
of the data available from the Shropshire SMR. However, it does demonstrate the feasibility of this type
of study, its potential for the understanding of patterns in the archaeological record; and it explores some
ways forward.
ACKNOWLEDGEMENTS
I would like to express my gratitude to Roger White, my co-worker in the WHP, who provided me with
much of the evidence and interpretations underpinning the work presented here – especially that of
section 3.
REFERENCES
Bassett, S 1989
In search of the Anglo-Saxon kingdoms, in Bassett, S (ed), The origins of Anglo-Saxon kingdoms. London:
Fox, HSA 1989
The People of the Wolds: Settlement History, in Dyer, CC (ed), The Rural Settlement of Medieval England: 77-101.
Gaffney, VL & PM van Leusen 1995
Postscript: GIS and Environmental Determinism, in Lock, G & Z Stancic (eds), GIS and Archaeology: a European
Perspective: 367-82. London: Francis & Taylor.
Gelling, M 1992
Signposts to the Past. Chichester: Phillimore & Co.
Hooke, D 1981
Anglo-Saxon Landscapes of the West Midlands: the charter evidence. BAR British Series 95.
Jones, RJA & R Evans 1975
Soil and crop marks in the recognition of archaeological sites by air photography, in Wilson, DR (ed), Aerial
reconnaissance for archaeology (CBA Research Report 12):1-11.
14 -
17
VAN LEUSEN: PATTERN TO PROCESS
Terrenato, N & AJ Ammerman 1996
Visibility and Site Recovery in the Cecina Valley Survey, Italy, Journal of Field Archaeology 23 (1996): 91-110.
Thirsk, J 1987
Agricultural Regions and Agrarian History in England, 1500-1700, Studies in Economic and Social History.
Macmillan.
van Leusen, PM 1993
Cartographic Modelling in a Cell-Based GIS, in Andresen, J, T Madsen & I Scollar (eds), Predicting the Past.
Computer Applications and Quantitative Methods in Archaeology 1992: 105-124. Aarhus: Aarhus University Press.
van Leusen, PM 1996
Unbiasing the Archaeological Record, in Archeologia e Calcolatori 7: 129-136.
van Leusen, PM & VL Gaffney 1996
Extending GIS Methods for Regional Archaeology: the Wroxeter Hinterland Project, in Kamermans, H & K
Fennema (eds), Interfacing the Past. Computer Applications and Quantitative Methods in Archaeology 1995 (Analecta
Praehistorica Leidensia 28): 297-305.
White, RH & Ph Barker 1998
Wroxeter. Life and Death of a Roman City. Stroud: Tempus.
Williamson, T 1993
The origins of Norfolk. Manchester UP.
Williamson, T 1987
Early Co-axial Field Systems on the East Anglian Boulder Clays, in Proceedings of the Prehistoric Society 53: 419-431.
Williamson, T 1988
Settlement Chronology and Regional Landscapes: the Evidence for the Claylands of East Anglia and Essex, in
Hooke, D (ed), Anglo-Saxon Settlement: 153-175. Oxford: Blackwells.
i OSGB 1928, Land Utilisation Survey of Britain, 1:50,000. Map sheets 60: Shrewsbury and Welshpool, 61:
Wolverhampton, and 70: Bishop’s Castle.
ii
Landsat TM floating quarter scene 5/203/023 obtained from NERC data centre.
iii
Details of the original data sets and processing steps are recorded in ATMPROC.DOC and DATA.DOC.
iv Incidentally, this picture is further biased by a set of barrow observations contributed by a special earthwork
survey of the Long Mynd area.
v Current thinking holds that some later alignments at least (tentatively dated to the Iron Age – Medieval period),
consisting of smaller pits, functioned as boundary markers. Nothing is known about the alignments consisting of
large pits, which may date to the Bronze Age.
14 -
18
C HAPTER 15
SETTLEMENT HIERARCHIES,
T E R R I T O R I A L D I V I S I O N S, A N D
VISUAL DOMINANCE∗
1
INTRODUCTION
The late protohistoric (Late Bronze Age/Early Iron Age) settlement pattern in the south, as in central
Italy, has been interpreted as one indicative of the transition of pastoral ways of life to one dominated by
agriculture and increasing hierarchisation of society, culminating in the rich graves of Iron Age elites in all
areas of Italy and in evidence for early urbanisation in many regions but apparently developing first in
Etruria and Lazio. Among the causative factors, the presence of exploitable mineral resources has been
suggested (Vanzetti, forthcoming); the persistence of tribal structures in the Sibaritide seems to confirm
this idea which may be combined with that of a difference in demographic ‘carrying capacity’due to
geological and climatological differences which translated into different potentials for land use and the
relatively late take-off of urbanisation.
Late protohistoric centralised settlement has been interpreted as evidence of a territorial division in which
each settlement laid claim to essential landscape resources, with an initial phase of peer polity interaction
being followed by one in which a single settlement obtained hegemony and others are relegated to 2nd
rank (Rome in Lazio, Torre Mordillo in the Sibaritide). The further development of this system seems to
have been aborted in the Late Iron Age and the Archaic/Classical period in the south when economic life
re-oriented on the successful colonies; in central Italy it was the hypertrophic development of Rome
which disturbed the equilibrium.
Peer polity interaction is an inherently unstable system because it will be upset whenever any of the
polities gets preferential access to resources. Rome, as a ‘border’polity of the Latial league, could enlarge
its territory through conquest to its north and west, and through its position on the Tiber could control
and profit from river-borne and coastal trade. These options were largely closed to the ‘central’polities of
the Alban hills. However, in its developed form the peer polities of the Pontine, Sibaritide, and Salento
regions were modelled using Thiessen polygons (Bouma & Van ‘t Lindenhout 1998, Peroni 1994:282ff.
and 1996, Burgers 1999) under assumption of equality and a simple rank-size hierarchy consisting of just
two levels.
Because of the research interests of the culture-historical paradigm (Formazione delle città, urbanisation, and
colonisation; see chapter 2) and the limitations inherent in site databases collected before the advent of
modern ‘landscape’surveying (chapters 4 and 13), theorising about settlement patterns has been limited
∗
These case studies are concerned with two of the three study areas only; case studies involving the rank-size analysis and X-tent
modeling of late Iron Age to Archaic settlement in the Salento Isthmus could not be completed in time for this thesis.
15 -
1
V A N L EUSEN : P ATTER N T O P ROCESS
to the top of the settlement hierarchy in the three study areas. Even in regions with a relatively highquality data base, typological classifications of these settlements have had to be based on locational
characteristics or size of inhabited or defended area. For example, Guaitoli (1977:22-25) classified the
protohistoric to Archaic settlements of Lazio into ‘coastal’, ‘crater-rim’,‘military-commercial’,and ‘minor’
groups; he gave no size-rule to support these classes. By contrast, Guidi (1985) applied a rank-size
analysis to the same set of settlements, estimating settlement size from the area bounded by natural
defenses, and concluded for the late 7th century BC (orientalizing period) that there was evidence for the
development of a three-level hierarchy with Rome at the top, Ardea and Gabii at level two, and other
towns dependent on these three at level three.
Settlement size plays a central role in all models of late protohistoric and archaic societies in Italy, from
Etruria and Lazio to the Salento and the Sibaritide. In the large majority of cases, sizes are estimated on
the basis of three assumptions: finds from slopes and valley floors originate from hilltops and plateaus;
these hilltops and plateaus were entirely rather than locally inhabited; and habitation is assumed to be
continuous even if some phases are not well represented at a settlement (Vanzetti, forthcoming). Moreover, given the current state of research many settlements of significant size still remain undiscovered.
For these reasons, any specific models of territorial organisation based on rank-size criteria must be
regarded as weak.
Whatever the problems in establishing the size and rank of protohistoric settlements, the locations of
what has been regarded as the upper-rank settlements are generally described as ‘dominant’, expressing
their elevated, easily defensible positions and large viewsheds within which essential landscape resources
such as transhumance routes were located. This opens up the possibility of studying the viewshed
properties of these settlement systems and the landscape that they are part of. In order to explore further
the issues raised in this introduction, three interrelated case studies are presented here in two
chronological sections. The following section presents a protohistoric and pre-colonial case study
regarding accessibility, visual contact, and territorial structure in the Sibaritide and the Pontine region; the
final section presents a case study of the Roman colonial landscape in the Pontine region. The
hierarchisation of protohistoric settlement systems on the basis of size and locational characteristics is
examined with reference to Peroni’s (1994) models for southern Italy, the territorial organisation of Late
Iron Age to Archaic early states is explored through visibility and accessibility analysis with reference to
Bietti Sestieri's (1985) models for Lazio; and the role of strategic considerations in the location of early
Roman strongpoints in the Pontine region (south Lazio) is considered through an investigation of 4th
century BC colonies on the Lepine scarp, in the context of Livy's historical references.
2
SETTLEMENT AND TER RITORY IN PROTOHISTO RY
THE SIBARITIDE
The locational characteristics of the larger settlements of the south Italian Bronze and early Iron Ages are
generally understood to be a function of both local (defensibility, available area, presence of sufficient
agricultural land) and regional criteria (a ‘commanding’position, sufficient distance from neighbouring
settlements, access to both low- and highlands). However, there is a measure of vagueness in the way
these criteria are applied. In the case of the Sibaritide, a review of the literature reveals the following
suggestions for the operationalisation of geographical models of Middle Bronze Age to Early Iron Age
societies:
• Defensibility - This is taken to require a cape- or promontory-like geomorphology; candidate locations
should be accessible from one direction (upslope) only, the other directions being characterised by
steep slopes. No specifications are given for the amount of steepness considered sufficient.
15 -
2
S E T T L E M E N T L O C A T I O N M ODELS
• Available area - Depending on whether a settlement is thought to fulfill mainly habitation or defensive
functions, the criteria for available area may specify a lower or an upper limit, or both. Except in
cases where the area of a settlement has in fact been measured from evidence of walls etc, the
available area can only be specified as a contiguous block of land of less than a specified slope, e.g. 16
percent.
• Presence of sufficient agricultural land - The two factors making up this criterion are ‘sufficient’and
‘agricultural’. The amount of land to be considered sufficient for a specific settlement depends
obviously on the amount and type of produce being consumed or traded from there. From the
Middle Bronze Age onwards, a spreading of dry farming techniques into the uplands has been
proposed by Barker (1985), while in the Final Bronze Age olive culture is added to the range of
agricultural techniques. In the Sibaritide the soils of the marine/fluvial terraces are considered fair
agricultural land for dry farming of grains and vegetables, as long as slopes are not so steep as to
cause erosion. However, given the relatively small size of even the major settlements of the Bronze
and early Iron Ages, this factor is unlikely to have much restricted settlement location. Kleibrink
suggests that the 25 ha of agricultural land available near the settlement of San Nicola (Peroni &
Trucco 1994, site 31) would be sufficient for it. The remaining land, including that located on lower
quality soils and on slopes, could have been used for olive culture.
• A ‘commanding’position - This criterion is not specified by the authors, but is usually interpreted as
signifying that the settlement should have an unusually large viewshed and/or an unusually complete
‘near’viewshed. A second potential characteristic of a commanding position is that of easy access to
whatever is being ‘commanded’,but the latter remains unspecified except as a minimum or maximum
vertical distance of the settlement from the nearest valley floor. For the Sibaritide one can think of:
summer and winter grazing and other economic interests; settlements, cemeteries and other social
interests; the coast, long distance routes and other strategic interests.
• Sufficient distance from neighbouring settlements - From the Middle Bronze Age onwards, and accelerating
toward the Final Bronze Age under the influence of technological change and demographic growth,
the undifferentiated settlement system of the Early Bronze Age in the Sibaritide is thought to have
crystallised into a hierarchical system of major and minor settlements. The territories of the major
settlements are bounded by the valleys of major streams; where such streams are close together the
operative criterion may be ‘social’distance as well as size of territory. In the Sibaritide in the Early
Iron Age this distance appears to be on the order of 15 km; major settlements of the Salento Murge
typically are some 12 km apart in this period. These central villages would begin to function as
redistribution centers from the Late Bronze Age onwards, as is shown by the evidence for storage of
large amounts of agricultural produce at Broglio di Trebisacce (Levi 1999:229). The minor
settlements in this system are thought to have strategic functions: either to provide safety from attack
or cattle raids from the direction of the uplands, or to control the crossings of the litoraneo protostorica
over the main river valleys.
• Access to both low- and highlands - The economic unit of which the settlement is the archaeologically
most visible part would have included parts of the coastal plain and the uplands as well as the
footslopes. In addition to the summer and winter grazing for livestock, sufficient untended and
wooded land would have been needed for hunting, gathering, fishing, fowling and the extraction of
wood, clay and other natural resources. For example, analysis of the animal bones excavated at
Broglio di Trebisacce indicates hunting on a small scale and animal husbandry of cattle, sheep, goats
and pigs; the livestock was kept for milk, labour, and wool rather than for meat; exploitation of the
coastal environment is evident from the occurrence of tortoise remains (A. Tagliacozzo in Peroni &
Trucco 1994: 56?-652). Social life would have demanded access to cult places such as caves, springs,
and mountaintops. In the Sibaritide the karstic cave sites, e.g. the ones in the S. Marco and Pollino
mountains, are situated upslope of the settlements, which they visually dominate. Kleibrink
(forthcoming) posits a major change in the cultic landscape of the Sibaritide following the Middle
Bronze Age, in that the ancestor worship as practised from at least the Neolithic until the Middle
15 -
3
V A N L EUSEN : P ATTER N T O P ROCESS
Bronze Age in caves, gave way to more public ceremonies in the Late Bronze Age, in which the
leaders were involved in votive deposits of metal implements in lakes, rivers and near springs and
mountain passes. The main infrastructure in the Sibaritide would have consisted of short
transhumance routes following the spines of the major hill systems until joining inland long distance
routes such as the ones postulated by Quilici (De Rossi et al. 1969:59-67); other routes would have
connected the settlements themselves. Kleibrink (forthcoming, par 2.7) suggests that the higher
settlements on calcareous outcrops will have been under the control of the lower, larger, ones and
linked to intensive pastoralist movement.
In all this, a number of other potential settlement location factors have not yet been mentioned: access to
and control over routes along the Crati and Coscile valleys into the hinterland and the opposite
(Tyrrhenian) coast of Calabria; the presence of a favourable microclimate; and the presence of a preexisting cognitive landscape. The significance of access to the coast, at least from the Late Bronze Age
onwards, has been said to lie in the general growth of overseas contacts with the eastern Mediterranean.
And indeed, as Kleibrink writes, the archaeologically attested exchange of objects and technology
indicates that during the Late Bronze Age overseas contacts were frequent (Peroni 1994:24) and
presumably profitable for both the Italic peoples and the traders from the eastern Mediterranean.
However, a more plausible reason may be found in the needs of pastoralism: could access to winter
grazing in the plain, which at this time would have been a heavily wooded and seasonally flooded
marginal area with soils generally too stony or clayey for paleo-technic agricultural use – have been so
vital as to be the determining element in settlement type and location choice? The role of the cognitive
landscape in settlement location is suggested by M Kleibrink (forthcoming, par 4.2), who posits a ‘tabu’
landscape around the San Marco cliffside near Cassano all Ionio, where ancestor worship is said to have
taken place in caves up to and including the Middle Bronze Age. During the Late Bronze Age, while ritual
dedications in caves severely declined, votive deposits of metal objects became much more popular all
over Italy (Bianco Peroni 1978/1979). Evidently the focus of cultic life shifted from ancestors in far-away
places to nature gods and places that could be reached more easily from the settlements.
IMPLEMENTATION
Our database consists of sites listed by Peroni & Trucco (1994); both habitation and cemetery sites are
here taken into consideration. For the Middle Bronze Age (1600 - 1300 BC) there are ca. 17 of these, for
the Recent Bronze Age (1300 - 1150 BC) ca. 19, and for Final Bronze Age/Early Iron Age (1150 - 900 700 BC) ca. 38. As shown in figure 1, in the Middle Bronze Age the Sibaritide foothills - marine and
fluvial terraces consisting of sands and conglomerates – are thought to have been in use for cereal
cultivation and some cattle breeding; in the Recent Bronze Age land use may have shifted back toward
pastoralism tending toward higher elevations, mainly in the interior, a tendency continuing in the Final
Bronze Age. In the Early Iron Age new settlements emerge mainly in the interior, possibly for strategic
reasons but as we shall see this may also be related to a stonger emphasis on agricultural territory. Peroni
(1994, fig. 96) suggests that in this period peer polities developed as well, most of which controlled
territories incorporating sections of the plain, foothills, and upland.
Table 1: Classification of protohistoric settlements in the Sibaritide (D'Angelo & Oräzie Vallino 1994).
15 -
x x x x x x x x x x
x x x x x x x x x x
x x x x
x x x x
x
x
4
x x x x x x
x x x x
29
16
20
19
26
18
17
4
22
25
13
28
10
24
3
32
1
35
27
30
21
14
11
12
6
36
2
34
31
33
23
9
15
8
c
l
a
1
s
2
s
3
4
5
6
5
7
Site number
S E T T L E M E N T L O C A T I O N M ODELS
Underlying Peroni's phase maps is a site type classification by D’Angelo and Oräzie Vallino (in Peroni &
Trucco 1994:827-8), who established the following classes for protohistoric settlement in the Sibaritide:
1. settlement and cultivable land in a well-defended area
2. elevated sites in protected conditions, but with little cultivable space – hence suitable for small
communities only
3. settlement in a well-defended position, but with cultivable unprotected land nearby – some suitable
for large communities, some for small
4. settlement and cultivation possible, but only limited natural defences present
5. sites with properties mainly useful for pastoralists1
6. sites whose main function lies in their viewsheds
Figure 1 - Protohistoric site distributions in relation to
lithotypes, after Peroni 1994, figs. 83, 86, and 96. Top left:
Middle Bronze Age subsistence economy. Top right: Final
Bronze Age animal husbandry. Bottom: Early Iron Age
territories. Open circle: site continued from previous period;
closed circle: new site; cross: site discontinued from previous
period. Oblique hatching: lithoid formations; Vertical
hatching: non-terraced sands and conglomerates; Stippling:
terraced sands and conglomerates suitable for crops; Blank:
recent sediments.
The totals by class are as follows: class 1: 10; class 2: 10, class 3: 4; class 4: 4; class 5: 7; class 6: 6. Classes 1
and 3 would contain the top-ranked settlements, classes 2 and 4 are minor centres, and 5 and 6 are special
1
Not specified, but possibly a location along, or near nodal points in, the network of transhumance routes was meant.
15 -
5
V A N L EUSEN : P ATTER N T O P ROCESS
purpose sites. The table below lists their classification of 36 sites in the Sibaritide. Note that the classes
are not separated by unambiguous criteria – the difference between classes 1 and 4 is in the quality of the
natural defences, that between 1 and 3 in the presence of cultivable land outside rather than inside the
defences, that between 2 and 6 (presumably) in the quality of the viewsheds. In all this, it must be
remembered that the actual archaeological evidence from many of these sites can be as little as a handful
of sherds.
ZONES OF VISUAL CONTACT
On the basis of the distribution of these site types in the landscape, pairs of settlements occupying single
hill systems have been identified, consisting of one larger settlement situated at lower altitude and with
sufficient agricultural land nearby, and one smaller defensible settlement at a higher altitude2. An example
of the former type is the site of Monte S.Nicola, occupying the saddle and sides of two hills overlooking
both the coastal plain and the valley of the Raganello river at an altitude of about 500m asl. If we assume
that the protohistoric settlement pattern was largely based around a pastoral land use pattern (with
transhumance route between summer and winter pastures following the radial ridges and streams of the
Sibaritide, and valleys forming obstacles rather than routes), then sites such as Monte S.Nicola should
have certain viewshed characteristics. In particular, the locations of any hilltops from which, coming from
the summer pastures in the uplands, winter pasture in the plain first comes into view. We can model such
locations by calculating viewsheds from several points in the plain.
The general context for Peroni’s models of protohistoric settlement systems is provided by the
physiography of the Sibaritide, in particular its radial geomorphology and hydrography. Within such a
landscape, areas with similar viewshed properties can be modeled without having recourse to the
locations of known sites and using ‘background’visibility properties instead (see also chapters 6 and 16).
Such areas can be defined by simple criteria and can be organised hierarchically, for example:
• all locations from where a significant part of the coastal plain or the major valley floors can be seen;
this includes the plain and valleys themselves, the edges and slopes of terraces, and the higher slopes
of the Pollino and Sila ranges which face the plain
• all locations from where no part of the plain or valley floor can be seen; this includes the highlands,
secondary river valleys, and the interior of the terraces
To explore such a model, four unrestricted viewsheds were calculated for points lying on the coastline at
the mouths of the Raganello and Crati rivers and at two other points to the north and south3, plus seven
more viewsheds of 10km radius based on four points located on the plain along the base of the foothills
and three points within the major valleys of the Coscile and Crati. When combined, these viewsheds do
indeed define the intended visual zones (see figure 2a). By including known protohistoric settlements in
this model, their degree of association with single zones or, conversely, their liminality with respect to
these zones can be assessed and interpreted. Liminal sites should be located near the edge of a visual zone
but still within the visible area. Protohistoric sites which lie on the outer edge of zone B can be
interpreted as essentially inland sites situated as close as possible to the coastal plain; protohistoric sites
that lie on the outer edge of zone C might be related to transhumance routes and the point where these
begin to descend into the plain; sites on the inner edge of zone C are ‘foothill’sites situated for visual
control of the largest possible area of the coastal plain; and sites that lie within the secluded parts of zone
C are ‘plain’(presumably agricultural) sites with no significant viewshed characteristics. A first inspection
of the model presented in figure 2a suggests that it might indeed be possible to group protohistoric sites
according to such viewshed properties; a better controlled and more detailed study will, however, be
necessary to substantiate this.
2 Authors disagree as to the hierarchical relation between the pair. Peroni regards the higher site of each pair as the main
settlement; Kleibrink the lower. This is not important for the GIS analysis and in the end may be a meaningless difference as
both sites of a pair can be said to form integral parts of one socio-economic system.
3
These points are located at co-ordinates 2650373/4412196; 2647225/4404081; 2651038/4398140; and 2652413/4390424.
15 -
6
S E T T L E M E N T L O C A T I O N M ODELS
Figure 2 - Left: Visual Zones of the Sibaritide, based unrestricted viewsheds from four points on the
coastline, four points along the inner margin of the coastal plain, and three points within the major
river valleys. A: coastal plain and lower slopes, B: highland zone, no visual contact with the plain,
C: foothill zone with areas of visual isolation (dotted lines: terraces and valleys). Boxes/diamonds:
protohistoric sites. Plus signs: Hellenistic/Roman sites. Grid lines represent the boundaries of IGM
topographic map sheets of the 1:25,000 series 25V. Right: 25km radius viewshed from the
important protohistoric settlement site of Torre Mordillo.
The utility of the concept of visual zones can be further explored by comparing them to the viewsheds of
individual protohistoric sites. Figure 2b presents one such viewshed, calculated from the site of Torre del
Mordillo, probably the most important indigenous settlement at the time of the first Greek colonisation4.
It can be seen that the sites has a very large viewshed, covering both the coastal plain and much of the
major inland valley floors and slopes. Because of its relatively elevated position on the rim of a marine
plateau, its viewshed also includes several of the ‘secluded’areas not visible from lesser elevations.
When the locations of Hellenistic/Roman farmstead sites in the Quilici dataset are included in the model
as well, it becomes apparent that the linear clustering observed by Quilici is related to specific
geomorphological settings, for which explanations may be sought not just in viewshed properties but also
in microclimatic variations. Finally, the significance of zones which tend to be ‘hidden’from most of the
plain and valleys (or conversely, from where these areas cannot be seen; indicated by dotted lines in figure
2a) could be further explored. Examples of such areas are the plateaux of Caccavato/Praineto north of
the Coscile river and Lauropoli in between the Coscile and the Raganello, and the valley of the Eianina at
S. Marco. Interestingly, Kleibrink in a forthcoming article suggests on the basis of other evidence that the
latter area might represent a protohistoric ‘tabu’landscape (Kleibrink, forthcoming).
THE ALBAN HILLS
The sites and monuments of the Alban hills were the subject of spatial studies by topographers early on.
Settlements, cult places, and ‘tombe principesche’from the later Iron Age onwards were related to
historical and infrastructural evidence, and interpreted in the context of the ‘formazione delle città’, the
4
By reason of its size and commanding position on a plateau overlooking the confluence of the Crati and Coscile rivers.
15 -
7
V A N L EUSEN : P ATTER N T O P ROCESS
formation of early city states. Such studies have continued more recently (cf. Bietti Sestieri 1985,
Chiarucci 1996, Arietti 1996), with the emphasis shifting towards models of the spatial organisation of the
indigenous Latial societies. The spatial organisation of the landscape of the Pontine region has most
recently been investigated diachronically in five phases between 700 and 300 BC by Bouma and Van ‘t
Lindenhout (1998) with the help of Thiessen polygons. These authors identify a general diachronic trend,
beginning in the Tiber Valley and continuing in the Pontine region, of centralization towards a smaller
number of increasingly urbanised settlements and an attendant simplification in Thiessen polygons.
Bouma and Van ‘t Lindenhout conclude that the Iron Age Latial system of peer polities was still intact
during the Archaic period (6th century), and began to collapse only towards the end of the 5th.
As in the single-period models advanced by earlier writers, this diachronic approach is based on some
very shaky assumptions regarding the status and contemporaneity of (proto-) urban polity centres. The
dangers involved can be illustrated by these authors’discussion of cult places in the context of their
location in a territorial centre, on a territorial boundary, or inside a territory (1998:97-100). In particular,
there is a potential circularity of argument involved in the fact that any particular set of Thiessen polygons
is the consequence of a decision to regard a certain group of sites as ‘equal’within a settlement hierarchy;
the polygons or their characteristics cannot then be used to prove that this was the case. More-over, their
argument that the presence of a cult place in a settlement is supporting evidence for its function as a
polity center is weak because it assumes that the absence of cult places is not due to the chances of
discovery. Again, their conclusions are often based on the appearance or disappearance of single centres
in the network of polygons, and so are very sensitive to the chance presence or absence of occupation
evidence for any single period. This sensitivity of Thiessen polygons to changes in the set of ‘seed’
settlements is demonstrated by overlaying Franco Arietti’s alternative hypothetical territorial model of the
central Alban area in the later Iron Age on that of Bouma and Van ‘t Lindenhout (Arietti 1996, fig. 3):
applying different criteria to decide which protohistoric sites were territorial centers, Arietti adds the
centers of Labigi, Lanuvium and Alba Longa (postulated at Castel Gandolfo) to the list used by Bouma
and Van ‘t Lindenhout with obvious consequences for the sizes and shapes of the resulting territorial
division (see figure 3).
Figure 3: Hypothetical protohistoric
territories of the Alban hills. Thiessen
polygons and central places after Bouma &
Van ‘t Lindenhout 1998, fig. 2-5 (in black)
and Arietti 1996, fig. 3 (in grey).
Topography: B Fidenae, C Antemnae, D
Rome, E Laurentina/Acqua Acetosa, G
Castel Decima, H Lavinium, I Frattocchie, J
La Rustica, M Gabii, N Tusculum, O
Ariccia, P Ardea, T Velletri, S Cisterna, U
Cori, V Palestrina, Y
Caracupa/Valvisciolo, 1 Labigi, 2 Castel
Gandolfo, 3 Lanuvium.
15 -
8
S E T T L E M E N T L O C A T I O N M ODELS
Whilst Thiessen polygons can be 'weighted' to reflect the differing sizes or populations of the polity
centers in a peer polity system, they were never intended to model the territories of a hierarchical set of
centers. For such a case, central place theory and X-tent models (Renfrew & Level 1979) are more
appropriate. Guidi (1985), applying a rank-size analysis5 to the settlements of south Lazio in four
chronological phases (corresponding to the 10th, 9th, 8th, and late 7th centuries BC), estimates settlement
size from the area bounded by natural defenses. For 7th century south Lazio he concluded (1985:232) that
there was evidence for the development of a three-level hierarchy with Rome at the top, Ardea and Gabii
at level two, and other towns dependent on these three at level three. If the large number of minor sites
for which no size data are available were included, he argues, then the rank-size graph would take on a
‘primoconvex’shape which indicates a high level of integration for the larger sites but a low level for the
smaller ones. These results suggest that the assumptions underlying the territorial analysis by Bouma &
Van 't Lindenhout and Arietti are not well supported. Given the lack of discussion of these issues, it
appears that these authors are not aware of the extreme sensitivity of the Thiessen polygon technique to
even small changes in the input data set.
3
ROMAN COLONIES OF THE LEPINE SCARP
As the influence of Rome over affairs in the Pontine region grew in fits and starts during the post-Archaic
period, and intermittent conflict with neighbouring tribes became more disruptive, so the development of
the indigenous Latial peer-polity system was replaced by one of a core-periphery system in which the
Pontine plain first became the scene of a drawn-out conflict between the expanding early Roman state
and the rather less clearly defined, but equally expansive, hill tribes, and later that of Roman demographic
and agricultural expansion. Attema (1993:231) suggested that the Roman colonies of Cora, Norba, and
Setia may have played an important role in the later Republican ‘colonisation’of the Lepine side of the
Pontine plain. The presence in this area of a large number of so-called ‘platform’villas which appear to be
of very similar date and design argues, he wrote, for a planned process of agricultural re-organisation and
exploitation, probably targeted at the production of olive oil and grain for the Roman market. The
position of the colonies themselves, located on the rim of the Lepine pre-mountains with magnificent
views across the Pontine plain to the sea and along the coast as far as Antium and the Monte Circeo,
expresses the control exercised over this agricultural area.
HISTORIC-LITERARY REFERENCES
However, these towns and other ones existed long before the late Republican period, and in order to
understand why they are located on the Lepine margin we have to trace these origins as far back as we
can. The main source for information about this early period is Livy’s Ab Urbe Condita, from which we
can gather the following information about each of these towns:
• Anxur (Terracina) - was captured and sacked by the Romans in 406 BC during precautionary
campaigns against the Volscans (IV 59). Four years later its garrison was overrun again by the latter,
re-taken by the Romans in 401 BC, but again under Volscan siege in 398 BC (V 8-16). In 329 BC
Rome sent 300 colonists to the town, each getting 2 iugera of land (VIII,21).
• Cora (Cori) - a Latin town, Cora joined the Aurunci against Rome in 503 BC, but was quickly defeated
by her (II 17). By 496 BC the town was under Volscan control because these were forced at that time
to send 300 hostages from Cora and Pometia to Rome as pledges against attack (II 22). The territory
of Cora was raided by the Privernates in 331 BC (VIII,19).
Rank-size analysis (Zipf 1949) can be used to examine the degree of socio-economic integration of a settlement system by
setting out the sizes and ‘ranks’(based on size and ordered by rank) of all settlements in a graph. An idealised rank-size graph will
have a log-normal shape; deviations from this line indicate higher (concave graphs) or lower (convex graphs) levels of integration.
5
15 -
9
V A N L EUSEN : P ATTER N T O P ROCESS
• Circeii - in 491 BC Volscans led by an exiled Roman expelled the Roman settlers sent there in 510 BC
by Tarquin (see Signia), ‘liberated’the town and handed it over to Volscan control (II 39).
• Anagnia (Anagni) and Ferentinum (Ferentino) - east of the Sacco (Trerus) river, Ferentinum was taken
by Rome in 412-411 BC, and given to her Hernician allies (IV 52). In 308-7 BC the Hernici of
Anagnia declared war on Rome but those of Ferentinum did not. The former were quickly defeated
(IX 42).
• Norba (Norma) and Setia (Sezze) - taking advantage of the weakened state of the Volscans following
an epidemic, Rome sent out fresh settlers to Norba in 493 BC, which became a fortified point for the
defense of the Pontine region (II 34). Setia was in Volscan hands (Dion. Hal. 6.61) before Rome sent
a colony there in 382 BC (Vell. Paterc. 1.14). The territories of Norba and Setia were raided by the
Privernates in 358 BC (VII,15), 342 BC (VII,42), and 331 BC (VIII,19).
• Privernum (Priverno) - Latial allies, the Privernates continued to raid the territories of neighbouring
Roman colonies in 358 BC (VII,15), 342 BC (VII,42;VIII,1), and 331 BC (VIII,19). Although
relatively quickly defeated by Roman armies each time, and deprived of two thirds of their territory in
retaliation (VIII, 1), only the siege and capture of Privernum in 330 BC (VIII, 19-20) appears to have
finally ended the conflict.
• Signia (Segni) - in 510 BC the Roman king Tarquin sent some surplus population out to Signia and
Circeii, both to increase Roman territory and to provide points of resistance to attack ‘by land or by
sea’(I 56). This appears to have not been a very successful venture, because already in 496 BC the
town had to be re-established with additional settlers (II 22). Inhabitants of the town attack fleeing
Hernici after their defeat by a Roman army in 362 BC (VII,8). Still a Roman colony in 340 BC, Signia
was apparently ruled by Latin allies (VIII,3).
From this brief overview it becomes apparent that no historic evidence predates the very end of the 6th
century; that none of the towns mentioned was actually established by colonisation from Rome; and that
most or all were therefore pre-existing Archaic and (by analogy with other areas) Iron Age settlements.
Late Iron Age and Archaic hilltop settlements such as certainly existed at Signia, Circeii, Norba, and Cora
were populated by a patchwork of indigenous lowland and highland tribespeople, apparently maintaining
some kind of political equilibrium punctuated by low-level raiding, presumably for cattle and prestige (see
also examples cited in Attema, in press). This would explain why these settlements would have been
located in places which afforded both safety and control over land and cattle.
Starting with the post-Archaic, groups of Roman colonists were sent out in an opportunistic manner to
safeguard Rome’s political and military interest. The towns of the Lepine margin, ‘outposts’from the
viewpoint of Rome, bore the brunt of the conflicts with the Volscan tribes which lasted throughout the
5th and the first half of the 4th century BC. More than once their allegiances swerved from safety under
Roman hegemony to independence of it; in addition to this, some Latial tribes continued to raid each
other’s territories, as is shown most clearly by the case of Privernum which, from its southerly position
may have felt itself to be as much akin to the Volscan way of life as it was to that of the Latial League.
While the objectivity of Livy’s accounts may be questioned, it seems clear that the conflict between Latins
and Volscans is acted out on the medium term, the conjoncture as defined by Braudel, and can be
understood perhaps in terms of the upland boom-bust cycle cited by Bintliff (1997:30-32; cf. chapter 2).
Most of the Lepine margin must have been effectively incorporated into the Roman state by the mid-4th
century when, in 358 BC (VII,15) she added the Pomptine and Publilian tribes – territorial units in which
citizens were enrolled for census, taxation, and military levies – and, following the final defeat of the
Privernates and the settling of their territory with colonists, the Oufentine tribe in 329 BC. By the end of
the 4th century the military Via Appia was completed as far as Anxur. The platform villas identified by
Attema (1993) appear only after this de facto incorporation into the Roman state was completed.
15 -
10
S E T T L E M E N T L O C A T I O N M ODELS
VIEWSHED ANALYSIS
In view of the above, the long-term Roman ‘policy’was not to establish colonies on the Lepine margin, but
to ensure that the important central places became or remained allied to her, an allegion that could at
times be strengthened by sending out colonists for reasons as much to do with demographic pressure at
Rome as with strategic interests (providing early warning and protection from Volscan raiding parties and
containing a local population which could not be trusted to choose Rome’s side in a conflict). It is not
unlikely that both sets of factors combined to determine which 4th century sites were deemed to be most
important.
The 4th century BC Roman colonies of the Lepine scarp provided bases from which both agriculture and
husbandry in the plain and uplands could be protected from Volscan inroads. But they also acted as visual
manifestations of Roman power in the lands of her former Latial allies. Their viewsheds might therefore
include areas in both the plain and the upland; especially the Lepine mountain passes from which raiding
parties might arrive. At the same time they must be positioned close by valuable cropland and grazing
herds to be able to protect these against sudden attack. Hence, if we model viewsheds for these colonies
we must take into account upland characteristics such as the location of mountain passes as well.
Figure 4: Higuchi viewsheds of three Roman colonies at Cori, Norba, and Sezze, and of protohistoric
Priverno on a shaded elevation model of the Lepine mountains. Black line: Via Appia. Higuchi
radius of 6600m indicated wby white broken line. Grid spacing: 10 km.
Higuchi viewsheds were recently introduced in archaeological research by Wheatley and Gillings (2000) as
a way of enriching traditional viewshed studies. Amongst other characteristics, Higuchi proposed
15 -
11
V A N L EUSEN : P ATTER N T O P ROCESS
that viewsheds should contain information about the distance and bearing to the objects in view. With
respect to distance, Higuchi viewsheds are subdivided into a short range (< 360 m) sector in which
objects are individually distinguishable and have a direct sensory impact, a middle range (360 – 6600m)
sector which constitutes the ‘pictorial’landscape where vision is paramount, and a long range (> 6600m)
sector which contains the ‘vertical backdrop’and horizon features. The distances at which each of these
sectors begin and end are variable, because they are relative to the typical tree size for the area under
study, but in the following description of Higuchi properties of the three Lepine colonies I simply use
Wheatley and Gillings’figures. Their middle range viewsheds are depicted in figure 4.
Cora is located on the western edge of the Lepini where it borders on the volcanic landscape of the Alban
hills, on top of a small hill situated at the mouth of a small drainage basin (391 m asl). Its views to the
north and southeast are obstructed by neighbouring higher hills. A viewshed from 2346428,4612210,400
shows that the whole of the Alban massif and the Pontine plain as far as Monte Circeo over 45 km to its
south can be seen from Cora, and its direct hinterland, up to 6 km distant, is also relatively well covered.
The Roman colony of Norba is situated on a promontory of the Lepine scarp, with steep slopes on three
sides but open to the interior, where several small streams and their tributaries form a modest agricultural
hinterland before descending into the plain at Valvisciolo. The highest point on this promontory, the
‘acropolis’hill, is at 490 m asl. The viewshed taken from this approximate point, which is located almost
600m from the Lepine scarp (2350125,4606607,492), is especially large toward the east and south-east,
and again the view across the Pontine plain includes both Monte Circeo and the Alban hills. However,
from this location one cannot see the nearby footslopes and valleys of the Brivolco stream system. As
field observations have shown that long stretches of the lower Lepine slopes are visible from its perimeter
wall, an improved Norba viewshed model should clearly be based on multiple viewpoints along its
perimeter6. Setia is located on a small hill next to the place where the Fosso Brivolco descends into the
plain, at about 280m asl (no elevations mapped within the town; over 250m above the plain), and is
naturally protected on all side by steep slopes. A viewshed taken from the approximate location of its
central church and a height derived from that of a neighbouring hilltop to the southeast
(2358047,4595884,310) extends into the upper valley of the Brivolco and into the pass leading east past
Roccagorga, all within 5 kms of the town. There is no viewshed due east. Toward the sea, there is an
almost 180 degree unrestricted view taking in the Pontine plain and coastal landscape, with the Monti
Ausoni and Circeo at up to 30 kms distance as a backdrop; however, views in both directions along the
Lepine scarp are restricted by the hills directly to the west and east, and the characteristic shape of the
Alban massif is not visible from the town.
The most easily accessible route between the Pontine plain and other parts of Italy to the east and
southeast is through the Lepini via the valleys of the Amaseno/Oufente. These valleys are surrounded by
several hilltop settlements probably dating back to the Iron Age: Roccasecca dei Volsci, Maenza,
Roccagorga and Priverno among them. A viewshed from Privernum (2368018,4592839,139 and 160) was
used to set off the other three views. It turns out that all of the major valleys here can be seen very well
from this location, even up to the pass between the Lepini and Ausoni, leading to Campania. There is no
view into the Pontine plain and, possibly significantly, no significant visible upland within the Higuchi
distance which the Romans might have needed to control by the installation of a colony following the
siege and capture of Privernum in 330 BC.
Summing up, the colonies have an excellent view of the plain, including the both the Via Appia and the
centuriated agricultural zone along it, but the viewsheds do not indicate that the lower Lepine slopes and
the ancient pedemontana route along it were of immediate concern. Toward the hinterland the colonial
viewsheds are complementary and mutually exclusive (that is, together they cover the whole of the
western side of the Lepine mountains, but they do not overlap); these viewsheds, and the fields and
pastures within them, are mostly within the Higuchi distance of 6600m and would therefore have
afforded strong visual control over the whole area. The hypothesis that the 4th century BC Roman
colonization in the Pontine region was mainly strategic in nature is therefore upheld.
6
As was done in the case of the Roman fortress and town at Wroxeter, see chapter 16.
15 -
12
S E T T L E M E N T L O C A T I O N M ODELS
4
CONCLUSIONS
A general conclusion that can be drawn from the case studies presented here, is that current economic
and cognitive models of the ordering of settlements and the landscape in protohistory (and, for
indigenous societies, even for many centuries afterwards) are of a very non-specific and intuitive nature.
Bias modeling (chapter 6) and corrective fieldwork (chapter 8) will be needed to test the many
assumptions on which these models are based.
• The developed Iron Age settlement pattern in the Sibaritide and Salento Murge displays remarkable
similarities in the geomorphological location and spacing of the settlements, located some 10-12 km
apart in defensible hilltop positions. It must be doubted that regular access to sea-born trade had a
major role to play in this, because the pattern continues into the Murge upland, and may in fact be
more strictly related to control over high quality agricultural and pastoral resources. This hypothesis
can be tested by targeted fieldwork in the Lepine uplands and the inland reaches of the Sibaritide.
• Current typologies of protohistoric settlements in central and southern Italy are insufficiently clear,
detailed, or supported by evidence to allow the definition of hierarchical and contemporaneous levels
with any degree of certainty. The basis for constructing territorial divisions, whichever method is
chosen for it, is therefore lacking.
• The ‘colonial’settlement pattern in southern Italy was centred on the coast rather than on the hill
country, and combined accessibility by sea with the presence of a substantial agricultural hinterland.
In contrast, Rome’s early colonies were as much or more intended to fulfill strategic functions, so
their locations meet other criteria of dominance – namely that of control over routes of attack and
advance. The fact that the viewsheds of the Roman colonies on the Lepine margin are both
complementary and fall within the Higuchi 'middle range' distance creates support for the idea that
these towns were located as much to control movement across the Lepine up- and highlands, as to
control and protect communications and agricultural resources in the Pontine plain.
• As a tool for archaeological spatial analysis of territories, Thiessen polygons have been used
extensively. The case study presented in section 2 demonstrates the weaknesses of the technique in
specific archaeological situations. The use of GIS and cost surface analysis allows the technique to be
refined by replacing the simple gravity model of space with one in which each centre can have its
own ‘weight’determining the relative size of its polygon, and in which characteristics influencing the
accessibility of the terrain are used to determine the location of territorial boundaries instead of
horizontal distance. Rank-size studies such as the one by Guidi (1985), although based on unreliable
settlement sizes, when combined with X-Tent modeling techniques provide a more credible
alternative to Thiessen polygons; another advantage is that they can be used to implement central
place models as well as peer polity models of society.
REFERENCES
Arietti, F 1996
Gli Albani e il loro territorio nell’VIII e VII secolo a. C., in Pasqualini, A (ed), Alba Longa, mito storia archeologia.
Atti dell’Incontro di studio Roma – Albano Laziale 27-29 gennaio 1994: 29-48. Rome: Istituto Italiano per la storia
antica.
Attema, PAJ 1993
An Archaeological Survey in the Pontine Region. A contribution to the early settlement history of south Lazio 900 - 100 BC. 2
Vols. PhD thesis, Archeologisch Centrum Groningen.
Attema, PAJ in press
Landscape archaeology and Livy: Warfare, colonial expansion and town and country in Central Italy of the 7 th
to 4 th c. BC, BaBesch 75 (2000): 115-125.
Barker, G 1985
Prehistoric Farming in Europe, Cambridge, New York, etc.
15 -
13
V A N L EUSEN : P ATTER N T O P ROCESS
Bianco Peroni, V 1978-79
Bronzene Gefässer und Hohenfunde aus Italien, Jahresbericht des Vorgeschichte der Universitaet Frankfurt a.M. 197879: 321-35.
Bietti Sestieri, AM 1985
The Iron Age community of Osteria dell’Osa. A study of socio-political development in central Tyrrhenian Italy (New Studies in
Archaeology). Cambridge University Press.
Bintliff, J 1997
Regional Survey, Demography, and the Rise of Complex Societies in the Ancient Aegean: Core-Periphery, NeoMalthusian, and Other Interpretive Models, Journal of Field Archaeology 24(1):1-38.
Bouma, JW & E van 't Lindenhout 1998
Light in Dark Age Latium. Evidence from settlements and cult places, Caeculus III: 91-102.
Burgers, G-J 1999
Antieke koloniale situaties in Zuid-Italië, een regionaal perspectief, TMA 21:19-26.
Chiarucci, P 1996
Viabilità arcaica e luoghi di culto nell’area Albana, in Pasqualini, A (ed), Alba Longa, mito storia archeologia. Atti
dell’Incontro di studio Roma – Albano Laziale 27-29 gennaio 1994: 317-334. Rome: Istituto Italiano per la storia
antica.
D'Andria, F 1999
Ricerche recenti sugli insediamenti indigeni di Puglia e Basilicata, in: La Forma della Città e del Territorio, Esperienze
metodologiche e risultati a confronto, Atti dell'Incontro di studio - S. Maria Capua Vetere 27-28 novembre 1998: 103-118.
Roma.
D’Angelo, S & F Ch Oräzie Vallino 1994
La Sibaritide. Lineamenti geografico-ambientali ed insediamento umano, in Peroni, R & F Trucco (eds)
1994:785-829.
De Rossi, GM, L Pala, L Quilici & S Quilici-Gigli 1969
Carta Archeologica della piana di Sibari, extract from Atti e Memorie della Societa’Magna Grecia, Nuova Serie IXX (1968-1969): 91-155. Roma: Societa Magna Grecia.
Guaitoli, M 1977
Considerazioni su alcune citta ed insediamenti del Lazio in eta protostorico ed arcaica, Römische Mitteilungen 84:526.
Guidi, A 1985
An application of the rank-size rule to protohistoric setlements in the middle Tyrrhenian area, in Malone, C & S
Stoddart (eds), Papers in Italian Archaeology IV, part iii: Patterns in protohistory (BAR S245):217-242.
Kleibrink, M forthcoming
The sacred landscape of the Sibaritide – the veneration of ancestors, nymphs and deities, in Attema, PAJ, G-J
Burgers, E van Joolen, PM van Leusen & B Mater (eds), Regional Pathways to Complexity. Proceedings of the RPC
conference, Groningen 2000 (BAR Int Ser).
Levi, ST 1999
Produzione e circolazione della ceramica nella Sibaritide protostorica. 1. Impasto e dolii. Firenze.
Peroni, R 1994
Introduzione alla protostoria italiana. Rome-Bari.
Peroni, R 1996
L'Italia alle soglie della storia. Roma: Laterza.
Peroni, R & F Trucco 1994
Enotri e Micenei nella Sibaritide, Taranto.
Rendeli, M 1993
Città aperte: ambiente e paesaggio rurale organizzato nell'Etruria meridionale costiera durante l'età orientalizzante e arcaica.
Roma: Gruppo editoriale internazionale.
Renfrew, C & EV Level 1979
Predicting Polities from Centers, in Renfrew, C & KL Cooke (eds), Transformations. Mathematical Approaches to
Culture Change: 145-167. New York, etc: Academic Press.
Vanzetti, A forthcoming
Results and problems of some current approaches to protohistoric centralization and urbanization in Italy, in
Attema, PAJ, G-J Burgers, E van Joolen, PM van Leusen & B Mater (eds), Regional Pathways to Complexity.
Proceedings of the RPC conference, Groningen 2000 (BAR Int Ser).
Wheatley, D & M Gillings 2000
Vision, Perception, and GIS: developing enriched approaches to the study of archaeological visibility, in Lock,
G (ed) 2000, Beyond the Map: Archaeology and spatial technologies: 1-27. Amsterdam, etc: IOS Press.
15 -
14
C HAPTER 16
WHP CASE STUDIES IN VISIBILITY
AND FRICTION∗
Two of the case studies presented here were originally developed in the context of the study of
Romanisation and urbanisation in the Wroxeter Hinterland, an area centring on the modern-day village of
Wroxeter in the middle Severn valley (Shropshire, UK); the third case study arose from my work on the
methodology of ‘cognitive landscape’analysis presented in chapter 6 of this thesis. All three are presented
here together not just because they cover the same geographic area, but also because the GIS techniques
on which they are based – line-of-sight and friction modelling – are related and tend to be used for
answering related archaeological questions (for a full technical discussion of these techniques and
questions and a review of the relevant literature, see chapter 6). For an introduction to the Wroxeter
Hinterland Project, see chapter 3 of this thesis; aspects of centralisation and Romanisation in the
Wroxeter hinterland have been sketched elsewhere by White and Van Leusen (1997) and again by White
and Barker (1998).
1
V I S I B I L I T Y A N D C O N TROL
Some aspects of the Iron Age – Roman transition within the territory of the Cornovii can be modelled
using only the highest-ranked settlements of either period. Cornovian society, especially in the later preRoman Iron Age, is thought to have become increasingly sophisticated and to have been dominated by an
aristocracy based on control over land, livestock, and mineral resources (especially salt). The Wroxeter
hinterland is well supplied with hillforts (some 40 in all if we include the ones that lie just outside the
WHP study area; most are presumed to date to the Iron Age although only a few have been investigated),
which has been taken to indicate that the tribe was politically fragmented and was organised in clans
around chiefs. However, an alternative view now takes ground (White & Barker 1998:36) that the hillforts
are expressions of conspicuous consumption in a society that had few other outlets for its wealth.
Whichever the case may have been, certainly the hillforts would have functioned as places of refuge and
control, and viewsheds from these hillforts may therefore tell us something about systems of control and
defence in the pre-Roman Iron Age.
∗
These case studies were prepared in 1996-7 as part of the Wroxeter Hinterland Project (WHP), directed by Vince Gaffney at
the University of Birmingham Field Archaeology Unit (BUFAU). They are based on pre-Conquest digital site data supplied in
December of 1996 by Ms Penny Ward of the Shropshire County Council on the basis of the Shropshire Sites & Monuments
Records. The data were subsequently checked and enhanced for the WHP by my colleague Roger White. I am particularly
grateful to Dr Gaffney, who set out many potential lines of research for me to follow and who himself with a student developed
models for the urban resource landscape around Wroxeter (Goodchild 1999). It should be noted that visibility/accessibility
modeling has moved on since these case studies were first conceived, and chapter 6 should be consulted for more recent work in
this area. Also note that DEM interpolation artifacts, visible in figures 16.3 to 16.6 as stripes or ‘steps’,have not been removed
before the analysis. To implement a decision as to whether sites should be visible in these areas, the individual viewshed maps
can be put through a simple neighborhood filter.
16 -
1
V A N L EUSEN : P A T T E R N T O P ROCESS
The indigenous ordering of the landscape of Britain was upset from the middle of the 1st century AD by
Roman military encroachment. The land of the Cornovii was first invaded by the Romans by the end of
the 40s (AD), and the tribe seems to have come to terms with the conquerors without putting up
significant resistance. While the main Roman force arrived in Cornovian territory north of the Wrekin, a
smaller force may have followed the Severn valley from the southeast and put up a vexillation fortress on
the Severn at Leighton just south of the Wrekin; archaeological evidence indicates that the Wrekin hillfort
was attacked and taken from there. Several temporary campaign forts were constructed in the area of
Wroxeter, which controlled the main routes across the Severn, in the following years. One of these is an
auxiliary fortlet on the Severn just south of Wroxeter which may have secured the main Severn crossing;
by the mid-50s the legionary fort at Wroxeter itself was established, probably by Legio XIII Gemina from
Mancetter (Warwickshire). The early Roman military strategies may be studied via viewshed analysis of
both the vexillation fort at Leighton, the legionary fortress at Wroxeter itself, and the auxiliary fort to its
south. As the legionary fortress developed into a town and civil civitas capital after 30 years of
predominantly military use, its viewshed may tell us much about its impact within a landscape that had
never seen such a population centre before.
1.1
IMPLEMENTATION
We have prepared a similar analysis for the WHP area by regrouping the traditional types of hillforts and
multivallate enclosures into more meaningful sets of large (over 2 hectares) and small (less than 1.5
hectares) multivallate enclosures (see table 1; for a more detailed discussion see section 2 on cost surface
analysis).
Table 1: 21 multivallate enclosures of the Wroxeter Hinterland, ordered by size. PRN: Primary
Record Number. Source: Shropshire County Council.
PRN
Name
Situation
Enclosed area (ha)
1108
1357
113
129
1438
1050
1069
226
357
60
135
1087
2000
1055
3970
1740
1048
2418
1256
472
2828
Wall Camp
Castle Ring
Ebury Hillfort
The Berth
Stevenshill
Earls Hill Camp
Wrekin camp
Caer Caradoc
The Ditches
The Burgs
Haughmond Hill camp
Nesscliff Hill Camp
Hurley Brook rect. enc.
Pontesford Hill Camp
Nills Hill
Callow Hill Camp
Bomere Heath
The Lawley, north
Cotwall No. 1
"British Camp"
marsh
hilltop
low hill
marsh
promontory
hilltop
hilltop
hilltop
hilltop
low hill
hilltop
hilltop
no hill
low spur
no hill
low spur
hilltop
no hill
hilltop
low hill
no hill
very large (14)
large (3.8)
large (3.6)
large (3.1)
large (3)
large (1.4), with annexe (1.6)
large (2.6)
large (2.6)
large (2.4)
large (2.1)
large (2)
small (1), with annexe (1)
small (1.2)
small (1.1)
small (0.75)
small (0.4)
small (0.4)
small (0.25)
very small (0.15)
very small (0.15)
unknown
Adding the viewsheds of sites within these two groups to obtain the cumulative viewsheds (see figures 1
and 2), it is clear that the areas most intensively viewed are all in the central upper Severn valley and its
main tributaries, with the maxima occurring on the western side of the Severn. These results are mildly
helpful in interpreting the results of the Thiessen polygon calculation, which argue for a system in which
three pairs of hillforts are spaced along the main basin, dominating opposite sides of it (see section 2).
The site of Wroxeter is in fact very near the point where four of these territories meet, making it ‘neutral
territory’. As it is also near one of the main Severn fords, we suggest that this location was well suited to
function as a (periodic?) trading post/market/fair, and forms a logical precursor to the legionary fortress
and town.
16 -
2
CASE STUDIES IN VISIBILITY AND FRICTION
Figure 1: cumulative 15km radius viewsheds of large multivallate enclosures (red diamonds) on
shaded DEM overlain with major streams and Roman road system (white lines). 8 by 1 km box in
the centre of the study area indicates zone which cannot suffer from edge effects. In this and all
further figures, grid spacing is 10 kms unless otherwise stated.
Figure 2: cumulative 15km radius viewsheds of small multivallate enclosures (red diamonds) on
shaded DEM overlain with major streams and Roman road system (white lines). 8 by 1 km box in
the centre of the study area indicates zone which cannot suffer from edge effects..
16 -
3
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 3: the multiple viewshed from the fortlet (5m viewing height, green) is almost completely
subsumed (yellow) into that from the fortress (7m viewing height, red). The partly reconstructed
pattern of Roman roads is overlaid.
Roman military campaigns into the region used two major routes, one from Greensforge in south
Staffordshire following the southern bank of the Severn via Morvill and Much Wenlock, crossing the
Severn near Cressage, the other from Mancetter following the later Watling Street via Red Hill and
curving north of the Wrekin. The precursor auxiliary fortlet to Wroxeter, discovered by aerial
photography and subsequent partial excavations (St Joseph 1951, Houghton & Wells 1978), is located not
on the elevated site of Wroxeter itself but nearly a mile to the south, right on the bank of the Severn and
some 15 metres lower. Given the military purpose served by this fortlet and the later fortress, viewsheds
based on them may well tell us what they were intended to control (see figure 3).
As expected, the viewshed from the fortlet is much smaller than that from the fortress. Whereas the
fortress, like the later town, has an uninhibited viewshed over two-thirds of the compass, the auxiliary
fort’s view is limited to just over half the compass. What is more, the fortress viewshed nearly completely
encompasses the fortlet viewshed, so whatever the reason was for placing the auxiliary fort where it is, it
cannot have been the viewshed. We may therefore speculate that the fortress was placed directly on the
river bank for tactical reasons (campaigning across the Severn) rather than strategic ones (control of
movement in the area).
In a separate analysis, a series of viewsheds were calculated from Wroxeter in order to explore its relations
with the known hillforts in the study area. In order to circumvent the problem of the low resolution (50
metres) DEM, the viewing position was chosen at 5 metres above ground level at the highest point within
the town walls. The result is depicted in figure 4 above and shows that the bulk of the effective (ie
ignoring small patches and far off hillsides) viewshed is to the west (from due N to due S) of Wroxeter,
and extending some 7.5 kms from the town in those directions. In order to explore the possibility that the
viewshed from Wroxeter might be enlarged when taking into account multiple viewing points (watch
towers) along the town walls, another viewshed was calculated using the whole of the town walls as the
seed area, and setting the viewing height at 5 metres (figure 5). We find that the viewshed is enlarged by
48.5 km2 (a 43% increase over the 112.4 km2 viewshed of figure 4), to cover areas to the northeast and
directly across the Severn to the west and southwest of the town. The viewshed for the legionary fortress
preceding the town is essentially identical to the latter. Most of the hillforts within a 15 km radius,
whether they were occupied during this period or not, are found to lie within this enlarged
16 -
4
CASE STUDIES IN VISIBILITY AND FRICTION
viewshed. However, since hillforts that posed a threat to the Romans were forcibly abandoned after the
Conquest, it is not clear that these results have any significance beyond that which was already proven,
namely that the location of Wroxeter was in common view and therefore a good ‘neutral’place to hold
markets.
Figure 4: unrestricted viewshed from the highest viewing point within Wroxeter. Red diamonds: large
multivallate enclosures.
Figure 5: 15 km max viewshed from Wroxeter wall circuit. Red diamonds: large multivallate
enclosures.
16 -
5
V A N L EUSEN : P A T T E R N T O P ROCESS
1.2
DISCUSSION
From figures 4 and 5 it is not immediately clear how many of the 13 large multivallate enclosures within
the study area fall within the Wroxeter viewshed. In fact, both in the unrestrained viewshed of figure 4
and the 15 km radius viewshed of figure 5, only 4 out of 13 are ‘visible’. Field observations have shown
this result to be incorrect for at least some of the hilltop locations, and point out a weakness in the lineof-sight technique employed: locations on the visible horizon are not reliably included in the viewshed.
One possible technical solution (not pursued here) might be to expand the viewshed by the addition of
‘horizon’cells; these can be identified by the fact that they a) must lie next to cells that are within the
viewshed, and b) must be further away from the viewing point than any neighbouring viewshed cell.
Overlaying the partly reconstructed pattern of later Roman roads on these maps of indigenous hillfort,
Roman military, and Roman civilian viewsheds, we can also observe that many sections of road fall within
the viewsheds. However, this appears to be due to the generally favourable position of Wroxeter within
the bowl-shape of the Severn valley rather than to any conscious decision to build the roads in such a way
that they would be visible from the fortress and town. Simulation studies will be presented in section 3
below to support the weakness of the statistical arguments generally adduced for deliberate placement of
archaeological feature (see also the example of the Arroux valley Celtic hillfort system discussed in
chapter 6, Madry & Rakos 1996).
2
STRUCTURATION OF T HE LANDSCAPE
The social, political, and economic organisation of space in the Wroxeter hinterland area throughout the
Late Iron Age and the Roman period can be studied from many angles, some of which are amenable to
GIS analysis. Specifically, the distances and effort involved in travel and transport can be studied through
cost surface analysis (CSA), and have ramifications into such areas as political control, economic spheres
of influence, and the social ordering of space. CSA techniques may be used to implement, and improve
upon, some classical types of archaeological analysis, including buffer, cluster and distance analysis,
tessellation of space, and site catchment analysis. They have also been used to explore some entirely new
concepts, relating to landscape accessibility and optimal routes and networks. In chapter 6 I have
reviewed the literature and current approaches on this subject; here I will present case studies based on
WHP data and problems.
2.1
CATCHMENTS AND T ERRITORIES
Although, in archaeological theory, site catchment analysis is a quite complex and flexible concept, the
definition of actual catchments (exploitation zones) has been approached generally in a very
straightforward manner as a circular area centred on the site focus, with a radius determined
experimentally or ethnographically (Chisholm 1968). Catchment boundaries could also be defined by
travel time instead of radius, but implementing this requires some way of taking into account the nature
of the intervening terrain, and in the absence of GIS has required considerable legwork in the past.
Traditional methods for constructing catchments remain, however, limited by their reliance on plane
geometry and choroplethe cartography. GIS-based techniques, in contrast, allow the simple ‘flat’
geographical space to be supplanted by a complex friction surface incorporating many relevant properties
of the terrain, and the distance-based rule for defining the catchment to be replaced by a time- or energy
expenditure based rule accumulating costs / encountering resistance as it moves further from the focus.
In order to lead up to these more complex techniques, I will first discuss some simple variants.
16 -
6
CASE STUDIES IN VISIBILITY AND FRICTION
DISTANCE-DEPENDENT TECHNIQUES
The calculation of catchments is only a preparatory step to the actual catchment analysis, which should be
based on archaeological theory. For early prehistoric societies this will often be foraging theory; however
this does not apply to the largely settled agricultural and pastoral landscape studied by the WHP. More
relevant is Bewley’s (1994: 65-8) work basing the economic basis of farming settlement sites in the Solway
plain on estimates of soil workability as derived from its moisture capacity and texture. Deriving a circular
catchment area within a GIS is a trivial operation, and reporting and tabulating the presence of resources
within each catchment (typically the goal of this type of analysis) can be automated. Following Bewley’s
classification, we used such a method to derive soil workability data for a 200m radius area around a
sample of 111 enclosures within 5 to 7 km of Wroxeter (figure 16a). Comparison of the workability
characteristics of these catchments to those of the sample area as a whole (figure 6b) shows that soil
workability did not significantly affect the siting of enclosures within the sample area. We could speculate
on the causes - some or many of the enclosures might specialise in cattle raising rather than arable; some
or many might be Roman in data and might therefore have had access to heavy ploughs - but that is not
the aim of this case study. Generally, such speculation points in one of two directions: either soil
workability was not among the most significant factors affecting enclosure siting, or the chosen method
(circular catchments) is too coarse. The latter direction could be pursued by modeling for 'a sufficient
amount of workable soil (e.g., 2 hectares) within a specific maximum distance (e.g., 400 m)' from the
enclosure sites; this type of model was in fact implemented for the Wroxeter hinterland by Gaffney and
Goodchild (Goodchild 1999).
800
8000
600
6000
400
4000
200
2000
0
0
Aa
A
C
D
E
F
Area
(background)
Area
(catchments)
Figure 6a: Soil workability and 200m radius
catchments for 111 enclosed sites in a 14 by 11 km
area around Wroxeter. Workability classes
according to Bewley 1994: dark green easiest –
dark red heaviest.
Figure 6b: comparison of workability
characteristics of enclosure catchments (dark
line) to background (light line). The graph
shows a slight avoidance of very heavy soils and
an equally slight preference for light soils.
Workability classes according to Bewley 1994:
Aa very easy, A easy, C average, D slightly
difficult, E moderately difficult, F extremely
difficult. Area on both vertical axes in hectares.
Workability
16 -
7
V A N L EUSEN : P A T T E R N T O P ROCESS
The binary (inside or outside) result obtained with this type of catchment can also be improved upon by
defining multiple distance-based ‘buffer’zones around sites (or, as will be shown below, around line or
area features). This opens up the possibility of deriving distance-dependant relationships between sites
and resources. We have employed buffer analysis to look at the most plentiful site group in the study area,
that of the enclosures, on the basis of Whimster’s (1989:35) system of morphological classification of
crop mark enclosures in the Welsh Marches. Whimster noted a (slight) tendency of rectangular enclosures
to cluster around Wroxeter; presumably this signifies or expresses some kind of economic, social, or
cognitive link, such as increased demand for agricultural produce or increased status attached to
‘Romanised’forms of settlement, focusing on the civitas capital itself. Whimster therefore tentatively
dates these rectangular enclosures to the Roman period, and the dating has received some support from
excavated evidence elsewhere (Bewley 1994). We can begin probing this hypothesis by checking that the
clustering is in fact present, and a simple one-sample test for randomness confirms that this is the case –
rectilinear enclosures do indeed cluster around Wroxeter more than curvilinear ones (figure 7a).
Figure 7: Expected/Observed number of a) rectilinear enclosures, b) curvilinear enclosures per
distance buffer from Wroxeter.
Figure 8: Expected/Observed number of a) rectilinear enclosures, b) curvilinear enclosures per
distance buffer from major streams.
Having established the fact of the clustering, we must now probe deeper for possible explanations. Since
the clustering is a property of both types of enclosed sites, and curvilinear enclosures are thought to
predate the establishment of the town at Wroxeter, we must look for other additional causes for the
clustering to occur. We may find such causes by examining other environmental and social factors, and by
examining potential bias factors. As a first approach, univariate preferences indicate that nearly all
enclosures occur on relatively flat land, of good to medium workability, and not far from major streams.
When we map the former two factors using the map algebraic function
Good_land=if((SLOPE <= 6) && (WORKABILITY [A - D]))
And limit our subsequent analysis to these area of good land, we can examine the relation of the various
types of enclosures to major streams (see figure 8 above). It now becomes clear that rectilinear
16 -
8
CASE STUDIES IN VISIBILITY AND FRICTION
enclosures were preferentially placed at a certain distance (ca. 500 m) from the nearest major stream, but
could easily occur up to about 3 km from such streams. Curvilinear enclosures have a more direct but
also weaker preference for closeness to streams.
Obviously, if rectilinear enclosures are indeed Roman in age, it may well be that a constellation of
proximity factors (to streams, to roads, to other enclosures, and to the market at Wroxeter) would have
been at work to ‘direct’the choice of a new settlement location. Other rectilinear enclosures may have
been 'Romanised' versions of existing enclosed sites, for which a different constellation of location factors
would be relevant (esp. the cultural, political and social ambitions of the inhabitant).
Since nearly all enclosures have been discovered by aerial photography, our models should also account
for the possibility that systematic biases related to modern crop sensitivity land use account for the
observed distance relationships. As I have shown elsewhere (chapter 14, section 2.3), there is a strong
univariate correlation between crop and soil marks and modern land use.
TESSELLATIONS
If catchment radii or buffer distances are extended until all available space is divided, a territorial division
or tessellation of the landscape results. Archaeological arguments for suspecting the existence of such
tessellations abound, albeit the type of tesselation differs by period. For the advanced Iron Age, the
sociopolitical structure of Wroxeter hinterland is thought to have been dominated by the sites known as
hillforts and perhaps also by multiple-ditched enclosures, and we may therefore be able to model
chiefdom territories using some method akin to Thiessen polygons. For the Roman period, a central place
model of market functions puts a ring of secondary markets at about 15 km from the main market and
civic center at Wroxeter, and introduces the element of ranking in the derivation of territories.
Ruggles and Church (1996) provide a thorough discussion of the theory, problems and possibilities
involved in creating more realistic Thiessen polygons, but in fact the major problem in most studies is
how to decide on a set of contemporaneous, equivalent sites to base the analysis on (see also my critique
of the tesselations applied to the Archaic polities of the Alban hills area, chapter 15). The definition of
potential late Iron Age chiefly residences within the Shropshire SMR is problematic in this respect. Within
the Wroxeter hinterland there are three known multivallate enclosures which are not recorded as ‘hillforts’
(PRN’s 472, 1055, and 2000; see table 1), even though one of them is situated on a hilltop; conversely,
many sites recorded as ‘hillforts’are not actually located on hilltops at all - notably, the ‘lowland hillforts’
at Wall Camp and the Berth at Baschurch. Given that hardly any of these sites has been excavated,
functional interpretations remain largely a matter of conjecture, and a less subjective partition could be
based on the size of their internal area (table 1). We can then distinguish groups of larger (2 to 4 ha) and
smaller (0.15 to 1.2 ha) multivallate enclosures; Wall Camp, at 14 hectares, and Bury Walls (12 hectares
but just outside the study area) are clearly outliers.
Plotting these two groups on a map of the study area (figure 10), we can see a crest of smaller multivallate
enclosures around the fringes of the Long Mynd upland (southwestern corner of the study area), with the
larger ones situated more toward its interior. The smaller sites would combine high status (as evidenced
through the effort spent on earthworks) with accessibility (= nearness to agricultural land and
infrastructure?), and might be residences of chieftains. The set of larger multivallate enclosures are fairly
evenly spaced across the landscape, with closer pairs occurring at Caer Caradoc/The Lawley and
Haughmond Hill/Ebury Hill. An argument can be made for the idea that these two pairs are either not
contemporaneous or not of the same type, and that we should perhaps exclude the lesser of each pair
(i.e., Haughmond and the Lawley) from our territorial analysis.
Constructing Thiessen polygons on the basis of the reduced set of larger multivallate enclosures and a
simple slope-related cost surface, we obtain a division of the landscape with some interesting properties
(see figure 9). Firstly, three pairs of hillforts occur at regular intervals opposite each other along the main
Severn corridor, with Wroxeter itself very near a common boundary between four of these. Although
pairing has been observed elsewhere (Bowden 1989), the siting of opposing pairs across a river is not
16 -
9
V A N L EUSEN : P A T T E R N T O P ROCESS
attested elsewhere, and is likely to be due to the particular physical geography of the study area. Given the
locations of known fords in the Severn, some of the boundaries between these hillfort territories
converge near Wroxeter, a result substantiating theoretical arguments for expecting the location of
markets and religious sites at or near territorial boundaries.
Secondly, hillfort territories to the northeast (Wall Camp) and southwest, with the possible exception of
the Lawley, are not oriented toward the Severn. Obviously these are marginal territories, the size and
shape of which will be influenced by edge effects (in particular, the presence of other large hillforts
outside the study area, such as Bury Walls). But the pure fact that these territories are visually and
physically isolated from the main Severn valley must make effective control that much harder, and it is
unlikely that they played a significant role in controlling the area. Further corrections can be applied by
assuming that territorial divisions followed the natural division of the landscape into watershed basins
(figure 10)1.
2.2
MODELLING IRON AGE/ ROMAN TRADE NETWORKS
Within the WHP we have attempted to take the tools of cost surface analysis further, using drainage or
shortest path calculations to construct a hypothetical Iron Age organic road network that complements
the known network of Roman roads. An extensive and intricate network of routes of various types must
have existed in the Iron Age landscape around Wroxeter. Not only were there established paths between
settlements, but there most also have been paths between settlements and central places such as the
hillforts, markets, and cult sites. Since cattle breeding now appears to have been the mainstay of both the
Iron Age and the Romano-British economy of the Cornovii, we must also expect to find networks of
wider trackways running between upland pasture and lowland settlement and market areas. Such
droveways are thought to have followed ridgelines in order to avoid, as far as possible, difficult terrain
and intrusion on farming land. Where cattle droves run through farming land, the movement of the
animals would have been restricted by parallel hedgerows or fences for which we have ample air
photographic evidence. The Shropshire SMR holds 33 trackway records, 10 of which are doubtful and
none of which have been dated with any degree of certainty. However, some can be assigned to the Iron
Age/Roman period on morphological grounds. A large number of these parallel linear crop and soil
marks are directly associated with enclosures and field systems. This is to be expected because crops
would need to be protected from cattle being taken along the tracks/droveways; outside these farming
areas, the trackways would not need any defined boundaries and could widen out considerably, making
them archaeologically invisible. Although the Roman military road network may to some extent have
followed and fossilised the main existing routes through the area (see especially the reconstruction of
infrastructure by Bassett 1990), we cannot assume that this network is representative of Iron Age
infrastructure in any sense. The object of this case study is therefore to reconstruct the Iron Age
infrastructure on the basis of what we already know about the locations of enclosed settlements. GIS
models based on cost surface analysis can be used to investigate the presence and properties of such
networks, either by directly probing relevant existing archaeological records, or by generating hypothetical
transport networks based on least cost principles.
A stream map and a watershed basin map can be constructed through the application of a drainage algorithm on a DEM; by
inverting the DEM the same algorithm can be used to generate ridge lines. Any resulting basin not fully within the study area will
suffer from edge effects.
1
16 -
10
CASE STUDIES IN VISIBILITY AND FRICTION
Figure 9: Distribution of smaller (diamond) and larger (box) multivallate enclosures in the study
area, with a Thiessen network generated on the basis of the larger enclosures.
Figure 10: Distribution of larger multivallate enclosures in the study area, with major watershed
basin boundaries generated from the DEM (yellow lines).
16 -
11
V A N L EUSEN : P A T T E R N T O P ROCESS
Least cost networks
The calculation of multiple least cost paths through the Wroxeter hinterland requires the definition of
start and end locations, which in turn depend on the system being studied:
• Our first approach was to use Wroxeter itself as a single ‘end’point given strong indications that it
had been the location of an important fair/market during the later Iron Age. ‘Start’points varied
depending on which type of route was being modeled; for the regular traffic of people and goods
start points were taken from the locations of known settlement sites (enclosures).
• A second set of route networks was generated using the larger multivallate enclosures as end points
and all enclosures as start points, in order to simulate the infrastructure of the Late Pre-Roman Iron
Age settlement system.
• Finally, a third set of networks was generated with end points chosen to represent the developed
Roman trade system. This includes, in addition to Wroxeter itself, known secondary roadside
settlements at Westbury, Church Stretton, Bridgnorth, Red Hill (Uxacona) and the lost settlement near
Harcourt (Rutunium).
Below we describe the process involved in generating the first of these networks, with a single end point
at Wroxeter. This is followed by a discussion of the other networks.
NETWORK GENERATION
The following calculation of the Wroxeter cumulative cost surface (‘cost to market’) and of the
subsequent least cost paths (‘cost from start point’) is based on a cost surface obtained using the Pandolf
formula for the physiological expenditure M (metabolic rate in Watts) involved in moving over natural
terrain, which incorporates total weight (body plus load) moved, a terrain factor describing ease of
movement, and percent slope:
M = 1.5W + 2.0(W + L)(L / W)2 + N(W + L)(1.5V2 + 0.35VG)
The grade G was mapped as percent slope, a derivative of the 50 metre resolution DEM obtained from
the OSGB for the study area. Because the slope calculation is non-directional, no distinction has been
made between downslopes and upslopes; my discussion in chapter 6 shows that this has a minor effect on
the quality of the model.
The terrain factor N, a cost surface (figure 11a), is constructed on the basis of terrain features known to
influence movement - marshy areas, roads, and streams of various widths, with a default value equivalent
to the presence of a dirt road assigned to the remainder of the study area. This default value is important
in that it assumes the presence of an existing intricate network of paths throughout the area, so that the
need to create an entirely new path (which would entail higher energy expenditure) would hardly ever
arise (an assumption I have argued for in chapter 6). Coefficients for these terrain features were mostly
taken from Marble (1996:5, quoting Machinova 1996 and Givoni and Goldman 1971). We assumed that
for small loads the Roman roads would not have a much lower coefficient than ‘organic’ones. Givoni
and Goldman (1971), for example, assign terrain coefficients of 1.0 and 1.1 respectively for metalled
roads and unmetalled paths. Although this slight difference in cost resulted in many least cost paths
running parallel to each other because of the effective lack of an energy penalty over large stretches of the
study area (see, for example, figure 15), it is preferable not to attempt to ‘correct’for this by artificially
lowering the coefficient of the Roman roads or raising that of the default value.
Modelling the terrain factor N revealed several implementation problems. The main problem in the
calculation of N proved to be the presence of a major river with tributaries in the study area; it was also
noted that in order to properly model seasonal variations in the accessibility of some land types (river
16 -
12
CASE STUDIES IN VISIBILITY AND FRICTION
alluvium and peat bogs) it would be necessary to calculate different versions of N. Stream coefficients had
to be set relatively high (88 for the Severn, 22 for the larger streams, and 5.5 for the smaller ones) in order
to prevent the least cost paths from crossing them repeatedly. We were also forced to include information
about fordable places in the Severn in order to differentiate between these and other stretches of the
river; fortunately, a cartographic record of Severn fords was available (Pannett 1989). This map identifies
16 -
13
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 11: a) Terrain Factor N; b) Energy cost surface M in Watts for the Wroxeter
Hinterland.Darker tones indicate higher cost.
nine good fords between Apley (below the Ironbridge Gorge) and Hayes on the Welsh border, in
addition to four lesser fords between Mytton and Cressage (in the centre of the study area), all of which
were assigned a value of 1. It was found that the incorporation of these detailed terrain features in our
calculations caused the resultant least cost routes to converge more on each other, on occasion skirt
around bends in streams, and make more use of the Roman road network. Finally, an unexpected
technical problem had to be circumvented. It turned out that, in the raster based GIS used, the cost
accumulation algorithm and the drainage algorithm both perform a local search of 8 neighbouring cells to
locate the lowest cost neighbour, whereas linear terrain features with a width near that of the map
16 -
14
CASE STUDIES IN VISIBILITY AND FRICTION
resolution (25 metres) were represented by a single corner-connected line of cells (see figure 13). This resulted
in the ‘skipping’of the important high cost terrain features by both the cost and the drainage algorithms,
and forced a rerun of the whole analysis - this time with all streams, roads and fords widened by 2 cells so
no more ‘skipping’of high cost linear features was possible.
The other variables involved in the determination of M were kept constant, with weight W at 70 kg, load
L at 4 kg, and velocity V at 4.8 km/h. Using these parameters the formula to calculate metabolic rate in
watts (M) becomes:
M = 105 + 0.483 + 74N(34.56 + 1.68G) = 105.483 + 2557.44N + 124.32NG
The resulting cost surface M, illustrated in Figure 11b, forms the basis for a series of cumulative cost
surfaces centering on the intended 'end point(s)'. In order to model a single route network converging on
Wroxeter, a cumulative cost surface was calculated using M as cost and NGR 356485 / 308705 (the main
forum entrance at Wroxeter) as the end point coordinate (figure 12). Cycling through a list of enclosure
site locations used as 'start points', this surface is then drained and the resulting least-cost paths added
together using map algebra to yield an 'organic' or 'natural' network for travelling from these sites to
Wroxeter. This network indicates not just where, but also how intensively used, routes were. One
example of a least cost path is illustrated in figure 14; the full network is depicted in figure 15.
Figure 12: Cumulative energy expenditure surface for Wroxeter, using shaded background.
16 -
15
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 13: The representation of linear terrain features in a raster cost surface can
introduce significant errors in the resulting cumulative cost surfaces and least cost paths,
depending on the algorithms used for cost accumulation and drainage. In this example,
on the left, a least-cost path can ‘jump’a one-cell-wide cost barrier such as a river if the
algorithm allows diagonal moves; on the right, the problem has been fixed by widening
the cost barrier - forcing the algorithm to look for low cost crossing points such as fords.
Figure 14: Simulated route network based on least cost paths through the cost surface of figure 12,
streams and Roman roads included.
The resultant network has some gratifyingly realistic aspects. In particular, it shows avoidance of streams
and convergence of routes, and the organic routes generated in the upper Severn valley coincide
approximately with the lines of presumed roads into north Wales (north-western part of the study area).
ROMAN SECONDARY MARKETS NETWORK
We can now investigate an alternative model of travel between enclosed settlements and markets, by
adding to the single ‘end point’(market) at Wroxeter five more end points located at known or presumed
second-level settlements in the study area – at Harcourt, Westbury, Church Stretton, Bridgenorth, and
Red Hill. In this model we will assume equality of ‘attraction’(that is, of cumulative travel costs)
16 -
16
CASE STUDIES IN VISIBILITY AND FRICTION
between Wroxeter and these other sites. The implementation of this model again starts from the cost
surface N and the energy surface M (figure 11); the cumulative cost surface (figure 15) uses the six sites
mentioned above as start points; and the cumulative least cost network (figure 16) again uses all
enclosures as start points.
Figure 15: cumulative cost surface for primary and secondary markets in the Wroxeter Hinterland.
Maximum cost boundaries (black), streams (blue) and roads (red) overlaid.
Figure 16: least cost network from all enclosure sites to primary and secondary markets.
16 -
17
V A N L EUSEN : P A T T E R N T O P ROCESS
Again the influence of terrain costs on territories and paths is clear. The ‘territorial boundaries’(maximum
cost lines) in the cumulative cost surface tend to follow streams, and the fordable places in the Severn
exert a strong ‘attraction’ on paths (see, for instance, the convergence of paths near the ford at
Strawardine - top left of study area). Also notable is the divergence between the simulated route and the
hypothetical Roman road where it crosses the Tern directly north-west of Wroxeter. Whereas the line of
the Roman road to the north of the Severn (postulated on the basis of regularities in 19th century field
boundaries) requires the presence of an additional bridge across the Tern, the simulated least cost path
takes a shorter route and uses the existing Tern bridge some 2.5 km to the south.
This model may be refined still further by adding constraints in the form of known nodes and paths. It
has, for example, been suggested that the site of Meole Brace a few km to the west of Wroxeter is
important because it is located at a node in the Roman road network, with excavated evidence for
redistribution functions (Ellis et al. 1992). Likewise, the known trackways of the Wroxeter hinterland can
be used as ‘attractors’in the cumulative cost surface. Even if the hinterland infrastructure in both the Iron
Age and Roman period was largely organic (and there is no reason to believe otherwise), routes may be
expected to follow natural lines from settlement to settlement, and to interconnect with the Roman road
system. We may therefore expect nodes to occur where natural routes connect with formal roads and
other routes routes, and may derive some idea about the relative importance of these nodes by the
number of individual least cost that intersect there. Such simulations will not be very accurate but should
still indicate areas where we might look for small markets or shrines, and will allow us to re-study the
archaeological records from this perspective.
3
E D G E E F F E C T S A N D B ACKGROUND INDICES
Cumulative viewshed analysis is a tool often used to investigate and interpret the ‘social’placement of
archaeological sites and monuments in the landscape. The placement of these sites and monuments in
areas of relatively high or (less often) low visibility is often seen as proof that they were intentionally put
there. A more sophisticated approach first calculates a ‘background’CVI which describes the ‘natural’
visibility of all parts of the terrain, then investigates whether the viewshed properties of the sites of
archaeological interest are significantly different from this, therefore presumably intentional. However,
the quantitative study of viewshed intensity gives rise to several further more or less subtle distorting
effects which must be taken into account. Two of these – the edge effect and the influence of viewshed
radius on the relation between elevation and CVI - are demonstrated here.
3.1
EDGE EFFECTS
The edge effect has been discussed in general terms in chapter 6, and in the case study presented in
section 1 above its maximum reach was visualised as a box outlined in red (figures 1 and 2), but no
attempt was made to further quantify it. In the following such an attempt is made, employing an idealised
circular raster ‘world’with a radius of 20 km and a resolution (cell size) of 500 by 500 m. The total
number of cells within this area is 4977. Three cumulative viewshed indices (CVI’s) were generated using
50 (1%), 250 (5%), and 500 (10%) samples of randomly chosen seed cells and a 10 km viewshed radius
(see figure 17a). It may be observed that the relatively large viewshed radius yields a consistent area of
high visibility near the centre of the world even with a very small sample of ‘seeds’. As the number of
seeds rises, so the CVI approximates the ideal distribution which would have resulted from a 100% seed
sample; but the 10% seed sample used for the right-hand CVI is already clearly sufficient for investigating
the edge effect.
16 -
18
CASE STUDIES IN VISIBILITY AND FRICTION
Figure 17a: Cumulative Viewshed Indices for a flat test universe of radius 40, using viewsheds of
radius 20 and random samples of 50, 250, and 500 viewpoints. Min=yellow, max=red. Grid
spacing: 20 cells.
In order to obtain a graph of the relation between CVI values and distance from the edge, a normalised
index of edge distance (EDI) was constructed in which 0 represents the edge and 100 represents the
centre of the world. Next, the median CVI per EDI value was tabulated and is represented in the graph
below.
CVI
Figure 17b: graph of edge distance index
(EDI, horizontal) against cumulative
viewshed index (CVI, vertical) for the righthand universe in figure 18a. Above a
distance of 20 (one viewshed radius), no edge
effect occurs.
100
80
60
40
20
0
1
4
7
10 13 16 19 22 25 28 31 34 37
The three CVI images come progressively nearer the ‘ideal’, where all areas over one viewshed radius (20
cells) from an edge have an identical CVI. It is noteworthy that even a 5% random seed sample can yield
a CVI that still deviates significantly from this ideal – so we cannot accept this as a rule of thumb (contra
Lake et al. 1998). Within the circular and ‘flat’universe used for this test, the edge effect diminishes
linearly with distance, as predicted. With a 10% sample of ‘seed’cells, CVI rises steadily from 12% at the
edge to 77% at one viewshed radius from the edge, then holds steady at about 80% before dipping to
72% at the center of the universe due to the random nature of the seed locations. The absolute CVI at the
edge of the universe and on the ‘plateau’depends on the precise combination of viewshed radius and
average distance between seed points used, and a formula could be derived to predict both, but in a
viewshed study of a real terrain both parameters would obviously be significantly and unpredictably
lowered.
3.2
VIEWSHED RADIUS EFFECT
The further away from the viewer, the more likely it is that an object will be masked by intervening
terrain, hence the more elevated it has to be in order to be seen. With increasing viewshed radius the
16 -
19
V A N L EUSEN : P A T T E R N T O P ROCESS
16 -
20
CASE STUDIES IN VISIBILITY AND FRICTION
amount of terrain far from the viewer increases exponentially whereas the amount of terrain near the
viewer remains constant; therefore elevated locations will obtain a higher CVI at larger viewshed radii. In
order to investigate and demonstrate the effect of changing viewshed radius on the types of
geomorphological units preferentially ‘seen’, digital elevation data from the WHP were used to calculate
several random ‘background’CVI’s at varying radii.
This showed up some important effects straight away. The cumulative short-range (2 km) viewshed
generated from a set of 5,000 points (approximately 4 points per km2) contains visibility values ranging
from 0 (completely secluded) to 54 (highly visible), but these are not randomly distributed. In fact, they
correlate directly with the size of the convex and concave geomorphic units in the study area, especially
when these are larger than an individual viewshed (say, over 4 km2). Conversely, a cumulative long-range
(10km) viewshed generated from 500 points (approximately 1 point per 6 km2; random3.los) shows the
high visibility index, not surprisingly, correlating with high ridges and hillsides, an effect which would be
expressed even more strongly if, as is current practice in many GIS studies, the viewsheds were
unconstrained.
4
CONCLUSIONS
As a consultation of chapter 6 will make clear, the case studies presented in sections 1 and 2 no longer
represent the ‘state of the art’in GIS modeling of visibility and accessibility. The modeling of ‘energy and
resource landscapes’ in a GIS environment with the help of cost surface techniques has become
increasingly sophisticated in the last few years. Cost is now measured in real terms as energy expenditure
in Watt or kCal; cost surfaces have become anisotropic to reflect the importance of the effect of direction
of movement on costs; the interpretation of visibility and accessibility models is now informed by a better
understanding of statistical complexities; and, last but not least, the limitations of the underlying theory
are now beginning to be understood.
Further work will be needed in order to develop sufficient understanding of ‘background’visibility and
accessibility indices such as the ones developed above in section 3, and by Llobera (2000). Of equal
importance is the testing of GIS-generated models, firstly by a comparison with extant historic,
archaeological, and cartographic data, and thereafter by targeted fieldwork.
Although the simulations presented in section 3 do not constitute conclusive proof, it would appear that
simply by increasing the viewshed range used, the higher visibility values will concentrate on areas of
higher ground, ridges and peaks. Any sufficiently large sample of archaeological viewpoints will tend to
generate a cumulative viewshed similar to these simulated ones, depending on the viewshed radius used.
Furthermore, any such viewshed based on points located on or near ridges and peaks will further
emphasise the visibility of other ridges and peaks. These results, together with those obtained in similar
simulation studies that found little correlation of viewshed intensity with elevation (Franklin & Ray 1994),
need further evaluation. Cumulative viewshed analyses must take such effects into account, or they
become nothing more than vehicles for our prior convictions.
REFERENCES
Bassett, SR 1990
The Roman and Medieval Landscape of Wroxeter, pp 10-12 in: Barker, PA (ed), From Roman Viroconium
to Medieval Wroxeter. Worcester: West Mercian Archaeological Consultants.
Bewley, R 1994
Prehistoric and Romano-British Settlement in the Solway Plain, Cumbria. Oxbow Monograph 36.
Bowden, M 1989
Little boxes: more about hillforts, Scottish archaeological review 6:12-16.
Chisholm, M 1968
Rural settlement and land use: an essay in location. 2nd (revised) edn, London: Hutchinson.
Franklin, WR and CK Ray 1994
16 -
21
V A N L EUSEN : P A T T E R N T O P ROCESS
Higher isn't necessarily better: visibility algorithms and experiments, in Waugh, TC and RG Healey (eds),
Advances in GIS research: proceedings of the 6th international symposium on spatial data handling: 75170. London: Taylor & Francis.
Goodchild, H 1999
An Appraisal of the Urban Dependency of Roman Wroxeter City. Unpublished MA dissertation, University
of Birmingham.
Houghton, AWJ & E Wells 1978
The Auxiliary Fort at Wroxeter, West Midlands Archaeol News Sheet 21: 60-4.
Llobera, M 2000
Understanding movement: a pilot model towards the sociology of movement, in Lock, G (ed) 2000, Beyond
the Map: Archaeology and spatial technologies: 65-84. Amsterdam, etc: IOS Press.
Madry, SLH & L Rakos 1996
Line-of-sight and cost-surface techniques for regional research in the Arroux River valley, in HD Maschner
(ed), New Methods, Old Problems. Geographic Information Systems in Modern Archaeological Research
(Southern Illinois University Center for Archaeological Investigations Occasional Paper 23): 104-126.
Marble, DF 1996
The Human Effort Involved in Movement over Natural Terrain: A Working Bibliography, Unpublished
report, Department of Geography, Ohio State University.
Pannett, D 1989
The River Severn at Wroxeter, Transactions of the Shropshire Archaeological Society 66: 48-55.
Ruggles, AJ & RL Church 1996
Spatial Allocation in Archaeology: an opportunity for reevaluation, in HD Maschner (ed), New Methods,
Old Problems. Geographic Information Systems in Modern Archaeological Research (Southern Illinois
University Center for Archaeological Investigations Occasional Paper 23): 147-173.
St Joseph, JK 1951
Roman Forts on Watling Street near Penkridge and Wroxeter.
Van Leusen, PM 1999
Viewshed and Cost Surface Analysis Using GIS (Cartographic Modelling in a Cell-Based GIS II), in JA
Barcelo et al. (eds), Computer Applications and Quantitative Methods in Archaeology 1998.
Whimster, R 1989
The emerging past: air photography and the buried landscape. London: RCHME.
White, RH & PM van Leusen 1997
Aspects of Romanisation in the Wroxeter Hinterland, in Meadows, K, C Lemke & J Heron (eds), TRAC 96.
Proceedings of the Sixth Annual Theoretical Roman Archaeology Conference, Sheffield 1996: 133-143.
Oxford: Oxbow Books.
White, RH & Ph Barker 1998
Wroxeter. Life and Death of a Roman City. Stroud: Tempus.
16 -
22
C HAPTER 17
INTERPRETING FIELD SURVEY
RESULTS IN THE LIGHT OF
HISTORIC RELIEF CHANGE:
THE FOGLIANO BEACH RIDGES
( S O U T H L A Z I O, I T A L Y ) ∗
Hendrik Feiken & Martijn van Leusen
1
BACKGROUND
The Pontine region, a low-lying and partly marshy area along the coast of south Latium, was taken up into
the Roman power sphere in a slow process completed only midway through the 4th century BC, which is
why the preceding period 500 - 350 BC is named ‘post-Archaic’rather than ‘early Republican’(Attema
1993). Much of it appears to have been marginal to the major political and economic developments of the
time, and this translates itself into the relatively low density of surface ceramics reported in field surveys
(Attema et al. 2001, Van Leusen 1998). Low finds densities present us with particular interpretation
problems because of the relatively large influence of biases in the type and amount of research conducted,
the visibility of the surface, statistical effects, and geopedological circumstances (Van Leusen 2001). It is
the effect of the latter factor that we have attempted to study in more detail, using the landscape of
ancient beach ridges around the Fogliano lagoon as our study area (figure 1).
This feasibility study into the use of historic elevation data for the mapping and correction of biases in the
archaeological record was conducted as part of the Regional Pathways to Complexity (RPC) project1 in
the Pontine region (south Lazio, Italy). The general aims of the RPC project are, first, to understand
indigenous versus externally-induced growth in complexity (especially urbanisation), and second, to
conduct detailed surveys in marginal areas to understand the scope and nature of dynamics that are
mostly known from urbanised sites. Specifically, the case study presented here is part of the project’s
methodological focus on GIS approaches to the detection of spatial patterning in the archaeological
record. One reason for concentrating on the Fogliano area is that it has been subject to two major surveys
in the last two decades (Voorrips et al. 1991; Attema et al. 2001), and the surface record is therefore
relatively well known; the other reason is that the Pontine plain as a whole was subject to major
∗
This chapter is a slightly revised and updated version of a study first presented as a paper at the 28th Annual Meeting of CAA in
Ljubljana (Slovenia, April 2000) and subsequently published in Stancic, Z & T Veljanovski (eds), Computing Archaeology for
Understanding the Past. Proceedings of the CAA2000 conference (BAR S931): 205-211. Oxford: Archaeopress.
The RPC project was conducted jointly from 1997 to 2001 by the Groningen Institute of Archaeology and the Archaeological
Institute of the Free University of Amsterdam. It studies protohistoric landscape and settlement dynamics in three Italian regions
- the Pontine region, the Salento Isthmus, and the Sibaritide. Processes of centralisation, early urbanisation and colonisation are
its main themes (Attema et al. 1998).
1
17 -
1
H I S T O R I C R ELIEF C H A N G E I N T H E P O N T I N E P LAIN
restructuring in the late 1920s and 1930s by the Italian government in a so-called Bonificá. These works
have changed many parts of the landscape, and must therefore be taken into account if we are to interpret
and understand our survey data. A detailed elevation survey of the Pontine plain was made in 1927 by the
Italian Military Topographic Institute (IGM), allowing us to compare the post-World War II relief to the
relief that was present before the Bonificá.
Figure 1 - Location of the Fogliano study area within the Pontine region (Lazio, Italy).
Before we proceed to describe the feasibility study itself, a brief overview of the settlement and land use
history of the area is in order. The landscape of ancient beach ridges around the Lago di Fogliano must
have attracted humans from earliest prehistory, and flint artefacts dating from the Middle Paleolithic
onwards were found during the surveys. This material seems to concentrate mostly along the banks of the
larger water courses and the lake itself, and one can imagine the water rich environment being very well
suited for fishing and fowling. However, throughout this period and into the late 2nd millennium BC
human presence would appear to have been quite rare and impermanent, the earliest indications for
agricultural activity and the use of ceramics dating to the Bronze Age2. Figure 2 shows the probable
settlement locations for the protohistoric period (running from the Bronze Age up to the beginnings of
Roman influence in the Pontine region, around 500 BC) as derived from the finds densities of the RPC
survey, on the background of the 1928 DEM. The absolute number of finds from this period tends to be
very low for most of the area surveyed - on the order of 1 to 5 finds per hectare. Settlement in this period
seems to be concentrated on relatively well-drained capes and banks along the larger streams, where
access to natural resources would have been easiest and preconditions for paleotechnic agriculture were
positive (Kamermans 1993: 100-4; see also Attema et al. 2001). By the end of this period (6th century BC),
the number of settlements begins to grow.
Figure 3 shows the post-Archaic, Roman Republican, and early Imperial sites identified by the RPC
survey, again covering a period of approximately one thousand years (500 BC to AD 500). The number of
settlements has greatly increased, but this occurs mostly between 200 BC and AD 200 - the late
Republican and early Imperial period. Settlement is concentrated on the relatively level and agriculturally
The evidence for this comes from an unpublished pollen core from Fogliano (pers. comm. E. van Joolen), and from the
unpublished finds database of the Agro Pontino Project (Voorrips et al. 1991).
2
17 -
2
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 2 – Results of the 1998/9 RPC survey in the Fogliano area, protohistoric period. Find
densities per hectare corrected for surface visibility. Background: 1927 shaded relief map and km
grid.
Figure 3 – Results of the 1998/9 RPC survey in the Fogliano area, Roman period. Find densities
per hectare corrected for surface visibility. Background: 1927 shaded relief map and km grid.
17 -
3
H I S T O R I C R ELIEF C H A N G E I N T H E P O N T I N E P LAIN
usable area to the east of the Lago di Fogliano, where a late Republican village seems to have sprung up.
We can tentatively relate this development to the inclusion of the area in the wider economy of Roman
society. Roman agricultural technology (drainage of lower-lying areas, heavy ploughs) enables more land
to be farmed; Roman hydrocultural technology (regulation of lake levels) enables commercial fisheries to
be established in the lagoons; Roman roads and markets enable the commercial exploitation of the clay
beds along the nearby Astura river; from the 1st century BC rich Romans even built their summer palaces
along the banks of the coastal lagoons of south Lazio. All this activity makes it likely that the growing
number of small farms became dependent on a few large specialised rural villae. However, this system
collapses from the 2nd century AD onwards, as a wetter climate3 led to the return of marshy conditions
and the expanding Roman empire found its supplies elsewhere.
2
T R A C K I N G H I S T O R I C RELIEF CHANGE
For the purposes of studying historic relief change in the Fogliano area, we have used historic maps of
land use and land form. We know from historical research that the landscape was little changed since the
Middle Ages, and historic maps dating to the 17th century show us an approximation of the landscape as it
was in the late Roman period. Such maps may even be a better guide to interpreting the proto- and early
historic archaeological record than the modern ones, which were made after major restructuring of the
region during the Bonificá.
When we digitised and compared a very detailed digital elevation model (DEM) from the 1:5,000 scale
maps prepared during the Bonificá (figure 4; IGM 1927) with a commercially available DEM of 1 arc
second (~25 m) resolution (figure 5)4, some major differences became immediately apparent. Whilst
obtaining a map of the differences between the two mappings was easy - all we needed to do was to
subtract the two DEMs from each other – most of our study was concerned with the identification and
removal of various mapping errors obscuring the ‘real’relief changes that might have taken place during
and following the Bonificá. Since it is usually one of the most important layers in a regional archaeological
GIS, deficiencies in the DEM can cripple much analysis. These sources of error, and our attempts at
removing them, are discussed in section 3; our interpretation of the archaeological evidence in the light of
the ‘cleaned’elevation data follows in section 4.
3
EXTRANEOUS SOURCES OF DIFFERENCES BETW EEN THE TWO DEMS
The first point we need to make is that any DEM is a model, that is, a simplified version of reality. The
type and amount of simplification that can be supported in any analysis depends on the questions asked,
and any analysis relying on DEMs should be explicit about its limitations. Secondly, the comparison of two
DEMs forces us to be even more precise in our description of the data. Differences between two DEMs
may be due to real changes in the morphology of the terrain they represent, or to errors committed in the
process of producing the digital elevation data, or to the precision with which the data were recorded, or
to the use of different projection parameters and coordinate systems. The following list summarises seven
distinct sources of differences we discovered between the two DEMs, which are not due to an actual (real)
change in the land form.
3
As deduced from pollen cores analysed by Haagsma (cited by Attema 1993:253) and Veenman (1996:59).
A matrix of numeric data, containing one value for each square arc-second, and based on the digitised contour lines and
elevation points of the 25 and 25V map series produced in the 1940s and 1950s at a scale of 1:25,000 (IGM 1996:13 and Table
42).
4
17 -
4
V A N L EUSEN : P A T T E R N T O P ROCESS
Scale: The scales of the topographic map sheets, from which our two DEMs are derived, differ. The 1927 DEM
derives from a 1:5,000 scale map; the more recent DEM derives from a 1:25,000 scale map. The reduced
mapping scale implies that features are simplified and smaller features may even be lost. When comparing DEMs
of different scales, such features will stand out as differences.
Resolution: Both the horizontal (X and Y) and the vertical (Z) resolution of our two DEMs differ. The horizontal
resolution of the 1927 DEM is approximately 5 by 5 m; that of the more recent DEM is 1 arc second
(approximately 25 by 31 m at the latitude of Italy). This means that an area represented in the latter by a single
elevation value, is represented by approximately 30 (5 by 6) values in the former. Unless that area happens to be
level, most of those values will be different from the single elevation value provided by the low resolution DEM –
they will lower downslope, and higher upslope (see figure 6a), giving rise to the ‘banded’appearance in some
areas of the raw differences map (figure 7a). The amplitude of the differences is related to the terrain slope and
the difference in resolution of the DEMs: at any given percent of slope, the error is directly proportional to that
difference. Because the error is systematic, we can devise a formula for deducing it: E = dR * S / 2, where E =
Error; dR = resolution difference; S = percent slope. The vertical resolution of the 1927 DEM is 0.1 m; that of the
more recent DEM is 1 m. Since the latter must round any values to the nearest whole meter this may lead to
differences of up to 0.5 meters compared to the former.
Mapping errors: Mapping errors may (and will) occur both during the original recording of elevation
measurements, and during the subsequent cartographic process. Proper control procedures are needed to
minimise the occurrence and effects of such errors. While we could not check the quality of the primary
cartographic data, we were able to compare both DEMs with the elevation data in the original map sheets, and
found some major mapping errors. One of these can be seen in figures 4 and 5 (area marked ‘A’) where a small
valley that existed in the 1928 DEM had mysteriously changed into a hill by the 1950s. Whilst such obvious errors
can be found and corrected fairly easily, there certainly remain many less obvious mapping errors in our DEMs a very worrying situation...
Interpolation: Our two DEMs were created by interpolation from digitised contour lines and elevation points, and
this has led to the introduction of three different kinds of errors in the data. Si nce many properties of the resulting
DEM are determined by the interpolation method (e.g., inverse distance weighting or thin plate splining; see
Hageman & Bennett 2000), it is important to study its effects:
While the interpolation method used in creating the 1927 DEM is known (a ‘flood fill’algorithm provided
by GRASS GIS), no such information was available for the more recent DEM. However, it appears from the data
that some sort of inverse distance weighting using a low number of data points was used, a nd this has led to a
large number of visible artefacts in the latter. Since the two DEMs were created using different interpolation
algorithms, the interpolated values will also differ.
Some softwares are unable to handle ‘0’(zero) as a real elevation value, in which case such values are
ignored during interpolation. The method used to create the more recent DEM apparently suffers from this error
which, for example, has caused the present dunes to ‘smear out’across the lagoons because the water’s edge
was digitised as a zero contour (see figures 4 and 5, area marked ‘B’). The elevations here are clearly in error,
and the area cannot be used in any further analysis;
Many interpolation algorithms that start from an input of digitised contour lines (including the ones used
to create both our DEMs) do not resample the input values. Since contour lines always represent cardinal
elevations, these end up being overrepresented in the resulting DEMs - the map histogram has a ‘saw tooth’
appearance, with the distance between ‘teeth’depending on the vertical resolution of the original map. While the
1927 DEM used contour lines every 0.5 m, the more recent DEM used contour lines every 5 m. In terms of
accuracy, this means that values are only accurate to half the dista nce between contours, i.e. to 2.5 m in the
case of the more recent DEM (see figure 6b).
Datum shift: One of the parameters of the coordinate system used by both DEMs is its datum (origin). The
horizontal datum of Italian topographic maps was moved by 2.53” (about 70 m) on one occasion; we could not
obtain any information regarding changes in the vertical datum used in either DEM 5. Left uncorrected,
comparison of the two DEMs using different datums would have led to measurement errors especially in areas of
steep slope; in this case, we were able to correct the horizontal datum shift.
We attempted to compensate for this by comparing the recorded elevations of relatively stable landscape features (e.g.,
buildings) in both DEMs. No vertical datum shift could be deduced from these.
5
17 -
5
H I S T O R I C R ELIEF C H A N G E I N T H E P O N T I N E P LAIN
Figure 4: Digital Elevation Model of the Fogliano area, derived by interpolation from digitised
0.1m contour lines and elevation points of the 1927 1:5000 IGM maps.
Figure 5: Digital Elevation Models of the Fogliano area, from 1arc-second numeric cartography
derived from the 1:25000 IGM topographic map series (permission IGM 26/05/98, no. 4805).
A: digitising error; B: interpolation error.
17 -
6
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 6a - On a slope represented by the thin continuous
line, the differences between two DEMs of differing
horizontal resolution (here, 5 and 25 m) alternate
between positive and negative (shaded areas). The
maximum difference depends on the terrain slope and the
difference between the resolutions of the two DEMs. All
units in m.
Figure 6b - The histogram of a typical DEM shows the
systematic ‘saw tooth’shape caused by overrepresentation
of cardinal values. Horizontal axis: elevation in m;
vertical axis: number of cells x 1000.
Our aim in identifying all these errors is, of course, to be able to correct them. Mapping errors (at least
those that were discovered by us) were corrected by re-digitising and interpolating from contour lines on
the original map sheets, and replacing the incorrect sections with these new data. Other areas of faulty
data could only be ‘corrected’by removing them entirely from the analysis (using ‘masking’functions in
the GIS). But the most interesting type of error was the one that could be estimated or calculated,
because these errors could be corrected by first mapping them and then using them as ‘noise filters’when
interpreting the differences between the two DEMs. Thus, the error formula referred to under
‘Resolution’in the list given above was used to calculate a map layer of the estimated amplitude and sign
of the error, given the known slope and resolution of the more recent DEM. We then applied the
correction by subtracting this calculated error from the observed difference between the two DEMs.
Figure 7 shows the differences between the 1928 and 1950s DEMs, both (a) before and (b) after the
corrections were applied. The general effect is one of producing a much less extravagant map, whose
values do not exceed 5 meters. We were now ready to evaluate these differences in terms of natural and
human processes that occurred in the area, rather than in terms of data error.
17 -
7
H I S T O R I C R ELIEF C H A N G E I N T H E P O N T I N E P LAIN
Figure 7 – Interpreting the differences between the two DEMs. A (top): uncorrected; B (bottom):
corrected..
17 -
8
V A N L EUSEN : P A T T E R N T O P ROCESS
Figure 7 C – Simplified corrected differences between the two DEMs, with the RPC (continuous
lines) and APP fieldwork areas (dashed lines) overlaid. A: area where large-scale sand removal may
have taken place; B: area where farmer has levelled his field.
30
20
10
0
0
500
1000
1500
-10
-20
-30
-40
-50
-60
Figure 8 – Scatterplot of average deflation/inflation (in dm) versus average find density per surveyed
unit of the Fogliano survey (Attema et al. 2001).
17 -
9
H I S T O R I C R ELIEF C H A N G E I N T H E P O N T I N E P LAIN
4
INTERPRETING THE E VIDENCE
Any real differences between the two DEMs might be the result of either natural or human causes, or to a
combination of both. In the simplified corrected map of differences between the two DEMs (figure 7c),
light zones have become up to 7 meters lower, dark zones up to 6 meters higher since 1927, and these
zones may therefore betray works carried out during the Bonificá. However, we must also consider other
processes such as plough-induced erosion and settling of zones with a clayey subsoil.
With the help of a soil scientist familiar with the region we began by evaluating the potential size of the
effects of natural causes. We could see that the lower-lying areas (especially the small valleys) had all
gotten considerably higher since the late 1920’s, whereas some areas to the north-east had been reduced
by 3 meters or more. The former is probably due both to intentional dumping of material into marshy
places and to the effects of ploughing the slopes of the adjacent ancient beach ridges. The latter may well
be partly caused by soil settling after the construction of the canal system during the Bonificá6, but it is
unlikely that a difference of more than one meter could be explained that way. We have some historic
evidence that a top layer of aeolian sands may have been removed from some terrains in order to provide
material for the founding of the new capital of the area, at Latina, and sand and gravel pits were still in
use in the northwestern part of the study area in the early 1980s and recently (Sevink et al. 1984:28);
archive research may reveal further evidence as to which terrains were historically involved in such
activities.
Certainly many parts of the study area were levelled in order to facilitate access and workability for
modern farming equipment, but such works were not often recorded7, so we have to rely on the
recollections of the local farmers and on the occasional evidence from soil cores. Raised terrains also
occur along the banks of the coastal lagoons, and we know from historical sources that these were
deepened and the material used to raise the surrounding land enough to prevent the formation of
marshes8.
Figure 7c shows that some of the fields surveyed in 1998/1999 by the RPC project (and in the 1980s by
the Agro Pontino Project) lie in zones affected by serious deflation or inflation. For the fields surveyed by
the RPC project, the relation between the average amount of change in the elevation and the average
density of ceramics found (see figure 8) gives rise to the suspicion that limited deflation (of ca. 0.5 m) is
helpful in bringing archaeological material to the surface, while strong deflation (more than 0.8 m)
destroys the record altogether. Conversely, even weak inflation tends to hide any archaeological material
that might be present on the surface. A closer look at the observations made in the field confirms the
general picture, and suggests ways in which our archaeological intrepretations may be improved:
• Areas with strong deflation (over 1.5 m) were observed during the survey to have no soil profile at all
and to be archaeologically sterile (area marked ‘A’); Sevink et al., in their account of the soils of the
area (1984:30), suggest that subsoil was brought from elsewhere, but this is flatly contradicted by the
evidence for deflation from our study. It is therefore quite possible that the top layer of soil, including
any archaeological remains, was removed from this area and used elsewhere in construction work9.
Evidence of such soil settling is plentiful in the Pontine region. The approaches to many of the small bridges built in the late
1920s have had to be lengthened, and the concrete bottom of the channel below them broken through, because the surrounding
land had sunk 70 cms or more.
6
7
With the exception of the local, often marshy, hollows which were mapped on the 1920s land cover maps.
In fact, this was one of the major aims of the Bonificá – to reclaim the Pontine marshes for agriculture by destroying the habitat
of the malaria mosquito.
8
It is known from historical records that large amounts of soil were brought in to found the new capital of the Agro Pontino at
Latina.
9
17 -
10
V A N L EUSEN : P A T T E R N T O P ROCESS
The late Republican village mentioned in the introduction may therefore have extended further to the
northwest than the results of our survey suggested.
• The reduction of local relief through ploughing or levelling can be traced in many fields:
1. Areas that were observed during the survey to have an unusual soil colour or material often
correlate with small hollows and valleys mapped in 1927/8 and afterwards levelled;
2. The raised banks of canals dug during the Bonificá on occasion also contain archaeological
material, which must presumably have come from the immediately adjoining stretch of the canal;
3. We have several examples where our study confirms that sites located on hillocks were, along
with the soil, ‘smeared out’across the surrounding fields by tillage;
4. One hollow, which the owner informed us he had filled in using soil (containing Roman ceramics
and building materials) from elsewhere in the same field, coincides exactly with a patch of
inflated soil (area marked ‘B’).
These observations confirm the importance of landscape history as a factor biasing the results of field
surveys, and the need for a structured ‘source criticism’so that we can trace and correct such biases.
5
CONCLUSION
This case study has demonstrated the feasibility of employing GIS to extract and interpret land form
changes from historic elevation and land cover data. Although it is generally recognised that the
interpretation of survey results requires knowledge of local geopedology and landscape history, workers
have not yet gone very far with this approach. The present case study shows that, provided certain
requirements regarding data quality are met, historic elevation data can be used to track one of the most
important factors biasing the results of field surveys today – changes in land form caused by human
agricultural and construction activity. By comparing historic to recent elevation data, using GIS, maps can
be made of the location and approximate amount of deflation/inflation influencing the presence and
visibility of the archaeological record. Such maps can be used both to target future surveys to areas that
are likely to have survived undisturbed, and to re-interpret the results of older surveys.
REFERENCES
Attema, P.A.J. 1993
An Archaeological survey in the Pontine Region, A contribution to the early settlement history of South
Lazio (900-100 BC), Groningen (Ph.D. thesis).
Attema P.A.J., G.J. Burgers, M. Kleibrink and D.G. Yntema 1998
Case Studies in Indigenous Developments in Early Italian Centralisation and Urbanisation: a Dutch
Perspective, in European Journal of Archaeology, Vol. I:326-381.
Attema, P.A.J., E. van Joolen & P.M. van Leusen 2001
A Marginal Landscape: Field work on the beach ridge complex near Fogliano (South Lazio),
Palaeohistoria 41/42 (1998/1999):149-162.
Hageman, J.B. & D.A. Bennett 2000
Construction of Digital Elevation Models for Archaeological Applications, in Westcott, K.L. & R.J.
Brandon (eds), Practical Applications of GIS for Archaeologists, pp 114-127. London: Taylor & Francis.
IGM 1927 (Year 5)
Bonificazione Pontina. Topographic map in scale 1:5,000, published by the Istituto Geografico Militare on
behalf of the Consorzio della Bonificazione Pontina. Firenze: IGM.
IGM 1996
Catalogo 1996. Produzione e Prestazioni dell’IGM. Firenze: IGM.
17 -
11
H I S T O R I C R ELIEF C H A N G E I N T H E P O N T I N E P LAIN
Kamermans, H. 1993
Archeologie en landevaluatie in de Agro Pontino (Lazio, Italië). Unpublished PhD thesis, University of
Amsterdam.
Sevink, J., A. Remmelzwaal & O. Spaargaren 1984
The soils of southern Lazio and adjacent Campania (Reports of the Fysisch Geografisch en Bodemkundig
Laboratorium 38). Amsterdam: University of Amsterdam.
Van Leusen, P.M. 1998
Archaic Settlement and Early Roman Colonisation of the Lepine Foothills, in Assemblage 4(1998),
http://www.shef.ac.uk/~assem/4/.
Van Leusen, P.M. 2001
Marginal Landscapes: Survey and interpretation biases in low finds density regions in Italy, in M. Gojda &
T. Darvill (eds), One Land, Many Landscapes (BAR S987):71-3. Oxford: Archaeopress.
Veenman, F.A. 1996
Landgebruik in de Pontijnse regio (1900-1100 BC). Een paleobotanisch en archeologisch onderzoek.
Unpublished MA thesis, University of Groningen.
17 -
12
C HAPTER 18
SUMMARY AND CONCLUSIONS
1 AIMS AND APPROACHE S
This thesis is the result of studies performed in the context of two separate multi-annual research projects
conducted in widely separated parts of Europe. Between 1994 and 1997 I was part of the Wroxeter
Hinterland Project (WHP) directed by Dr Vincent Gaffney at the Birmingham University Field
Archaeology Unit (BUFAU). The aim of this project was to relate the growth, flowering, and decline of
the Roman civitas capital Viroconium (modern-day Wroxeter, Shropshire) to its largely indigenous
hinterland, which in the late pre-Roman Iron Age was settled by the tribe of the Cornovii. From 1997 to
2001 I conducted my dissertation research at the University of Groningen within the Regional Pathways
to Complexity (RPC) project. Within this umbrella project directed by Dr Peter Attema (RUG) and Dr
Gert-Jan Burgers (VU), I studied methods of comparison of settlement dynamics and land use from late
protohistory until the Roman Imperial period in three Italian regions – the Pontine region in southern
Lazio, the Salento Isthmus in Puglia, and the Sibaritide on the Ionian gulf in northern Calabria. Both
projects are similar in that they focus on the combined ‘classical’ issues of Romanization and
urbanization, apply a regional scale of analysis, and attempt to restore indigenous populations and elites to
their ‘rightful’place in history.
PROBLEM ORIENTATION (CHAPTERS 1 & 2)
The archaeological problem definition in both projects revolves around the relationship between the
internally driven dynamics of the indigenous societies and the role of external colonizers. Whilst the latter
was widely perceived until recently to be the driving force behind supra-regional processes such as
centralization and urbanization, perspectives have changed since the 1980s and indigenous roles are now
seen to be as important as that of the colonizing powers, if not more so. Because such a perspective can
receive little or no support from historical sources, archaeologists must employ other tools such as
ethnographic comparison; they must also compensate their lack of knowledge of the indigenous nonurban landscape with new targeted fieldwork and the study of indigenous patterns of land use and
settlement.
The methodological problem consists in the combination of two facts. Firstly, the available archaeological
data was for the most part not collected with modern landscape archaeological aims in mind; rather, for
the past century or more, archaeologists have worked within a classical culture-historical paradigm which
has defined the goals and scope of archaeological research. More-over, it has become evident that
patterns in archaeological data at any spatial scale can be caused by systematic biases in the these data have
been collected and turned into archaeological records. When, as in landscape archaeology, it is intended to
study archaeological remains in conjunction with the landscape, ways must be found to address this
problem. Secondly, there is no accepted methodology for deriving regional and supra-regional
interpretations of settlement dynamics on the basis of the archaeological remains alone. That is, we do not know
how to do landscape archaeology without the culture-historical prejudice. In the terminology of New
Archaeology, there is a general lack of middle range theory, needed to link the archaeological evidence to
the culture-historical interpretation in a substantive manner. Much of this thesis consists of explorations
into this gray area. The comparison of regional patterns and long-term trends in settlement and land use is
approached here from a geographical point of view, which has mainly been implemented through the
(study of the) application of geographical information systems (GIS) software. A large part of this thesis
18 -
1
SUMMARY AND CONCLUSIONS
therefore consists of articles presenting and illustrating aspects of GIS methodology that were found to
be directly relevant to the analysis and interpretation of regional archaeological data.
Chapter 2 presents these archaeological and methodological problems in detail, using examples taken
from the three Italian RPC project study regions, and beginning with a brief review of regional settlement
dynamics as these are presented in recent literature. The societal processes considered most important for
the 1st millennium BC, and the most important archaeological concepts, theories, and methods that are
used in this thesis, are isolated and discussed.
Centralization, urbanization, and colonization are introduced as the most important, but at the same time
problematic, concepts. Centralization is a process affecting all of society: power (whether religious,
economic, political, military or administrative) becomes progressively concentrated in the hands of fewer
families or persons, settlement structure is tending toward nucleation, and economic life (production,
storage, trade, and exchange), cult, and construction is progressively mobilized and concentrated in a
fewer number of locations. The process, occurring generally in the Mediterranean between the Bronze
Age and the Archaic, is thought to have ultimately been driven by demographic expansion, through the
gradual increase in supra-regional contacts and the successful introduction of Aegean technology (olive
culture, storage of oil in pithoi). Urbanization (and, in its early stages, proto-urbanization) in its
indigenous form is a direct consequence of the process of centralization, but has in all four study areas
been deeply affected by the external imposition of urban forms – the Greek colonies in the Italian south,
and the Roman colonies in south Lazio and central Shropshire. This colonization by Greeks and Romans
was not a unitary phenomenon but took place for many different reasons and across several centuries. In
its early phase, lasting perhaps a century, it was characterized by relatively small populations planted for
economic or strategic reasons (late 8th – 7th century Greek colonies in the Tarantide and the Sibaritide, 5th
century Roman colonies in south Lazio). Population and settlement expansion, and its attendant
encroachment on and eventual domination of indigenous societies, occurred only later (6th century
expansion of the Greek colonies, 4th century expansion of the Roman colonies). The geographical
progression of this latter phase can to some extent be attested archaeologically in the material evidence
associated with Hellenization and Romanization: this occurs in the 5th/4th century on the Lepine side of
the Pontine plain, in the late 4th/3rd century in the Salento Murge and the foothills and uplands of the
Sibaritide, with the late 3rd/2nd century Roman conquest of the Greek south, in the late 2nd/1st century in
the coastal side of the Pontine plain, and in the 1st century AD in outlying provinces of the Roman
empire, such as Shropshire.
This review of concepts is followed by an analysis of the theoretical basis for interregional comparison,
presenting the advantages and disadvantages of the various approaches and opting for a quantitative
approach that stays closer to the archaeological data than the hitherto usually applied qualitative, sociopolitical explanatory models. Finally, a brief preliminary exploration of qualitative and quantitative
comparison between the three study regions is presented.
The introductory section of the thesis ends with a chapter (3) introducing in more detail the problematic
of Romanization in the Wroxeter Hinterland and the aims and approaches of the WHP, using two articles
published in 1996-7 in conference proceedings. The first article defines the aim of the project – explaining
the anomalous existence of Wroxeter itself, which was the fourth largest town in Roman Britain but
seemingly lacked the developed rural hinterland needed to support it. The project’s goal is defined as an
attempt to establish Wroxeter’s place in Millett’s (1990) models for urbanization and Romanization in
outlying provinces of the Roman Empire, both through the study of the available archaeological records
and through an extensive program of systematic field survey. The second article uses both direct evidence
and theoretical arguments to reject three current explanations for the existence of a flourishing provincial
Roman town in a barely Romanized hinterland; namely 1) that Wroxeter had been an overly ambitious
foundation and never reached high population levels, 2) that the rural population remained hostile to the
colonizer and resisted acculturation, and 3) that Wroxeter never developed a strong economic base. The
first of these explanations was refuted directly by the project’s full geophysical and aerial photographic
study of the town itself, and the two remaining explanations are rejected as well on the basis of the
argument that the obvious economic and demographic health of the town implies that a wealthy pre-
18 -
2
V A N L EUSEN : P A T T E R N T O P ROCESS
Roman hinterland existed and that the native Cornovian elite was happy to invest this wealth in the town.
An alternative explanation for the apparent economic and cultural contrast between town and hinterland
is then presented, based on two arguments: firstly, that the wealth of the native pre-Roman Cornovians
took on archaeologically invisible forms (cattle, salt) probably as a consequence of poor access to foreign
trade routes (hence a lack of means to generate prestige through the acquisition and redistribution of
foreign goods), and secondly, that our current knowledge of late Iron Age and Roman settlement and
land use in the region is biased by a lack of systematic research.
2 FIELD WORK IN ITAL Y ( 1 9 9 8 – 2 0 0 0 )
Our knowledge of the archaeology of all three RPC regions was, at the start of the project, composed
mainly of Italian studies that took place from the 1960’s onwards, and Dutch research projects combining
excavation and survey from about 1980 onwards. Gaps in this knowledge can, for most of this period, be
traced in a fairly straightforward manner to the archaeologists’preponderant interest in ‘high’classical
culture to the detriment of earlier and later periods, linked to a disregard of ‘low’culture and the rural
landscape in favor of cult places and urban settlements with their architectural remains and their
cemeteries. It is unfortunate that the lack of information about the spatial coverage of these studies means
we cannot even take the relatively robust classical patterns for granted. Likewise, the large-scale Agro
Pontino survey conducted by the University of Amsterdam, which applied a random transect sampling
strategy from which to extrapolate over all of the Pontine plain, remains unfortunately unpublished, and
their methods of collecting and recording ceramic data do not seem to allow detailed pattern analysis.
Regarding the later more intensive and systematic Dutch surveys of the 1990’s, their location within the
study regions was evidently biased toward urban settlements and their immediate hinterlands and ignored
rural and ‘marginal’parts of the landscape. In short, the available data displayed significant geographical,
chronological, and typological biases. During the period 1998-2000 the RPC project team, through its
program of intensive and systematic archaeological field surveys in all three study regions, has contributed
towards the filling of these various gaps in the existing archaeological record. Four preliminary reports on
this fieldwork, published with other members of the RPC project team, were presented in chapters 9-12,
preceded by a summary of the overarching goals and results of the RPC fieldwork program in chapter 8.
At the same time attention was also turned to the development of a suitable methodology for conducting
the field surveys themselves and for the analysis of the resulting data. These two topics are summarized in
the following sections.
2.1 FIELD WORK
PONTINE REGION
The 1998-1999 RPC surveys at the foot of the Lepine mountains near the deserted Medieval village of
Ninfa, and around the Lago di Fogliano on the Pontine coast, were reported in chapters 9 and 10.
Although the original aim of the Ninfa fieldwork was the continuation of a mapping program of so-called
‘platform’villas, the study area also fell just inside one of the map sheets of the Forma Italiae series of
archaeological maps (Cora, Vittucci Brandizzi 1968) and could therefore be used to examine the degree to
which this older Italian data set had succeeded in capturing the diachronic archaeological landscape. It
was demonstrated that the study area contained, in addition to the almost exclusively Roman monumental
remains mapped by Vittucci, many smaller and less obtrusive Roman sites. Moreover, the area contained
ample remains of an intensively used pre-Roman (Archaic and post-Archaic) landscape which had not
been registered by her at all. In view of these results, the settlement history of this landscape unit (the
‘northern colluvium’, which also includes the proto-urban settlement at Caracupa/Valvisciolo) now
appears to have more in common with that of the core area of Latial society in the Alban hills than with
that of the Pontine plain proper, for which intensive settlement and land use has not been demonstrated
until the middle Republican period. Following the recognition of variations in size, situation, and status of
18 -
3
SUMMARY AND CONCLUSIONS
the Roman rural villa sites in the Ninfa area, we have now begun to delineate the contours of a more
detailed rural Roman settlement hierarchy. The following conclusions regarding urbanization and
colonization were drawn:
• Archaic settlement in the Ninfa area, while dispersed, appears to have been quite dense – an
indication that the population certainly was not ‘urbanized’(in the sense of ‘nucleated’) to a very high
degree. It may be that incipient Latin urbanization was halted in the northern colluvium by the end of
the Archaic as circumstances became less favorable through sporadic warfare, and the inhabitants
were forced to resettle into smaller and more easily defended sites on the Lepine scarp during the late
and post-Archaic (550 – 350 BC). Some of the latter would then have been targeted for the early (i.e.,
early 5th century) strategic Roman colonization.
• The fact and nature of Roman colonization in the Lepine slopes is to a large extent predicated on the
presence or absence, between the Archaic and Roman periods, of a post-Archaic hiatus in the
settlement history of the area. Dispersed post-Archaic settlement existed in the Cisterna area only a
few kilometers to the west (Attema 1993:181ff) and we may therefore assume that there was
settlement continuity throughout the post-Archaic. The nature of the change from slight buildings
and thick augite tempered pottery in the Archaic, to tile-roofed buildings with cisterns and depurated
amphora and fine wares in the Republic remains to be explained. Are we looking at the
transformation of temporary shelters into permanent habitations? Was the indigenous population
moved to make way for new Roman settlers whose farmsteads were constructed according to some
colonial base plan, or do the changes visible in the archaeological record reflect the gradual
integration of the local Latial rural economy into the same regional economic and cultural networks
that also include their Roman allies?
The Fogliano fieldwork area was selected because it represents the coastal landscape of the Pontine
region, regarded as economically, politically, and demographically ‘marginal’on the basis of both classical
and more recent historical sources. The results of the survey support this hypothesis for the time up to
and including the middle Republican period, establishing a basic settlement history that can probably be
extrapolated to the whole Pontine coastal landscape. Sporadic ceramic evidence was encountered for
what was probably non-intensive land use and impermanent settlement in this landscape of ancient beach
ridges, valleys, and coastal lagoons from the Bronze Age onwards, and the number and ceramic density of
sites only begins to increase in the Archaic period. In the absence of sufficiently diagnostic ceramic types,
the intensity of land use during the post-Archaic and middle Republican periods must remain unclear for
the present, but a clear departure from the status quo ante occurs with the remarkable growth in the
number of rural villas during the late Republican period (200 – 50 BC). As this growth takes place mostly
in the single largest available continuous and flat area of suitable sandy (eolian) soils, we interpret it as the
development of a rural village. This rural socio-economic development was tentatively connected to the
production and supra-regional trade in fish and fish products that emerged in this period along the
Pontine coast, but apparently lasted not more than two centuries since none of the sites appears to have
been in use after the early Imperial period; a decline that is in line with general economic trends in the
expanding Roman empire.
The results of these surveys were put in the context of diachronic developments in the wider region: the
development during the Bronze Age, Iron Age and Archaic period (i.e. to the 6th century BC) in the core
areas around the Alban massif and Rome of, first, centralized settlements and, later, peer polity city states
is reflected by a similar, but late and stunted, development of more marginally located polities such as
Caracupa / Valvisciolo on the Lepine footslopes and Cisterna di Latina on the south-eastern margin of
the Alban massif. During the post-Archaic and Republican period the growing political, military and
economic influence of Rome expressed itself archaeologically first in the establishment of colonies on the
Lepine margin and mixed farming on the colluvial slopes and (though much less so) along the Via Appia.
Only much later did it result in the exploitation of the coastal landscape for fish farming, pottery
production and leisure industry. The apparent dismantling of the Lepine olive culture and the near
abandonment of settlement there and in the coastal area following the early Empire indicates that the
18 -
4
V A N L EUSEN : P A T T E R N T O P ROCESS
Pontine region generally became economically marginalized as the Roman Empire moved its large-scale
agriculture and service industries elsewhere.
SALENTO ISTHMUS
Just as at Fogliano, the aim of the field survey conducted in 1999 near the town of Ostuni in the Salentine
Murge (chapter 11) was to map in detail some previously barely studied ‘marginal’landscape units. The
limestone plateau of the Murge had long been seen as constituting one of the social, economic, and
geographical margins of the lowland urban society developing in the early Hellenistic period in the
Salento. Whereas research by the University of Lecce and the Free University of Amsterdam had been
concentrated on these lowland central places and their immediate hinterlands, the Ostuni survey for the
first time offered a chance to chart in detail the long-term history of sections of both the high Murge and
the transitional zone toward the coastal plain of the Adriatic. On the micro-regional scale of
interpretation, one of the major conclusions that were drawn on the basis of the survey is that, in broad
outline, both landscape units demonstrate parallel shifts in artefact densities and distributions from the
Bronze Age to within the early Imperial period:
• With regard to the landscape history of the protohistoric, classical, and archaic periods, the almost
complete absence of finds dating from the Late Bronze Age until the 4th century BC confirmed that
society during that period probably had a strongly nucleated structure, centering on strategically
located cliff and hilltop settlements. Archaeological material dating to these centuries is restricted to
the cave site of S. Maria d’Agnano, where the Archaic formalization of cult activities can be argued to
have supported territorial claims on the surrounding land. The fluidity of protohistoric land use
strategies was illustrated when the survey unexpectedly found widespread and often dense scatters of
a very homogeneous impasto dating to the Middle Bronze Age. These scatters, occurring in large
numbers in both upland and lowland settings, must be interpreted as the remains of a system of
shifting cultivation that was in use for a relatively brief period, and in which family groups
periodically relocated to exploit fresh or regenerated parts of the landscape. During this time no large
and permanently inhabited settlements would have existed in the area.
• For the Hellenistic and Roman periods the survey confirmed the expected low intensity of land use,
in that individual farmstead sites were found to be located approximately 1 km apart in both
landscape units. On the other hand the survey provided surprising evidence that ‘colonial’ceramics,
building materials, and building styles had already penetrated far into the Murge during the early
Hellenistic period. The development of an indigenous-Hellenistic urbanized society on the Salento
Isthmus was therefore complemented by a contemporaneous Hellenization, possibly even
colonization, of even the most remote areas. This suggests that essentially all of the population was
involved in and affected by this process. For both survey areas, a basic continuity of occupation
throughout the Roman Republican and Imperial periods can be deduced, with the possible exception
of the late Imperial period in the upland area.
SIBARITIDE
For the survey in the Sibaritide (2000; chapter 12), field work was aimed at testing several hypotheses
generated on the basis of Lorenzo Quilici’s large-scale topographic survey of the late 1960’s, while at the
same time charting in more detail part of the archaeological hinterland of the protohistorical settlement
and cult place on the Timpone della Motta near Francavilla Marittima, excavated in the 1990s by M.
Kleibrink. From Quilici’s survey, as from other similar surveys conducted for the Forma Italiae series, had
emerged an intensively settled but almost exclusively classical (Hellenistic-Roman) landscape. Within the
survey zone, these farmstead sites appeared to cluster into loose-knit villages linked by Quilici to a
(hypothetical) regional and supraregional infrastructure. The RPC survey tested these ideas by means of a
field work transect through the foothill zone, designed to reveal the existence of any systematic
chronological and geographical biases or lacunae.
18 -
5
SUMMARY AND CONCLUSIONS
The results of the survey confirmed most of Quilici’s ideas. In view of the very sporadic presence of
protohistoric surface ceramics the foothill zone appears not to have been intensively used before the
archaic expansion of the sphere of influence of the Greek colony at Sybaris. The adverse conditions for
retrieval of protohistoric ceramics on the intensively worked terraces must be taken into account when
interpreting the archaeological record of relatively well-preserved upland protohistoric sites. Despite the
historical date for the establishment of Sybaris around 720 BC and the archaeological evidence for
continued use of the sanctuary and necropolis at neighboring Timpone Motta into the late 6th century BC
(Attema et al. 2000), no securely datable materials from this period were found. Given our experience
with the very low visibility of coarse Iron Age impasto wares among the naturally occurring stones in the
survey area, we concluded that neither ourselves nor the Quilici team were able to identify such material
with any degree of reliability; we cannot therefore infer much from its absence. For the Archaic, much
will depend on a closer dating of the coarse wares, which make up more than half of the finds by weight,
through association with datable fine wares or through typological comparison with excavated material
within the region. The large numbers of classical sites do indeed originate, for the most part, in the early
Hellenistic period; and they do indeed cluster along the edges of specific landscape units (gently sloping
plateaus of marine origin). Among the ‘new’small classical sites identified by the intensive survey several
are located within clusters identified by Quilici, while others are scattered all over; the material recovered
from these mostly small Hellenistic farmsteads is remarkable for its uniform poverty.
In other respects, however, Quilici’s thoughts on site distributions and settlement history have had to be
modified as a result of our survey. Despite the deterioration of the soil archive in the intervening decades
about twice the number of sites mapped by Quilici were recorded in the same area by our more detailed
and complete coverage. Large and small Hellenistic sites were found to occur in other landscape settings
(among them secondary plateau edges of fluvial origin) besides the ones identified by Quilici as well, and
most appeared to be discontinued in the Roman period. The survey also established that just over half of
the undiagnostic surface ceramics in the transect displays fabric characteristics potentially placing them in
the Classical/Archaic rather than the Hellenistic period. If this dating can be confirmed it has obvious
consequences for the settlement history of the Sibaritide as a whole, substantiating historic accounts of
the imperium of Sybaris.
2.2 FIELD METHODS
Topographical and field surveys can cover relatively large areas, but the diversity of approaches, the
potential for significant bias, and the lack of an accepted approach to statistical analysis of the resulting
data mean that their interpretation is a matter often left exclusively to the judgement of the survey
director. The promise of the large field survey projects based on regional sampling designs of the 1970’s
and 1980’s, of providing a firm basis for regional extrapolation and statistical inference, has not been
fulfilled, and more recent survey projects (including those of the RPC) have elected to map the surface
record at a certain resolution, rather than sampling it at a certain fraction. Accordingly, the RPC surveys
were used to experiment with methods for more objective recording of archaeological landscapes
(summarized in chapter 8). In order to alleviate the resulting substantial load of administrative procedures,
experiments were also conducted with automated and digital field recording (chapter 7).
FIELD PROCEDURES
With regard to methodology, the 1998 Ninfa survey provided a valuable first insight in and confirmation
of the limitations of the old topographical style of regional survey, confirming the need to develop
methods for the registration of continuous spreads of ceramics across the landscape (as opposed to discrete
spreads in the form of ‘sites’) in later field surveys - a lesson taken to heart in experiments during the 1998
survey campaign at Fogliano. The conduct and recording of the fieldwork there and in all later surveys
was based on geographical collection units (land parcels of circa 1 hectare) instead of archaeological units
(sites) or agricultural units (fields), and the problem of selective recording was fought by adopting a policy
of collecting and bagging all surface material per unit, leaving the classification of finds by an expert to a
later stage.
18 -
6
V A N L EUSEN : P A T T E R N T O P ROCESS
In addition to the realization that the history of settlement and land use in the Pontine region can be
more easily understood if we analyze it in terms of relatively small physical landscape units than if we
attempt to do so for the region as a whole, the Fogliano survey also highlighted a problematic recognized
earlier by British archaeologists during the analysis of surveys in the Aegean. This problematic regards the
interpretation of the often very low ‘off-site’finds density of ceramic from various periods carpeting the
Mediterranean landscape. Efforts to obtain a better understanding of the factors that influence the
probability of retrieving surface finds by revisiting find spots from the 1998 campaign in 1999 and 2001
led me to believe (with others) that even a single ceramic find should, under certain circumstances, be
interpreted as indicative of the existence of a local subsurface reservoir (site). The importance of factors
influencing the survival and visibility of ceramics in the plough-soil was shown to vary greatly, more-over,
depending on the manner of production and age of the material. This causes protohistorical ceramics, for
example, to have a much lower probability of retrieval in a field survey than is the case for classical
Roman and Hellenistic ceramics, and is an obvious cause of bias.
The experience gained in these earlier survey campaigns in the Pontine region was employed to improve
aspects of methodology, and later surveys in the Salento and Sibaritide were conducted with a higher
spatial resolution (units of 0.25 ha) and a more rigorous registration of factors affecting visibility. An
appendix to chapter 8 presents the annotated field forms developed in the course of these surveys; they
represent the current stage in an ongoing process of developing field administration procedures that
satisfy internationally accepted standards of good practice in the conduct of field surveys.
• By applying a detailed survey method, focusing on the documentation of the density distribution of
artefacts rather than of sites, it has become possible to assess accurately the variability in quality and
quantity of this surface material in the light of both cultural and natural formation processes.
• The goal of formal comparison of survey results can only be reached if procedural standardization
and rigor are applied in the recording and processing of field data. This will involve greater use of
digital recording and wireless transmission methods to increase the precision and efficiency of field
surveys; the more detailed recording of landscape parameters that may affect not only site location,
but also site visibility; the realization that survey biases may differ, and must therefore be assessed
separately, for each landscape unit and category of archaeological material; and the development of
widely accepted definitions of such crucial concepts as ‘site’,‘off-site’,and ‘scatter’.
DIGITAL RECORDING METHODS
Chapter 7 is dedicated to a description of experiments conducted in collaboration with Dr Nick Ryan of
the University of Kent at Canterbury during the October, 2000, fieldwork in the Sibaritide. The aim of
these experiments was to increase the accuracy and efficiency of the recording and processing of
information during and after archaeological field surveys by using programmable, lightweight, and semiautomated digital registration tools (digital field assistants or dFA’s). By conducting instant digital
recording of landscape parameters and collected materials in the field (that is, without first passing
through a ‘hardcopy’stage) the efficiency of the work is increased and the danger of transcription error
removed. By pairing such digital field recording with automatic and accurate GPS location methods, the
mapping of collection units and archaeological entities is no longer dependent on less precise manual
methods often involving outdated topographical maps. This apparatus was used successfully to record the
surveyors’routes and observations, agricultural field boundaries, and the center and circumference of
archaeological sites. The experiments confirmed the potential of the dFA system for both speeding up
field recording procedures and reducing the number and size of errors made during the recording
process. The system’s potential for easing navigation and the sharing of information during surveys was
not fully explored, but our experience in (re-)locating archaeological sites mapped in the 1960's indicates
that it will also prove useful in that area. With limited enhancements to functionality, and further
improvements to the standard spatial accuracy, it was demonstrated that the system can profitably be used
in any type of archaeological field survey. With the full technical and procedural integration of a
professional version of the kit into fieldwork methodology dFAs may transform fieldwork practice, but
18 -
7
SUMMARY AND CONCLUSIONS
this will require further extensive testing of system components, software, and field procedures.
In recent years, the use of professional GPS surveying equipment in archaeological fieldwork has become
much more popular, and some teams are adopting commercially available products in order to obtain
GIS-like capabilities in combination with GPS. These high-powered approaches, while providing very
high accuracy and versatility, require considerable expense and highly skilled personnel, and cannot yet
provide a true field information system. The digital Field Assistant system was argued to be preferable
over such alternative approaches, because it is relatively inexpensive, can provide immediate feedback in
the field, is portable and unobtrusive, and is designed to perform typical and frequent archaeological
fieldwork tasks. With respect to the GPS component of the dFA system, the availability of a good
location device is a crucial feature in the recent shift in emphasis of archaeological survey work away from
the well-mapped and well-controlled coastal zones of Italy, to the more rugged and less well-mapped
inland zones. For archaeological applications where the accuracy requirements are higher than what can
be achieved with a single receiver, the addition of a GPS base station for differential correction would
give the most satisfactory results. Post-processing, of course, would not offer any improvement in realtime positioning in the field but so far we have not identified any reason why this should be a high
priority. Should it become necessary, corrections could be broadcast from the base station and received at
the rovers by using conventional wireless-modems.
3 T H E M E T H O D O L O G Y O F REGIONAL COMPARISON
The process of compilation, comparison, and interpretation of regional data sets from a quantitative point
of view was illustrated in chapter 13, using the Pontine region as an example. Included is a discussion of
the desired structure of regional relational archaeological databases, of the need for unambiguous and
standardized definitions of archaeological entities such as the ‘permanent habitation site’, and of the
remarkable lack of standardization in fieldwork methods and publication which hampers even the most
basic comparison between two or more archaeological data sets. It is shown that traditional regional site
databases do not provide a good basis for storing the new data types and emphasis on uncertainty and
fuzziness inherent in modern landscape archeological data. In future, archaeological databases such as
that of the RPC project should contain the new types of entities and relations needed to describe
accurately all types of archaeological field observations; they should fully document the process of
interpretation of these observations along with the interpretations themselves; they should contain
mechanisms for keeping track of data quality and for improving data quality through bias modeling if
necessary; and they should use formal authorities for the chronological and typological classification of
source data and interpretative constructs. To ensure that such databases can be used by others, the need
for metadata describing the database itself, and for the development and implementation of explicit and
formal classifications was argued.
From the quantitative comparison of the results of all the available surveys in the Pontine region, it
emerges very clearly that less intensive survey (such as practiced until the early 1990s) has tended to result
in the discovery of a predominantly classical landscape, because sites from this period are the most
obtrusive (large and dense scatters containing both tile and ceramics, standing architecture, many
diagnostic wares). More intensive survey results in an explosive growth in the number of small and/or
diffuse ceramic scatters identified, for which chronological and functional typology tends to be much less
clear-cut. A further hurdle to the comparison of survey data sets across projects is presented by the lack
of a standardized approach in almost every aspect of their collection, description, interpretation, and
publication. This is demonstrated by the example of several incongruent chronological and typological
site classifications in recent use in survey projects in central Italy.
It is concluded that priorities for future work in the regional or interregional comparison of survey data
must lie with the removal of weaknesses in the core classifications employed. Local ceramic
chronotypology must be further developed through fabric classification and seriation of survey
assemblages, especially for the post-Archaic period. Site type classifications, particularly for the pre18 -
8
V A N L EUSEN : P A T T E R N T O P ROCESS
Roman periods, should be improved through a program of field tests involving surveys, trial trenches,
and excavations at a representative sample of site types. Surveys should address the current lack of data
for what has traditionally been considered the marginal parts of the ancient landscape; especially the upand highlands. Furthermore, the intensive surveys conducted in the various landscape units of the
Pontine region since the early 1990s now provide a solid basis to revive the idea of a regional stratified
sampling design. The conduct of future survey campaigns within such an overarching strategy can result
in a more efficient use of limited resources and should generate more easily comparable data for the
region as a whole.
3.1 MODELING DATA F ORMATION PROCESSES
The interpretation of large-scale (wide-area) patterns in archaeological landscapes has in the past always
taken place within the context and limits created by the available written sources from classical antiquity.
In order to escape from these limitations and base one’s interpretations directly on patterns in the
available archaeological data, suitable methods must first be developed. Two types of methodological
studies were accordingly undertaken. Firstly, studies that aimed at obtaining an understanding of, and
control over, the quality of the archaeological data that lies at the heart of regional settlement histories
and the comparisons based on them (chapter 4). Secondly, studies that aimed to assess the utility of the
GIS toolkit for the analysis and interpretation of patterns in those same archaeological data (chapters 14
and 17).
DATA QUALITY
On the regional scale, geographical, typological and chronological biases abound. There is, for example, a
great lack of survey data for what has traditionally been considered the marginal parts of the ancient
landscape - especially the up- and highlands in the case of Italy; and if theoretical arguments for the
‘invisibility’of the majority of small protohistoric settlements are correct, then what appears to be a clearcut process of proto-urbanization taking place in Italy, may in fact be the preferential discovery of the
highest-ranked settlements within a much more widely settled landscape. Chronological biases are
especially insidious, as they are built in to typochronological classifications which are often imported from
outside the region being studied. For example, the early settlement histories of the Pontine region and the
Sibaritide are greatly influenced by our (in-)ability to recognize post-Archaic, respectively Archaic
ceramics. Rather than continuing to rely for dating on fine wares originating in other regions and often
itself dated only by typological association with wares from even further afield, it is therefore of singular
importance to study the local fabrics and wares in excavated contexts. Local ceramic chronotypologies
must be further developed through fabric classification (cf. Attema 2000) and seriation of survey
assemblages.
The problems that currently haunt the interpretation of the results of surveys by one and the same group
operating in the same area, are shared to an even greater extent by those intending to assess the
archaeological record at a wider regional, or even supra-regional, scale. Currently, students are forced to
choose between two equally unattractive approaches to regional and interregional comparison: either to
compare the broad characteristics of the data sets while ignoring most of the associated problems, or to
spend a huge effort on devising a multitude of correction methods for low quality data sets. Qualitative
comparison of high-level interpretative constructs is doable because these are provided ready-made by
period and area experts, but unsatisfactory because the interpretations 1) rely on a shared and limited set
of theoretical constructs, and 2) are heavily biased by past research trends and results. Rather than
continuing to collect data that suffer from these biases, research should be targeted at the development of
more reliable methods of collecting data that are truly representative of the extant archaeological
landscape, and of standardized ways of describing, archiving, and publishing results. Clearly there is an
urgent need to begin to develop landscape archaeological data sets of a sufficiently high quality to allow
(inter-)regional comparison – guidelines for which should become embodied in an international standard
defining ‘best practice in landscape archaeology’.
18 -
9
SUMMARY AND CONCLUSIONS
BIAS MODELING
The strength of our interpretations of regional archaeological databases, such as are compiled by desktop
study and occasional fieldwork, rests entirely on the quality of the data within it. With the exception of a
few survey projects designed in the late 1970’s under the influence of the New Archaeology, and
implemented mostly in the 1980’s, none of those data were produced with the express aim of obtaining a
representative picture of the regional archaeological landscape. Regional interpretations must therefore
explicitly take into account the possibility that the data on which they are based are not representative for
that landscape. A similar problem has plagued the interpretation of the results of archaeological field
walking surveys, which had become increasingly popular during that same period as well. Ever more
intensive and systematic field work has brought to light the significant role of numbers of factors biasing
the objective retrieval of archaeological surface data. Chapter 4 deals with methods that can be used to
detect and counter such biases, both proactively by introducing procedural improvements in the design
and execution of contemporary fieldwork, and retroactively by applying extensive source criticism on data
sets formed in the past:
• Our ability to record surface archaeological material is not perfect; it is biased by visibility and
research biases. Causing the former are factors such as current and historical land use / land cover
(LULC); causes of the latter include the recording and classification methods used by the field
archaeologist. The amount and type of bias varies strongly depending on the type of data and scale of
analysis; no hard and fast rules can therefore be given, other than that a bias study should be a
requisite part of any regional data collection exercise or analysis.
• Neither the intensive interest and study conducted in the early 1980s, nor the growing popularity of
surveying and use of GIS since then, have so far led to anything resembling a successful approach to
the recording and correction of biases which is valid across projects. We must follow up on
Terrenato’s (1996) urgent call to record bias factors if we are to attempt ‘the correction, at least
partially, of incomplete distributions’, and conduct the ‘series of methodological experiments dealing
with the various aspects of how to document surface scatters’,advocated by him.
The case studies presented here were conducted to demonstrate a) the relevance of bias factors to the
interpretation of survey data and of landscape archaeological data in general; and b) methods by which
bias factors can be included in geographic models of archaeological landscapes. At the regional scale,
studies of the data collected by the Wroxeter Hinterland Project and the Agro Pontino Project (Voorrips
et al. 1991) demonstrate this for systematically surveyed data and general archaeological records; at the
scale of a ‘local’ survey such as the Ninfa and Fogliano surveys conducted 1998-9, case studies
demonstrate this for specific visibility and research biases.
CHANGES IN LAND USE AND LAND FORM
The geological history and the history of land use of the landscape have a great influence on the design
and results of archaeological fieldwork. In the WHP surveys conducted in 1994-6, the choice of fields was
limited by modern land use and land cover (LULC), in particular the availability of recently ploughed
agricultural fields. Since such fields are not randomly distributed over the landscape – relief, distance to
the river Severn, soil type, modern infrastructure and hydrology all play a role – the surveys result in the
taking of a potentially biased sample which cannot be used to make straightforward extrapolations about
the study area as a whole. In the surveys conducted by the Agro Pontino project during the 1980s,
paleosurfaces dating to the Paleolithic period had been covered in most parts of the Pontine plain by
more recent alluvial and colluvial deposits (Kamermans 1993), and similar though less clearly evident
biases must have been present for material dating to later periods. In chapters 14 and 17 of my thesis I
presented examples of bias modeling applications focusing on the recent and subrecent land use history
of the Wroxeter hinterland and the Pontine plain. It was demonstrated that 20th century land use in the
former region is correlated with the large-scale distribution patterns of several of the most numerous site
types in the former area. Land form changes as a result of the fascist and later land improvement schemes
18 -
10
V A N L EUSEN : P A T T E R N T O P ROCESS
in the latter region were likewise shown to have a significant influence on the results of the RPC field
surveys near Fogliano (chapter 10).
The study of LULC for regional archaeological research can be said to have both a methodological and a
historical purpose. The former is perhaps best approached by the use of GIS to store and compare
historical cartographic data about land use. Such data may be derived from archival records made for
military, legal or taxation reasons; from studies of agricultural productivity; and, more recently, from
historic aerial photography and satellite imagery. The resulting LULC maps can then be compared and
correlated to the visibility and discovery mode aspects recorded in the archaeological archives. The case
study presented in chapter 14 found that recorded discovery mode in the Shropshire SMR does indeed
correlate with historic LULC. Examples are given of the positive correlation of aerial photographic data
to arable land use, of chance finds to areas more likely to be frequented in recent times, and of
earthworks to uncultivated land. Using the example of barrows and ring ditches, it was demonstrated that
analysis of the archaeological record within a particular discovery mode can also flag up significant
deviations. Whilst both feature groups represent a single underling class of sites (ring ditches being the
ploughed-out remains of burial mounds), barrows discovered by air are located mainly on poor soils while
ring ditches discovered by air occur mainly on rich soils. When the two data sets are recombined the latter
correlation becomes insignificant. The second, historical, use of LULC studies is demonstrated by
reconstructing a regional pattern of arable vs. woodland for the late Roman period supported by archival
studies of diocesan boundaries and literary topographic references. It is argued that such studies are
needed if we are to distinguish between archaeological patterns relating to ancient LULC, and those
relating to modern LULC.
Whilst the scope of this case study was limited both by the available time and by the relatively low quality
of the data available from the Shropshire SMR, there is no reason to assume that other regions will have
been much better documented. Our archaeological understanding of the Wroxeter hinterland should not
be conditioned by the ‘accidental’but systematic absence from the landscape of certain kinds of evidence
as recorded under certain kinds of conditions. Such hiatuses can be managed efficiently in a GIS by
mapping the type, coverage, and intensity of archaeological research events1 across the landscape. With
regard to the potential of land use mapping in Mediterranean studies, one may point to the many and
sometimes very detailed topographic maps of the Pontine region, which contain land use data dating back
as far as the 16th century. Attema (1993), for example, used military topographic maps of 1851 to derive a
map of the typical transhumant settlement and land use practiced until the early part of the 20th century in
the Pontine plain.
The case study presented in chapter 17 demonstrates the feasibility of employing GIS to extract and
interpret recent and subrecent land form changes from historic elevation and land cover data. Although it
is generally recognized that the interpretation of survey results requires knowledge of local geopedology
and landscape history, workers have not yet gone very far with this approach. The case study shows that,
provided certain requirements regarding data quality are met, historic elevation data can be used to track
one of the most important factors biasing the results of field surveys in Europe today – changes in land
form caused by agricultural and construction activity. By detailed comparison of historic and recent
elevation data in the GIS environment, maps can be made of the location and approximate amount of
soil deflation/inflation affecting the presence and visibility of the archaeological record. Such maps can be
used both to target future surveys to areas that are likely to have survived undisturbed, and to re-interpret
the results of older surveys.
Taken together, chapters 14 and 17 provide an especially clear demonstration of the strength of the
correlation between most small-scale (wide area) patterns in regional archaeological data sets, and the
combined factors of recent and subrecent land use and local research methods and interests. A more
The term ‘event’is here used in accordance with accepted archaeological recording practise, where it indicates a single episode
of relevant actions/observations affecting a monument.
1
18 -
11
SUMMARY AND CONCLUSIONS
extensive and detailed meta-analysis of Italian surveys is now needed in order to map in detail their
correlation with contemporary land use.
3.2
MODELING SETTLEMENT PA TTERNS
SITE LOCATION MODELS
In order to arrive at an assessment of the two main schools current in archaeological GIS studies of the
last decade, chapters 5 and 6 presented a full analysis of the theory, methodology, and methods
underlying ‘predictive’ models (location models mostly based on the extrapolation of correlations
between the locations of archaeological remains and characteristics of the physical landscape) and
‘cognitive’models (mostly models that relate the location of archaeological remains to the degree of
visibility and accessibility of the surrounding landscape). Predictive models were developed internationally
mainly in the context of heritage management and preservation, but in the form of location models have
long been the object of academic research as well. In the latter case the aim is usually to explain existing
settlement and land use patterns by relating them to landscape parameters. The potential of GIS software
as a tool for this kind of study was quickly recognized and in Europe led to a steadily increasing flow of
publications since the early 1990’s. During that same period, however, post-modern theory also gained an
increasing number of proponents within European archaeology, giving rise to heavy criticism of the
‘ecological determinism’ inherent in predictive models, and proposals to replace it by ‘cognitive’
alternatives. This led to a lively but chaotic debate regarding both the theoretical underpinnings and the
aims and methods of this type of geographical model. Based on an overview of the international literature
on the subject, chapter 5 lists and evaluates all of the arguments employed in this debate, in which a series
of dichotomies stemming from the polarized theoretical stances appear to predominate.
The main conclusion arising from this review is that procedural transparency rather than theoretical purity
should be the main characteristic of predictive models, a goal that can be reached only by formalizing all
the stages in the modeling process as presented in the chapter, increasing the quality of the data and
methods employed, and adequately testing the resulting models in the real world. Wide-area predictive
modeling using GIS is poised to play a very important role in archaeological heritage management at the
national level in the European Union because of the imminent implementation of the Valletta Treaty on
the protection of the cultural heritage, but at the same time has remained an important tool for
archaeological research as well. The ability to generate formal, rule-based, and testable hypotheses in the
form of predictive maps is fundamental to both types of use, and requires a better understanding of the
underlying theory, data and methods. Specific recommendations include the need to improve the spatial,
functional, and temporal resolution of the models, to arrange for the formal inclusion of archaeological
theory and expertise through the use of expert systems, and to incorporate formal stages of source
criticism (bias correction) and quality testing. In both CRM-type and academic models there is sufficient
scope for procedural improvement, including the proper and transparent use of statistical techniques and
inclusion of bias models. It should be clear to students and end users alike to what extent models are
supported by statistical inference, and what can and cannot be inferred from them. ‘Source criticism’of
both archaeological and environmental input data, including especially the absence of such data, should
be an integral part of the modeling methodology. This can perhaps be implemented through the use of
“taphonomical map layers” that assess the nature and extent of the distortions of the known material
heritage, and suggests a link with the area of error propagation modeling in Geography.
COGNITIVE MODELS
A heightened attention to the landscape as perceived and conceived by humans both in the past and
today is one of the more significant contributions of post-modern archaeology to the debate regarding the
nature and goals of GIS applications (chapter 6). In contrast to the external and physical characteristics of
the landscape, this approach targets its internal, cognitive, aspects. In everyday research practice it
translates into archaeological applications and development of two GIS tools in particular: line-of-sight
analysis (LOSA) and cost surface analysis (CSA). Leaving aside the fruitless theoretical debate
accompanying this development, research efforts in this direction have allowed a more realistic
18 -
12
V A N L EUSEN : P A T T E R N T O P ROCESS
approach to predictive modeling that takes into account the human experience of being in the landscape.
In chapter 6 I reviewed the international archaeological literature on these two techniques. It is concluded
that, a decade after the first LOSA / CSA studies were conducted, an initial phase characterized by naïve
applications and constrained by the capabilities of generic GIS has drawn to an end, and is being replaced
by a phase in which specific procedures are being proposed in order to implement more realistic models
of human perception of the landscape. Although research in this area is still characterized by the existence
of ‘schools’ - the ‘environmental’ school continues to explore refinements to approaches current
throughout the past decade while proponents of the ‘enrichment’ school advocate a landscape
architectural approach – it is becoming increasingly clear that deterministic analysis in GIS can become
more accurate by adopting a flexible approach to cognitive criteria. Early explorations of such a ‘cognitive
processualist’approach include Wheatley and Gillings’(2000) investigations in the framework of Higuchi
viewshed properties, Llobera’s (2000) implementation of ‘attractive’ and ‘repellent’ features in the
landscape, and several of the case studies presented elsewhere in this thesis.
The apparent, and vocal, conflict between adherents of processualist and post-processualist approaches to
archaeology was shown to be beside the point from a pragmatic point of view. A much more significant
watershed exists between studies that fail to adduce proper supporting evidence to their interpretations,
and those that do. A case in point is the general lack of supporting evidence given for claims of unusual
(non-random) cost or viewshed properties for particular locations within a region. Yet the techniques to
provide such evidence exist:
• To substantiate the significance of visibility and accessibility properties obtained for a sample of
archaeologically meaningful locations, they can be compared with similar properties of either one
large sample, or many similar-sized samples of randomly chosen locations. The former is typically
done by first generating a cumulative visibility or accessibility index for all, or a representative subset
of, locations within the study region. The result obtained for the sample of interest can then be
formally compared to the population (one-sample tests) or to a representative sample of it (twosample tests). As is shown by Fisher and others (1997) and Kvamme (1999), Monte Carlo tests can be
employed in the latter approach to demonstrate that the results obtained are unlikely to have arisen
by chance, now that computing power is no longer an issue.
• A different method by which LOSA- or CSA-based models may be supported is by comparison with
independent archaeological evidence. For example, networks of least-cost paths may be compared to
historically known networks such as the mule-paths that criss-crossed the Italian uplands until
recently.
• Finally, two potential approaches have been suggested to make cognitive models more robust: firstly,
since there are a large number of potential sources of error in the input parameters and algorithms
involved in these models, Wheatley and Gillings (2000) suggest that we should instead study the
trends emerging from an accumulation of models with a wide variety of such input parameters and
algorithms. Secondly, rather than attempting to interpret the viewshed or accessibility properties of
sites directly, we could study the differences among sites and between sites and ‘background’. Both
methods avoid a lot of the uncertainty and errors usually associated with this type of analysis.
Case studies from the Wroxeter hinterland, illustrating the application of line-of-sight and cost surface
analysis techniques in models of the Late Iron Age and Roman cognitive landscape, were presented in
chapter 16. Developed originally in 1996-7, these case studies no longer represent the ‘state of the art’in
GIS modeling of visibility and accessibility, as the modeling of ‘energy and resource landscapes’with the
help of cost surface techniques has become increasingly sophisticated in the last few years. Cost is now
measured in real terms as energy expenditure in Watt or kcal; cost surfaces have become anisotropic to
reflect the importance of the effect of direction of movement on costs; the interpretation of visibility and
accessibility models is now informed by a better understanding of statistical complexities; and, last but not
least, the limitations of the underlying theory are now beginning to be understood. Keeping these
limitations in mind, line-of-sight analysis is used to model and visualize the potential degree of control
exercised from Iron Age hillforts and the later Roman legionary fortress at Wroxeter over parts of
18 -
13
SUMMARY AND CONCLUSIONS
the central Severn valley. Cost surface analysis is used to model, first, the accessibility of the region in the
Roman period, and then, on the basis of that accessibility, the locations of potential routes and nodes in
the local infrastructure.
An exploration of the properties of ‘background’visibility and accessibility indices is also included in
chapter 16. Although the simulations presented there do not constitute definite proof, it would appear
that simply by increasing the viewshed range used, the higher visibility values will concentrate on areas of
higher ground, ridges and peaks. Any sufficiently large sample of archaeological viewpoints will tend to
generate a cumulative viewshed similar to these simulated ones, depending on the viewshed radius used.
Furthermore, any such viewshed based on points located on or near ridges and peaks will further
emphasize the visibility of other ridges and peaks. It is argued that the properties of cumulative
‘background’indices should become more widely known and used by practitioners. Of equal importance
is the testing of GIS-generated visibility and accessibility models, firstly by a comparison with extant
historic, archaeological, and cartographic data, and thereafter by targeted fieldwork.
CENTRAL PLACES AND TERRITORIAL DIVISIONS
In chapter 15, spatial models deriving from the theoretical models of centralization, urbanization, and
colonization presented in chapter 2 were investigated, using examples from the Pontine region and the
Sibaritide. These examples concerned centralization, urbanization, and the formation of territories during
protohistory, and early and middle Republican colonization of the Lepine margin.
Protohistoric settlement dynamics in widely separated regions of central and southern Italy, as presented
in the literature, demonstrate a remarkable similarity, so that it is quite possible to draw comparisons
between the regions on this basis. The development of indigenous central places and territories in the late
Bronze Age and early Iron Age is one such dynamic. The process by which archaeological sites of that
period are identified, the criteria by which they are classified, and the arguments and methods used to
segment the surrounding landscape into territories, are investigated and weaknesses in the process are
exposed. In general it emerges that current economic and cognitive models of the ordering of settlements
and the landscape in protohistory (and, for indigenous societies, even for many centuries afterwards) are
of a very non-specific and intuitive nature. Bias modeling (chapter 6) and corrective fieldwork (chapter 8)
will be needed to test the many assumptions on which these models are based. For the early (4th century
BC) Roman colonization of the Lepine margin, a viewshed study is conducted to test the hypothesis that
these colonies were established as strategic strongpoints, as much for the purpose of controlling the
Lepine uplands from which direction attacks by mountain tribes could be expected as for protecting the
agricultural resources and infrastructure of the Pontine plain. A brief review of literary historical sources
regarding Rome's early colonies in south Lazio is given to substantiate the need for such strategic decision
making in a landscape which remained contested between Latins, Volscans, and Romans for a century
and a half. The results of the study show that the colonial viewsheds cover the whole of the western
Lepine mountains in a complementary fashion, and support the hypothesis.
Specific conclusions regarding settlement patterns in the three study regions, drawn in chapter 13,
include:
• The developed Iron Age settlement pattern in the Sibaritide and Salento Murge displays remarkable
similarities in the geomorphological location and spacing of the settlements, located some 10-12 km
apart in defensible hilltop positions. It must be doubted that regular access to sea-born trade had a
major role to play in this, because the pattern continues into the Murge upland, and may in fact be
more strictly related to control over high quality agricultural and pastoral resources. This hypothesis
can be tested by targeted fieldwork in the Lepine uplands and the inland reaches of the Sibaritide.
• The ‘colonial’settlement pattern in southern Italy was centered on the coast rather than on the hill
country, and combined accessibility by sea with the presence of a substantial agricultural hinterland.
In contrast, Rome’s early colonies were as much or more intended to fulfill strategic functions, so
their locations meet other criteria of dominance – namely that of control over routes of attack
18 -
14
V A N L EUSEN : P A T T E R N T O P ROCESS
and advance. The fact that the viewsheds of the Roman colonies on the Lepine margin are both
complementary and fall within the Higuchi 'middle range' distance creates support for the idea that
these towns were located as much to control movement across the Lepine up- and highlands, as to
control and protect communications and agricultural resources in the Pontine plain.
• As a tool for archaeological spatial analysis of territories, Thiessen polygons have been used
extensively. The case study presented in section 2 of chapter 13 demonstrates the weaknesses of the
technique in specific archaeological situations. The use of GIS and cost surface analysis allows the
technique to be refined by replacing the simple gravity model of space with one in which each center
can have its own ‘weight’determining the relative size of its polygon, and in which characteristics
influencing the accessibility of the terrain are used to determine the location of territorial boundaries
instead of horizontal distance. Rank-size studies such as the one by Guidi (1985), although based on
unreliable settlement sizes, when combined with X-Tent modeling techniques provide a more
credible alternative to Thiessen polygons; another advantage is that they can be used to implement
central place models as well as peer polity models of society.
4 C O N C L U S I O N : R E G I O NAL ARCHAEOLOGICAL DA TA ANALYSIS
My research has shown that the type of regional archaeological data analysis required by landscape
archaeological approaches is an area where both theory and method are still in their infancy. High-level
theories about the occurrence, scope, and effects of processes such as centralization, urbanization, and
Hellenization/Romanization cannot yet be supported by middle range theory, which itself cannot be
developed until the basic business of generating information of sufficient quality about the archaeological
record has been tackled. Currently, archaeological data can be made to fit almost any interpretation
generated, ultimately, on the basis of the ancient written sources. If we are to escape from this selfreinforcing cycle, research should perhaps no longer be focused on the classical themes generated by
culture-historical approaches, but should seek its own proper field of operation.
In the area of methods and methodology, I have demonstrated the pervasive influence of systematic
research and visibility biases on the patterns that are present in the archaeological data generated over the
past 50 years or so. There are mechanisms at work, both in the traditional archaeological interpretation of
limited numbers of excavated sites and historical sources, and in the landscape archaeological approach,
that cause the systematic undervaluation of unobtrusive remains. The significance of systematic biases in
both the coarse site-based data sets resulting from desktop and ‘topographic’studies and the more
detailed site-based or ‘continuous’data resulting from intensive field surveys has become much clearer as
a result of the studies reported here. This should have practical consequences for the ways in which we
study the existing archaeological record, plan future landscape archaeological research, and conduct field
surveys. Site databases, the traditional starting point for regional archaeological studies, can no longer be
taken at face value; rather, they require careful source criticism before being used to support specific
arguments and hypotheses about settlement and land use dynamics. My studies have also shown that
future data collection, whether through field survey, excavation or other methods, has to take place in a
much more methodical manner if we are to produce data that are sufficiently standardized to be
successfully exchanged, compared, and interpreted by others – guidelines for which should become
embodied in an international standard defining ‘best practice in landscape archaeology’.
18 -
View publication stats
15
Martijn van Leusen