Late Bronze Age Social Landscapes of the Southeast Balkans: A spatial analysis of local ceramics and site distribution 9781407316819, 9781407355573

Table of contents :
Front Cover
Title Page
Copyright
Dedication
Acknowledgements
TABLE OF CONTENTS
List of figures
List of tables
List of plates
Abstract
1. Introduction and background
2. Change and identity: a theoretical perspective
3. Data and method
4. Archaeological overview and prior research
5. Site characterisation and issues of preservation
6. Quantifying site distributions
7. Characteristics and classification of the local LBA pottery
8. Spatial analysis of ceramic distributions
9. Regional patterning and a revised chronology
10. People and places in LBA Thrace
Conclusion
Bibliography
Appendix I. Plates

Citation preview

A spatial analysis of local ceramics and site distribution

Denitsa Nenova

B A R I N T E R NAT I O NA L S E R I E S 2 9 3 6

2019

E

Late Bronze Age Social Landscapes of the Southeast Balkans

AL N N LI IO N IT O D IAL AD ER AT

M

Denitsa Nenova has a PhD in Archaeology from University College London and is currently working on a number of archaeological projects in the Eastern Mediterranean and the Balkans. Her main professional interests and research areas include European Bronze Age, Landscape Archaeology, Survey Methodology, Remote Sensing and Computer Applications in Archaeology.

Late Bronze Age Social Landscapes of the Southeast Balkans

The book explores settlement and burial patterns across the southeastern corner of the Balkan Peninsula during the second millennium BC and offers a new, detailed cross-border examination of the local pottery data. The volume offers a comprehensive analysis based on the existing cultural-historical framework and calls into question established constructs such as the ‘Plovdiv-Zimnicea’ culture. The work offers a chronologically structured analysis of pottery sequences and is methodologically innovative in the way it applies a rare combination of settlement-scale analysis using advanced spatial-statistical methods alongside artefact-scale typological and stylistic study on local ceramics also subjected to spatial-statistical mapping. As a result, the research highlights clusters of attributes and cycles of micro-regional interaction. On that basis it also addresses the formation, development and decline of the Late Bronze Age tradition(s) in Thrace and examines the degree to which this trajectory was influenced by wider patterns of regional development.

NENOVA

‘The aggregation of data that crosses political and geographic boundaries in SE Europe is new and highly valuable.’ Dr. Adela Sobotkova, Macquarie University

2019

‘Denitsa Nenova’s work covers a significant gap in Thracian archaeology. This is a synthetic study of the Late Bronze Age in Thrace that explores several aspects of the regional material culture. The study is theoretically well embedded and may become a reference point in regional archaeology.’ Dr. Stefanos Gimatzidis, Austrian Archaeological Institute

BAR S2936

BAR INTERNATIONA L SE RIE S 2936

Late Bronze Age Social Landscapes of the Southeast Balkans A spatial analysis of local ceramics and site distribution

Denitsa Nenova

B A R I N T E R NAT I O NA L S E R I E S 2 9 3 6

2019

Published in 2019 by BAR Publishing, Oxford BAR International Series 2936 Late Bronze Age Social Landscapes of the Southeast Balkans ISBN 978 1 4073 1681 9 paperback ISBN 978 1 4073 5557 3 e-format DOI https://doi.org/10.30861/9781407316819 A catalogue record for this book is available from the British Library

© Denitsa Nenova 2019 Cover Image Pots do not equal people but they often travel with them. The Author’s moral rights under the 1988 UK Copyright, Designs and Patents Act are hereby expressly asserted. All rights reserved. No part of this work may be copied, reproduced, stored, sold, distributed, scanned, saved in any form of digital format or transmitted in any form digitally, without the written permission of the Publisher.

BAR titles are available from: Email Phone Fax

BAR Publishing 122 Banbury Rd, Oxford, ox2 7bp, uk [email protected] +44 (0)1865 310431 +44 (0)1865 316916 www.barpublishing.com

To the countless unrecognized workers and students whose labour generates archaeological data and whose names remain unknown.

Acknowledgements This book is based on my PhD research completed at the Institute of Archaeology, University College London (UCL) and partially funded by the British Academy and the British School at Athens (BSA). I especially thank my supervisors Andrew Bevan, Todd Whitelaw and Cyprian Broodbank, as well as Evangelia Kiriatzi for their guidance, comments and support and specifically Andrew Bevan for commenting on the draft of this book and for his help with a number of analyses. I express my gratitude also to Georgy Nekhrizov, Stelios Andreou and the Department of Archaeology at the Aristotle University of Thessaloniki, Mehmet Özdoğan, and Catherine Morgan and the British School at Athens for granting me access to materials and their publication. I also thank Thilo Rehren, Nadezhda Kecheva, Nataliya Ivanova, Estelle Strazdins, Hans and Diny Geerts, and my family for facilitating my research in different capacities and throughout different stages. Three anonymous reviewers provided constructive comments that encouraged me to restructure the manuscript and to refine some of my ideas. Charles Sturge, with his much-needed, rapidly yet impeccably made corrections, played a significant role in the final form of the text. The wealth of errors that remain are my very own. Last but not least, I want to express a number of positive feelings to Hüseyin Çınar Öztürk, who influenced several of my ideas throughout the years and offered his comments on the earlier draft, but most importantly who has always been there for me.

Contents List of figures.................................................................................................................................................................... vii List of tables...................................................................................................................................................................... xi List of plates..................................................................................................................................................................... xii Abstract............................................................................................................................................................................ xiii 1. Introduction and background........................................................................................................................................1 Introduction......................................................................................................................................................................1 Background......................................................................................................................................................................2 Geography........................................................................................................................................................................3 Topography..................................................................................................................................................................3 Geology and natural resources....................................................................................................................................5 Climate........................................................................................................................................................................6 Soils and vegetation....................................................................................................................................................7 Natural routes..............................................................................................................................................................7 2. Change and identity: a theoretical perspective..........................................................................................................11 Culture-historical archaeology.......................................................................................................................................15 Pots as politics ...............................................................................................................................................................17 Style and social boundaries............................................................................................................................................18 Physical environments and human landscapes..............................................................................................................19 3. Data and method...........................................................................................................................................................21 Analysis of pottery data.................................................................................................................................................21 Site analysis....................................................................................................................................................................23 Site definition and exploratory site-based analyses.......................................................................................................23 Statistical location modelling.........................................................................................................................................24 4. Archaeological overview and prior research..............................................................................................................27 Archaeology is born.......................................................................................................................................................27 Archaeology after the Second World War......................................................................................................................29 The ‘Archaeological Enlightenment’ ............................................................................................................................30 Archaeology in the twenty-first century.........................................................................................................................32 Problems of interpretation..............................................................................................................................................33 5. Site characterisation and issues of preservation ......................................................................................................35 Archaeological recovery ...............................................................................................................................................35 Site classification............................................................................................................................................................36 Settlements.....................................................................................................................................................................36 Sanctuaries.....................................................................................................................................................................38 Other categories.............................................................................................................................................................38 Site distribution..............................................................................................................................................................39 6. Quantifying site distributions......................................................................................................................................45 Selection of zones and parameters ................................................................................................................................46 Locational modelling.....................................................................................................................................................50 Zone one: the eastern part of the Upper Thracian Plain.................................................................................................50 Zone two: the entire Upper Thracian Plain....................................................................................................................52 Zone three: Rhodope Mountains....................................................................................................................................53 Zone four: western Rhodopes........................................................................................................................................55 Zone five: eastern Rhodopes..........................................................................................................................................55 Zone six: eastern Rhodopes and Sakar..........................................................................................................................56 Zone seven: western Thrace the northern Aegean hinterland........................................................................................57 v

Late Bronze Age Social Landscapes of the Southeast Balkans Zone eight: upper and middle Struma Valley.................................................................................................................58 Zone nine: Central Macedonia and Chalkidki...............................................................................................................59 Cemeteries......................................................................................................................................................................60 Multivariate categorical analysis: Eastern Rhodopes....................................................................................................60 Settlements.....................................................................................................................................................................60 Sanctuaries.....................................................................................................................................................................61 A multinomial model as an alternative..........................................................................................................................61 7. Characteristics and classification of the local LBA pottery......................................................................................63 Function and technology ...............................................................................................................................................63 Coarse ware....................................................................................................................................................................63 Fine ware........................................................................................................................................................................64 Technological features ..................................................................................................................................................65 Classification..................................................................................................................................................................68 Kantharoi (K).................................................................................................................................................................68 Bowls (B).......................................................................................................................................................................71 Kylikes (KY)..................................................................................................................................................................75 Amphorae (A)................................................................................................................................................................77 Jugs (J)...........................................................................................................................................................................81 Cups (C).........................................................................................................................................................................83 Double vessels (D).........................................................................................................................................................85 8. Spatial analysis of ceramic distributions....................................................................................................................87 LBA pottery complexes ...............................................................................................................................................87 Pottery distributions revisited. Spatial analysis of diagnostic ceramic features ..........................................................89 9. Regional patterning and a revised chronology...........................................................................................................99 Traditional chronological framework ...........................................................................................................................99 10. People and places in LBA Thrace............................................................................................................................ 111 Sub-regional patterning of ceramic and site distribution ...........................................................................................111 Area one. Eastern Rhodopes and Sakar .....................................................................................................................111 Area two. Western Rhodopes.......................................................................................................................................113 Area three. Upper Thracian Plain.................................................................................................................................114 Area four. Struma Valley..............................................................................................................................................116 Area five. Western Thrace and the northern Aegean hinterland...................................................................................117 Comparanda.................................................................................................................................................................119 The appearance of LBA incised ware in the southeast Balkans...................................................................................120 The origins of the LBA in the southeast Balkans.........................................................................................................122 Conclusion.......................................................................................................................................................................127 Bibliography....................................................................................................................................................................129 Appendix I.......................................................................................................................................................................157 Plates................................................................................................................................................................................157 Site and pottery data are available to download as Excel files from: www.barpublishing.com/additional-downloads.html

vi

List of figures Figure 1.1. Map of the study area ........................................................................................................................................2 Figure 1.2. Map of the area with main places mentioned in the text ...................................................................................4 Figure 1.3. Map of the region indicating natural routes following main river arteries and mountain passes, associated with present-day locations ..................................................................................................................................8 Figure 4.1. Map of the area with main places mentioned in the text .................................................................................28 Figure 5.1. Map of the area showing the main places mentioned in the text.....................................................................36 Figure 5.2. Distribution of LBA sites in western and northern Thrace, and Macedonia in comparison to the distribution of sites in eastern Thrace ................................................................................................................................37 Figure 5.3. Distribution of sanctuaries across the study area ............................................................................................39 Figure 5.4. Levels of archaeological investigation across the study area ..........................................................................40 Figure 5.5. Distribution of tell sites across the study area .................................................................................................41 Figure 5.6. Distribution of sites with stone architecture across the study area ..................................................................42 Figure 5.7. Distribution of burial sites across the study area .............................................................................................43 Figure 6.1. Map of the area showing the main places mentioned in the text.....................................................................45 Figure 6.2. Map of the study area showing the extents of the proposed zones .................................................................46 Figure 6.3. Raster maps representing the twelve selected covariates: elevation, slope, aspect, distance to large rivers, distance to small- and medium-sized water sources, distance to marble, distance to limestone, distance to conglomerate, and distance to alluvium ........................................................................................................................47 Figure 6.4. Histograms of independent variables showing raw and log-transformed distributions ..................................49 Figure 6.5. Cook’s distance plots of raw data of zone one ................................................................................................50 Figure 6.6. Cook’s distance plots log-transformed data of zone one .................................................................................50 Figure 6.7. Q-Q plot the raw data of zone one...................................................................................................................50 Figure 6.8. Q-Q plots of the log-transformed data.............................................................................................................50 Figure 6.9. Histogram of the site and non-site distribution across an elevation range covering the entire Rhodope Mountains ...........................................................................................................................................................54 Figure 6.10. Histogram of the site and non-site distribution across a log-transformed elevation range covering the entire Rhodope Mountains ...........................................................................................................................................56 Figure 6.11. Two-class geological map of the eastern Rhodopes showing the locations of sites identified as cemeteries ....60 Figure 6.12. Two-class geological map of the eastern Rhodopes showing the locations of sites identified as sanctuaries ..........................................................................................................................................................................61 Figure 6.13. Multinomial regression juxtaposing settlements, cemeteries and sanctuaries from the eastern Rhodopes against elevation and distance to water .............................................................................................................62 Figure 7.1. Map of the area showing some of the main places mentioned in the text .......................................................64 Figure 7.2. Distribution of vessel shapes across the main technological pottery groups and surface treatment types rate across the dataset ...............................................................................................................................................65 Figure 7.3. Distribution of inclusions rate across the main technological pottery groups; comparative chart used for estimating the frequency and the size of inclusions in ceramics..................................................................................65 Figure 7.4. Distribution of surface treatment types across the main technological pottery groups and surface treatment types across the main colour groups ..................................................................................................................66 vii

Late Bronze Age Social Landscapes of the Southeast Balkans Figure 7.5. Incised decoration: Koprivlen, Kastanas, Toumba Thessalonikis, Toumba Thessalonikis, Kastanas, Toumba Thessalonikis, Toumba Thessalonikis, and Nikisiani ..........................................................................................67 Figure 7.6. Furchenstich decoration: Ada Tepe, Ada Tepe, Dolno Lukovo, Pastrook, Pelevun, Dambala, Mishevsko, Zhelezino, Vishegrad ......................................................................................................................................68 Figure 7.7. Kantharos. Type K1: Kastanas ........................................................................................................................69 Figure 7.8. Kantharos. Type K1.1: Sandanski ...................................................................................................................69 Figure 7.9. Kantharos. Type K2: Razkopanitsa .................................................................................................................69 Figure 7.10. Kantharos. Type K2.1: Sandanski .................................................................................................................70 Figure 7.11. Kantharos. Type K3: Plovdiv ........................................................................................................................70 Figure 7.12. Kantharos. Type K3.1: Plovdiv .....................................................................................................................70 Figure 7.13. Kantharos. Type K3.2: Ada Tepe ...................................................................................................................70 Figure 7.14. Kantharos. Type K4: Lilovo ..........................................................................................................................71 Figure 7.15. Kantharos. Type K4.1: Stomantsi ..................................................................................................................71 Figure 7.16. Kantharos. Type K5: Upper Thrace ...............................................................................................................71 Figure 7.17. Kantharos. Type K6: Upper Thrace ...............................................................................................................71 Figure 7.18. Bowl. Type B1: Sandanski ............................................................................................................................72 Figure 7.19. Bowl. Type B2: Chokoba 18A.......................................................................................................................72 Figure 7.20. Bowl. Type B3: Mylopotamos.......................................................................................................................72 Figure 7.21. Bowl. Type B3.1: Toumba Thessalonikis ......................................................................................................72 Figure 7.22. Bowl. Type B3.2: Stathmos Aggista ..............................................................................................................73 Figure 7.23. Bowl. Type B4: Toumba Thessalonikis .........................................................................................................73 Figure 7.24. Bowl. Type B4.1: Chokoba 18 ......................................................................................................................73 Figure 7.25. Bowl. Type B4.2: Bikovo ..............................................................................................................................73 Figure 7.26. Bowl. Type B4.3: Vratitsa .............................................................................................................................73 Figure 7.27. Bowl. Type B5: Chokoba 18A.......................................................................................................................74 Figure 7.28. Bowl. Type B6: Ada Tepe ..............................................................................................................................74 Figure 7.29. Bowl. Type B7: Ada Tepe ..............................................................................................................................74 Figure 7.30. Bowl. Type B8: Pastrook ...............................................................................................................................74 Figure 7.31. Bowl. Type B8.1: Ada Tepe ...........................................................................................................................74 Figure 7.32. Bowl. Type B8.2: Ada Tepe ...........................................................................................................................74 Figure 7.33. Kylix. Type KY1: Exohi ................................................................................................................................75 Figure 7.34. Kylix. Type KY2: Kamenska Chuka .............................................................................................................75 Figure 7.35. Kylix. Type KY3: Borino ..............................................................................................................................75 Figure 7.36. Kylix. Type KY4: Exoh .................................................................................................................................75 Figure 7.37. Kylix. Type KY5: Stathmos Aggista .............................................................................................................76 Figure 7.38. Kylix. Type KY6: Ada Tepe ..........................................................................................................................76 Figure 7.39. Kylix. Type KY6.1: Asprovrysi .....................................................................................................................76 Figure 7.40. Kylix. Type KY6.2: Dolno Lukovo ...............................................................................................................76 Figure 7.41. Kylix. Type KY7: Borino ..............................................................................................................................77 Figure 7.42. Kylix. Type KY7.1: Faia Petra ......................................................................................................................77 Figure 7.43. Kylix. Type KY8: Photolyvos .......................................................................................................................77 viii

List of figures Figure 7.44. Amphora. Type A1: Lilovo ............................................................................................................................78 Figure 7.45. Amphora. Type A1.1: Potamoi ......................................................................................................................78 Figure 7.46. Amphora. Type A1.2: Toumba Thessalonikis................................................................................................79 Figure 7.47. Amphora. Type A2: Potamoi .........................................................................................................................79 Figure 7.48. Amphora. Type A2.1: Borino ........................................................................................................................79 Figure 7.49. Amphora. Type A3: Plovdiv ..........................................................................................................................79 Figure 7.50. Amphora. Type A4: Kastanas ........................................................................................................................80 Figure 7.51. Amphora. Type A4.1: Kamenska Chuka .......................................................................................................80 Figure 7.52. Amphora. Type A5: Plovdiv ..........................................................................................................................80 Figure 7.53. Amphora. Type A6: Sandanski ......................................................................................................................81 Figure 7.54. Amphora. Type A6.1: Ada Tepe ....................................................................................................................81 Figure 7.55. Jug. Type J1: Toumba Thessalonikis .............................................................................................................81 Figure 7.56. Jug. Type J1.1: Kamenska Chuka ..................................................................................................................82 Figure 7.57. Jug. Type J2: Plovdiv.....................................................................................................................................82 Figure 7.58. Jug. Type J3: Nova Zagora ............................................................................................................................82 Figure 7.59. Jug. Type J3.1: Plovdiv..................................................................................................................................82 Figure 7.60. Jug: Lilovo .....................................................................................................................................................83 Figure 7.61. Cup. C1: Izvor ...............................................................................................................................................83 Figure 7.62. Cup. C1.1: Vishegrad.....................................................................................................................................83 Figure 7.63. Cup. C2: Chokoba 18A..................................................................................................................................84 Figure 7.64. Cup. C3: Ada Tepe.........................................................................................................................................84 Figure 7.65. Cup. C4: Tatul................................................................................................................................................84 Figure 7.66. Cup. C5: Chokoba 18 ....................................................................................................................................84 Figure 7.67. Cup. C6: Chokoba 18 ....................................................................................................................................84 Figure 7.68. Cup. C6.1: Chokoba 18A...............................................................................................................................84 Figure 7.69. Cup. C6.2: Dolno Lukovo .............................................................................................................................85 Figure 7.70. Examples of double-vessels from LBA contexts in Thrace ..........................................................................85 Figure 7.71. Hypothetical reconstruction of the functionality of the LBA double-vessels ...............................................86 Figure 8.1. Map of the area showing some of the main places mentioned in the text .......................................................88 Figure 8.2. Relative risk surface of amphora types A1, A2, A3, A4, and A6 ....................................................................90 Figure 8.3. Relative risk surface of kantharos types K1, K2, K3, and K4.........................................................................91 Figure 8.4. Relative risk surface of bowl types B1, B3,B4, and B8.1 ...............................................................................92 Figure 8.5. Relative risk surface of jug types J1, J2, and J3 ..............................................................................................93 Figure 8.6. Relative risk surface of kylix types KY2, KY3, KY5, and KY6 ....................................................................94 Figure 8.7. Relative risk surface of furchenstich decoration, and incised decoration .......................................................95 Figure 8.8. Arbitrary regions defined from unsupervised clustering combining the relative risk surfaces of all pottery types .......................................................................................................................................................................96 Figure 8.9. Comparison among 10, 100, 1000, and 10000 decision trees in the random forest algorithm set at five kernels .........................................................................................................................................................................96 Figure 8.10. Comparison among 100, 500, 1000, and 1000000 iterations allowed in the k-means at five kernels ..........97

ix

Late Bronze Age Social Landscapes of the Southeast Balkans Figure 8.11. Arbitrary regions defined from the unsupervised clustering presented in this chapter combining the relative risk surfaces of all pottery types ...........................................................................................................................97 Figure 8.12. Arbitrary regions defined from the unsupervised clustering combining relative risk surfaces of all sites by type........................................................................................................................................................................98 Figure 9.1. Map of the area with the main places mentioned in the text .........................................................................100 Figure 9.2. Model of the second millennium radiocarbon dates from the study area collected so far ............................104 Figure 9.3. Summed probability distribution of a larger sample covering calibrated BCAD dates from the beginning of the Early Bronze Age until the end of the Iron Age: raw data distribution overlaid with normalised curve and normalised SPD overlaid with a bin density distribution ................................................................................105 Figure 9.4. Summed probability distribution of a larger sample covering calibrated BCAD dates from the beginning of the EBA until the end of the Iron Age indicating a potential drop in the values around the beginning of the sixteenth century BC.............................................................................................................................105 Figure 9.5. Hypothesis testing of exponential SPD with a 95% confidence interval addressing the null model of a larger sample covering calibrated BCAD dates from the beginning of the EBA until the end of the Iron Age ...........106 Figure 9.6. Comparative distribution of dated sites across the study area combining calibrated radiocarbon dates and cultural synchronisations .................................................................................................................................108 Figure 9.7. Periodical distribution between different sub-regions in the study area compared with established chronological phases from central and northern Greece..................................................................................................109 Figure 10.1. Map of the area showing the main places mentioned in the text.................................................................112 Figure 10.2. Comparative chart of vessel types from the entire study area juxtaposing EBA 3, MBA 1, MBA 2, and LBA vessel types .......................................................................................................................................................115 Figure 10.3. Pottery illustration comparing assemblages of shapes typical for western Rhodopes, eastern Rhodopes, central Macedonia, western Thrace and the northern Aegean during the end of sixteenth – thirteenth century BC .......................................................................................................................................................................117 Figure 10.4. Pottery illustration comparing assemblages of shapes typical for the Upper Thracian Plain in LBA1, the Upper Thracian Plain in LBA2, and Struma Valley (end of fourteenth – twelfth century BC) .....................118 Figure 10.5. Pottery illustration comparing mixed assemblages of shapes typical for LBA 1 and LBA 2 .....................121 Figure 10.6. Map of chronologically defined micro-areas based on pottery variation ....................................................124 Figure 10.7. Map of the study area showing two potential border zones identified along the Struma Valley to the west and the Ergene/lower Maritsa Basin in eastern Thrace to the southeast............................................................125

x

List of tables Table 6.1. Analytical zones used in the location modelling ...............................................................................................46 Table 6.2. A list of variables used in the multivariate regression models presented in this chapter ..................................48 Table 6.3. Non-constant variance score test of zone one ...................................................................................................50 Table 6.4. Variance inflation factor test for multicollinearity applied to the data of zone one ..........................................51 Table 6.5. Linear model of the raw data of zone one .........................................................................................................51 Table 6.6. Linear model of the log-transformed data of zone one .....................................................................................51 Table 6.7. Stepwise Akaike model of the raw data of zone one ........................................................................................51 Table 6.8. Stepwise Akaike model of the log-transformed data of zone one.....................................................................51 Table 6.9. ANCOVA model of the raw data of zone one ...................................................................................................51 Table 6.10. ANCOVA model of the log-transformed data of zone one .............................................................................52 Table 6.11. Model comparison of zone one .......................................................................................................................52 Table 6.12. Regression values and confidence intervals of the covariates in zone one .....................................................52 Table 6.13. Zone two, comparison of models ....................................................................................................................52 Table 6.14. Zone three, comparison of models ..................................................................................................................54 Table 6.15. Zone four, comparison of models ...................................................................................................................55 Table 6.16. Zone five, comparison of models ....................................................................................................................56 Table 6.17. Zone six, comparison of models .....................................................................................................................57 Table 6.18. Zone seven, comparison of models .................................................................................................................58 Table 6.19. Zone eight, comparison of models, based on sample 1 with 41 random points .............................................58 Table 6.20. Zone eight, comparison of models based on sample 2 with 500 random points ............................................58 Table 6.21. Zone nine, comparison of models ...................................................................................................................59 Table 6.22. Binomial model of settlement type locations from the eastern Rhodopes ......................................................60 Table 6.23. Probability values of the predictor from the binomial model of cemetery type locations from the eastern Rhodopes ...............................................................................................................................................................61 Table 6.24. Probability values of the predictor from the binomial model of sanctuary type locations from the eastern Rhodopes ...............................................................................................................................................................61 Table 9.1. Traditional chronology of the LBA in Thrace .................................................................................................101 Table 9.2. Radiocarbon samples from the study area, calibrated using IntCal13 (OxCal 4.3) ........................................103

xi

List of plates Plate 1. Kantharos. Type K1 ............................................................................................................................................158 Plate 2. Kantharos. Type K1.1 .........................................................................................................................................159 Plate 3. Kantharos. Types K2 and K2.1 ...........................................................................................................................159 Plate 4. Kantharos. Types K3, K3.1, and K3.2 ................................................................................................................160 Plate 5. Kantharos. Types K4 and K4.1 ...........................................................................................................................161 Plate 6. Kantharos. Types K5 and K6 ..............................................................................................................................162 Plate 7. Bowl. Types B1 and B2 ......................................................................................................................................163 Plate 8. Bowl. Types B3 and B3.1 ...................................................................................................................................164 Plate 9. Bowl. Type B3.2 .................................................................................................................................................164 Plate 10. Bowl. Types B4 and B4.1 .................................................................................................................................165 Plate 11. Bowl. Type B4.2 ...............................................................................................................................................166 Plate 12. Bowl. Type B5 ..................................................................................................................................................166 Plate 13. Bowl. Types B6, B6.1, and B7 ..........................................................................................................................167 Plate 14. Bowl. Types B8, B8.1, and B8.2 .......................................................................................................................167 Plate 15. Kylix. Types KY1 and KY2 ..............................................................................................................................168 Plate 16. Kylix. Types KY3, KY4, and KY5 ...................................................................................................................169 Plate 17. Kylix. Types KY6 and KY6.1 ...........................................................................................................................170 Plate 18. Kylix. Types KY6.2, KY7, KY7.1, and KY8 ...................................................................................................170 Plate 19. Amphora. Types A1 and A1.1 ...........................................................................................................................171 Plate 20. Amphora. Types A1.1 and A1.2 ........................................................................................................................172 Plate 21. Amphora. Types A2 and A3 ..............................................................................................................................172 Plate 22. Amphora. Type A4 ............................................................................................................................................173 Plate 23. Amphora. Types A6, A5, and A6.1 ...................................................................................................................174 Plate 24. Jug. Type J1.......................................................................................................................................................175 Plate 25. Jug. Types J1.1, J2, J3, and J3.1........................................................................................................................175 Plate 26. Cup. Types C1, C1.1, C2, and C3 .....................................................................................................................176 Plate 27. Cup. Type C4 ....................................................................................................................................................177 Plate 28. Cup. Type C5, C6, C6.1, and C6.2....................................................................................................................177

xii

Abstract In a period when complex systems of Bronze Age social life and urban communities were wellestablished in the Aegean and Anatolia, the southeast corner of the Balkans, commonly known as Thrace, appears to have followed a different trajectory. Nevertheless, the presence of artefactual parallels between Thrace and its neighbours during the second millennium BC has provoked scholars to propose various forms of cultural interaction. This possible connection has diverted the research focus in Thrace from a thorough examination of local characteristics, which has limited our understanding of internal social patterns of development, a problem further exacerbated by an overall lack of systematic study. Moreover, the contemporary division of the area between Bulgaria, Greece, and Turkey and repeated adjustments to the borders throughout the twentieth century has further constrained synthetic archaeological investigation. Bearing in mind that modern political divisions do not necessarily coincide with any prehistoric social pattern, this research approaches the Late Bronze Age (LBA) in Thrace from an alternative contextual and geographic perspective. More than a hundred LBA sites have now been identified in contemporary Bulgaria. There is also a wealth of complementary information from northeastern Greece but, in contrast, a lack of evidence from European Turkey. In this context, a new, detailed cross-border examination of local ceramics calls into question existing culture-historical constructs such as the ‘PlovdivZimnicea’ horizon and offers a chronologically structured analysis of local pottery sequences. The classification tree employed in this study is based on a set of stylistic, morphological, and technological characteristics, which is then subjected to relative risk analysis of spatially-sensitive types. As a next step, an unsupervised clustering technique using a Random Forest algorithm was applied to a stack of relative-risk surfaces of all the pottery types identified for this study, in order to define the extent to which different ceramic characteristics are spatially correlated and whether their combined distribution forms clusters of preferred pottery types. Complementary multivariate spatial analysis of site distribution was applied to reveal possible settlement patterns with different micro-regional characteristics. This second spatial analytical component, consisting of site locational modelling, is utilized to examine the role that environmental factors played in local site distribution patterns and the relative density of sites across different potential micro-areas. Both analytical processes were supplemented by recently accumulated radiocarbon evidence, which facilitated the tracking of the appearance and the origin of the LBA incised ware in the region and the discussion on the provenance of other local pottery traditions. The overall outcome of the study suggests some spatiotemporal discrepancies in material culture across the study area. Chronologically, these are expressed in at least two phases of the LBA and one that can be attributed to the end of the Middle Bronze Age (MBA), which are not evenly distributed in the region. Besides spatial variability in the pottery traditions, this research suggests various levels of population mobility throughout the landscape along with models of ‘shifting agriculture’. Overall, this project highlights clusters of attributes and cycles of micro-regional interaction and independence at a strategic location of potential Eurasian articulation such as the southeast Balkans. Finally, a general pattern of isolation in the final phases of the Bronze Age can be rethought by considering a combination of both persistent vectors of long-range contact and also a high degree of local cultural diversity.

xiii

1 Introduction and background

Introduction

Age tends to describe the area under study as an independent phenomenon connecting central Europe with the Aegean (Leshtakov 2006, 145). Some studies have also applied the term ‘periphery’ as relevant to the southeast Balkans in relation to a South Aegean ‘core’ during the Late Bronze Age (Bozhinova 2008, 55; Leshtakov 2006, 144; Popov Jockenhövel 2011, 280). Large sections of the northern Aegean coast and its hinterland, however, show little sign of adopting the same kinds of organisation and lifestyle and instead almost certainly developed social structures that emphasised smaller-scale and potentially more socially egalitarian principles.

The archaeology of the Late Bronze Age (LBA) in the southeast corner of the Balkan peninsula, commonly known as Thrace, presents some particular challenges. Standing not only between Late Bronze Age southern Aegean and Anatolian states but also central and eastern European societies from the second half of the second millennium BC, this crucial area is poorly understood. It appears that at the junction where Asia meets Europe, certain conditions and specific kinds of evidence have combined to discourage fully integrated research. The essence of the problem lies in more recent dynamics: the area has been floating between several countries, Bulgaria, Greece, and Turkey, for the entire history of Archaeology as an academic discipline (Fig. 1.1). Three different national archaeologies have consequently had a considerable impact on the study of the Bronze Age in the region. Not only have there been practical difficulties, but it is also worth emphasising that current geopolitical divisions rarely coincide with any prehistoric social pattern.

One of the significant problems in approaching this topic lies in the assumption of a general cultural unity in the area from the northern Aegean coast to the Danube river and perhaps beyond. While it is true that such unity sometimes existed in this part of the Balkans, more evident in certain periods and less apparent in others, this hypothesis has not been tested with regard to the second millennium BC. The construction of cumulative concepts such as the ‘Zimnicea-Plovdiv’ group, which was assigned to the entire Bulgarian territory, is indeed outdated and needs consideration in the light of more recent research. The aforementioned archaeological construct is likely provoked by the lack of a neat chronological sequence caused by the insufficiency of scientific dates. Not only that of the LBA, but the chronology of the entire Bronze Age in the area is problematic. Recently obtained evidence allows us to update and reconsider the general chronological scheme that might have an impact on our understanding of the dynamics under consideration and the appearance of changes.

Archaeologists working in the area have often acknowledged the need for cross-border research, but such a venture has become possible only now in the twenty-first century, building upon recent advances. The available information was insufficient for most of the twentieth century and thus has resulted in limited foreign research in this problematic area, mainly by those seeking links between the Aegean and central Europe. The possibility for local research co-operation was obstructed by the complex political conditions in the region and by the strong language barrier dividing the three countries. Distribution maps often stopped at contemporary national borders, assuming a ‘different’ archaeology on the other side. Nevertheless, local researchers often pursued direct connections between Thrace and the southern Aegean, where, by the later part of the second millennium BC, complex political and economic systems had emerged alongside the growth of urban centres. The Upper Thracian Plain, structurally open to the Aegean through the major river valleys and the passes of the lower eastern part of the Rhodope Mountains, inspired some scholars to define the area as belonging culturally to the Mediterranean (Katincharov 1974; Treuil 1983, 2, 15-18). Others link Thrace and Macedonia with the Troad and defend the cultural, ethnic, and linguistic unity of the population (Georgiev 1987,127). One approach to the local Bronze

After underlining the pitfalls associated with characterising the LBA in the southeastern part of the Balkans, this study is organised in a way that can address some fundamental questions. First and foremost, I will present and discuss some of the principal cultural characteristics existing in the study region during the LBA, especially concerning pottery production, settlement patterns and burial practices. It is essential to understand whether we can treat the record as originating from a culturally homogenous system and if not to assess the level and the extent of the heterogeneity that is visible in the study area. The chronology is a central factor in the attempt to understand the dynamics in the region. That said, it has been nearly impossible to attempt a diachronic study. However, due to some recently available evidence, I will approach the 1

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 1.1. Map of the study area.

the north can be observed by the second half of the fifth millennium BC. Although somewhat developed socioeconomic environments, both of the latter were different from Anatolian and Near Eastern social structures with no signs of urbanisation in the Near Eastern sense (Özdoğan 2004, 392). During that period, the eastern part of Thrace, that bridge of cultures and continents, was already left out of both cultural horizons, settled only sporadically with small non-tell settlements, while the population inhabiting the rest of the Balkans were building settlement mounds with a continuous long-term occupation.

study with a refined chronological dataset, aiming to trace the formation, development and decline of the local LBA tradition(s). Based on all of the above, I will then address the nature and the extent of regional and interregional networks of social and economic interaction across the region. Background Today it is broadly accepted that by the end of the seventh millennium BC, a large portion of southeast Europe was already populated by Neolithic farmers, while it is evident that for a few preceding centuries an earlier Neolithic ‘Balkan-Anatolian block’ was in the process of formation (Hoddinott 1981, 15; Todorova 1986; 1995). As such, the southeast corner of the Balkans was playing a critical part as a bridge for transmitting Neolithic life out from Anatolia and into Europe.

The situation in the southeast Balkans during the period between the end of the fourth millennium and the end of the second millennium BC is puzzling in terms of the geographical extent of different cultural groups as well as their identity and sociopolitical relations. Some studies of Early Bronze Age (EBA) ethnocultural affiliations suggest a persuasive connection between northwestern Anatolian societies and what is referred to as Thrace through intermittent similarities in pottery and metal assemblages as well as local architectural traditions. However, while on the Anatolian side urban life flourished, in the Balkans several new local, as well as intrusive, cultural elements can be recognised throughout the entire EBA, including examples assigned to the Yamnaya and the Corded Ware

The development of Chalcolithic societies marks an important later episode of Thracian prehistory (see Gimbutas 1956; 1974; Renfrew 1972). By the beginning of the Late Chalcolithic, in the Near East, the prerequisites for increased urbanisation were already present (Özdoğan 2004, 392). Simultaneously, in the Balkans, the formation of the Kodzhadermen-Gumelnitsa-Karanovo VI (KGK) complex in the south and Cucuteni-Tripolye complex in 2

Introduction and background complexes (Panayotov 1989). In the south, the Ezero, Mihalich, Sveti Kirilovo and Yunatsite cultural groups have been defined based on mostly particularly distinctive pottery styles. A certain level of social and commercial interaction seems to have existed through communication routes within the southern Balkans and northwestern Anatolia during the EBA, but that network nearly vanished at the end of the EBA. What is known about the centuries corresponding to the Middle Bronze Age (MBA) in the southern Aegean and Anatolia is very little and represents a significant lacuna in the archaeological record and the interpretation of the regional Bronze Age dynamics. Similarly, the beginning of the LBA and the definition of the LBA are questions still challenging many researchers working in the area. In general, the characterisation of the entire Bronze Age period in this crucial area is a problem confronting any prehistorian dealing with European archaeology.

reason, it has been traditionally considered a critical zone and a cultural bridge transmitting people, ideas and things (Özdoğan 2004, 390). Leaving aside the historical baggage a term such as Thrace usually carries, I will employ it here when appropriate but avoid overly detailed geographic definitions throughout the study. Generally, the extent of the area engaged in the current research spans the Balkan Mountains to the north, the Aegean to the south, the Black Sea and the Marmara Sea to the east and the valley of Strymon River to the west. Within these limits, the area can be abruptly divided into several sub-regions: 1) Sub-Balkan Fields1, 2) Upper Thracian Plain2, 3) Sakar Mountains, 4) Strandzha (tr. Istranca) Mountains, 5) southwest Black Sea coast, 6) Tekirdağ Uplands 7) Rhodope Mountains, 8) Lower Maritsa (gr. Έβρος, tr. Meriç) Valley and Ergene Plain, 9) western Thrace and the northern Aegean hinterland, and 10) Struma (gr. Στρυμόνας) Valley. Such segmentation has often predisposed different living conditions, caused by a combination of features such as drainage and water balance, soils, vegetation, raw materials, natural routes or other environmental factors.

Besides these strictly archaeological challenges, the recent geopolitical environment in the region has only obstructed further any thorough research. The traditional network mechanisms of the twentieth-century archaeology have entrapped most research agendas strictly within the borders of existing nation-states. It is often the case that topographically defined units have been cut across to serve the needs of modern political agendas. The borders on any archaeological map appear as political as the current state borders are, limiting possible research syntheses. Although scholars have often acknowledged that nationstates’ territories were not necessarily the best analytical or interpretative units when it comes to the study of past societies, such a bias was unavoidable, and its importance was often ignored; the practical and ideological difficulties involved to overcome it was more than the potential advantages.

Topography The northernmost zone of Thrace, traditionally referred to as the ‘Sub-Balkan Fields’ is geomorphologically enclosed by the large natural barrier of the Balkan Mountains to the north and the hilly forests of Sredna Gora to the south, where it borders the ‘Upper Thracian Plain’ (Fig. 1.2). Several rivers fragment both sub-regions into plains consisting of mostly alluvial terraces with altitude reaching 400 m a.s.l. (Galabov et al. 1977, 228-9, Kiradzhiev 1990). This considerably flat terrain does not enclose any orographic barriers that have the potential to obstruct movement.

Currently, as the region is somewhat more accessible for cross-border research, an approach incorporating topographic units to provide a spatial framework seems more appropriate. Although studies based on grouped data corresponding to geographical areas are also problematic because of issues such as ecological fallacy (see Openshaw 1984), replacing political units with topographically defined ones offers at least a less anachronistic methodology. Spanning three contemporary nation-states, the study area employed in this research represents an attempt to overcome bias prompted by illdefined divisions. The section below offers an attempt to briefly outline some of the geographic characteristics of the area relevant to this research.

The eastern part of the Upper Thracian Plain, all the way to the Black Sea coast, consists of small hills, hilly valleys, and low fields. This area is discrete, uniting the flow accumulation of the middle stream drainage of Tundzha to the lower slopes of the Sakar Mountains. Because of the Sakar’s low mountainous topography, the transition between the two sub-regions is smooth, although the mountain’s elevation reaches 865 m a.s.l. (Borislavov 1999, 46). The Sakar are surrounded by two other mountain ranges – the Rhodope Mountains and the Strandzha Mountains, separated from them by two of the largest rivers in the area, Maritsa and Tundzha. The Sakar’s slopes are steep, mostly denuded of forests and sharply dissected by the valleys of the numerous river tributaries, which makes them uninhabitable and less suitable for agriculture. The highest part of the mountain, however, consists of a row of flat plateaus forming a long ridge, providing topographic advantages for human occupation. Geomorphologically similar to the Sakar

Geography The southeastern corner of the Balkan peninsula constitutes a geographic region, often referred to as Thrace, which is geomorphologically and climatically diverse. Furthermore, it is the primary contact point between Asia and Europe as well as the natural link between mainland Europe, the Aegean, the Black Sea and Anatolia (Fig. 1.2). For this

1 2

3

bg. Подбалкански полета или Задбалкански котловини. bg. Горнотракийска низина.

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 1.2. Map of the area with main places mentioned in the text.

lower slopes of the Strandzha Mountains and constitutes the isolated watershed between the Ergene Plain and the Marmara coast. Gelibolu, on the other handconversely, is a small plateau deeply dismembered partitioned by numerous streams. The inland landscape is small-featured and hilly, reaching an altitude of 426 m a.s.l., naturally protected by the coast consisting mainly of cliffs and narrow gorges (Erol 1976; Robertson and Dickson 1984).

Mountains are the Strandzha Mountains, defined by their low, sprawling hills, dissected by river valleys. Along the eastern half of the Strandzha, stretches a comfortable saddle, facilitating natural passage. This mountain range is currently geopolitically divided by the international border between the Republic of Turkey and the Republic of Bulgaria. The Black Sea coast of Thrace stretches from Cape Emine to the Bosporus, which includes parts of Balkan Mountains, the Gulf of Burgas, and Medni Rid in Strandzha Mountains (Galabov et al. 1977, 328; Popov and Mishev 1974). It should be noted that a significant portion of the coast along with the entire Gulf of Burgas was subject to intensive sinking, also recorded in historical times, and this submergence was accompanied by a substantial accumulation of alluvial material (Baltakov 1988, 232), significantly obstructing any archaeological enterprise. Similar is the morphology of the shores of the Bosporus, the Sea of Marmara, and the Dardanelles, once functioning as riverbeds (Erol 1976; Ustaömer and Robertson 1995; 1997) and later played the role of a passage-way between the Mediterranean and the Black Sea.

To the west, the Tekirdağ uplands meet the Ergene Plain and the lower Maritsa Valley. The areas of both basins consist of small hills composed of compact tertiary and later sediments, along with many smaller valleys prone to become marshy most of the year. A flood plain has also formed above the point of confluence of both rivers. Further upstream, there are many marshy tracts between one or the other edge of the plains and their riverbanks, raised by flood deposits. Two major right-hand tributaries, Luda Reka (gr. Ερυθροπόταμος, tr. Kızıldelisu) and Arda (gr. Άρδας, tr. Arda) join the Maritsa at Edirne and, along with its main left-hand tributary, Ergene, draining the slopes of the Strandzha Mountains, periodically flooding the area (Ayanov 1938, 13, 26). The origins of the Maritsa Valley itself, in its section south of Edirne, stem from the sinking of the land blocks that produced the North Aegean Basin (Angelova et al. 1993, 41-59). Moreover, at its lowest portion, the Maritsa is divided into three parallel

Tekirdağ uplands are the southern uplands behind the Marmara coast, which reach the Gelibolu peninsula to the southwest. On the northeast, the area is connected to the 4

Introduction and background rivers instead of flowing in a single channel. Because of this and due to its high-water state, the river forms a significant natural barrier.

the area consists mostly of the delta and the marshy areas of its hinterland on the Greek side of the Greek-Bulgarian border. As a whole, the terrain of the Struma basin is a mixture of mountains and valleys, where the middle stream has a higher altitude than the upper one.

The Rhodope Mountains are a part of the Rilo-Rhodope mountain chain, which is the oldest landmass on the Balkan Peninsula. The average altitude of the entire mountain range is 785 m a.s.l. It is geomorphologically split into the ‘eastern’ and ‘western’ Rhodopes, divided by the rivers Kayakliyka, Borovitsa, Arda and Chepinska (Batakliev 1963; 1969). The western part is significantly higher and more substantial, constituting ca. 66 per cent of the entire Rhodope range (Deliradev 1953) with typical mountainous topography, while the eastern part is hilly and considerably lower in elevation. The average altitude of the western Rhodopes is 1150 m a.s.l., but many peaks and ridges reach 1500 m a.s.l. and above, while the average height in the Eastern Rhodopes is only 329 m a.s.l. with the highest distribution of landforms ranging between 200 and 600 m a.s.l. Broad tectonic basins with vast fields located on top of high ridges, surrounded by deep river valleys, define the geological look of the western Rhodopes. The northern part of the western Rhodopes is considerably lower and the river valleys are wide and shallow, which makes it more inhabitable, while their middle streams in the heart of the mountain range are deep with steep and rocky slopes. The eastern Rhodopes, by contrast, are broken into broad plains, divided by three mountain ridges. Although still mountainous in structure and appearance, the relatively low terrain of the eastern Rhodopes facilitates both movement and living conditions.

Geology and natural resources Geologically the study area is as diverse as it is topographically. The Sub-Balkan Fields and the Upper Thracian Plain are tertiary sedimentary plains consisting of mostly alluvial terraces, defining the flat appearance of the terrain. Thick quaternary and river layers cover the tertiary seabed (Donchev and Karakashev 2002). The Sakar Mountains consist of a dome-like structure formed by a massive volcanic (batholith) core, surrounded by highgrade metamorphic rocks, namely gneiss, amphibolite, and schist (Kamenov et al. 2010). To the east, the Strandzha Mountains are a metamorphic complex, the northern end of which is an ‘east-west-trending flysch-volcanic zone’ (Yılmaz et al. 1997). The Strandzha are also known for their karst landscape, although their highlands are composed of masses of gneiss and granite. Paleogene deposits and related volcanic and tuff constitutions take a central place in the geomorphological plan of the eastern Rhodopes (Georgiev 2007; Vaptsarov 1962, 75; Vaptsarov 1964). Metamorphic rocks, mostly gneiss, schist, marble, and amphibolite with volcanic intrusions, form the entire mountain range. In the eastern Rhodopes, there are also quartzite and diabase. This metamorphic layer is often covered sediment and volcanic rocks, mostly sandstone, conglomerate, limestone, andesite, and tuff (Michev et al.1980, 412). The western Rhodopes sub-region consists mainly of vast masses of granite plutonium covered by a set of metamorphic rocks and occasionally sandstone, conglomerate or tuff (Marchev et al. 1998; Yaranov 1960, 97).

The area between lower Mesta (gr. Νέστος) and lower Maritsa, often referred to as western Thrace or Greek Thrace, is a structural basin. It consists of deep-cut river valleys and basins collecting the waters from the surrounding mountains to the north. There are broad alluvial valleys to the north and vast marshes to the south of the area. It is divided by the lagoon of Lake Vistonis and Porto Lago Bay into an eastern and a western part. The latter differs by the absence of large streams and consists of high coastal hills with a separate drainage system independent from that of the lower flow of the Mesta. The northern margin reaches the area of modern Komotini and ends with the southern edge of the western Rhodopes, while the southern boundary is formed by the northern Aegean coast, including the island of Thasos.

Flatbeds of yellowish sandstones, chalky and marly limestone and marls contour the shores of the Dardanelles. The Black Sea coast of the Strandzha was formed predominantly by Upper Cretaceous volcanic rocks (andesites and andesite tuffs) in layers of marine sediments, where solid andesites, granites and conglomerates form rocky capes (Galabov et al. 1977, 328; Popov and Mishev 1974). The Tekirdağ upland, locked in between the Strandzha, the Rhodopes, and the Sakar higher massifs, is a sedimentary basin (Görür and Okay 1996). The predominant rocks are a sequence of micaceous sandstone and shale (Okay 2008, 37).

To the west, the lower Mesta-Maritsa basin meets the lowest section of the Struma River. In general, the Struma Valley can be roughly divided into two parts: the northern upper stream and southern lower stream. The upper stream cuts into the Osogovo-Belasitsa mountain chain and runs through the southwestern corner of the Republic of Bulgaria. Its western limit is the international land border with the Republic of North Macedonia. The area consists of a river gorge with steep slopes enclosed by the Rila and Pirin Mountains from the north and the east. The average altitude of the area is ca. 806 m a.s.l. The southern part of

To the west, the lower Maritsa Valley and the Ergene Plain consist of small hills composed of soft tertiary and later sediments, along with many valleys prone to be marshy during most of the year. The simple pattern of the valleys is evidence for the geological youth of the drainage system. Old lake deposits of soft limestone, marl and gravel fill the basins of the lower Maritsa Valley and the

5

Late Bronze Age Social Landscapes of the Southeast Balkans Ergene lowland, while the area between the lower Maritsa and Mesta consists of mostly conglomerate and weaker tertiary sediments which at height transition to crystalline rocks.

22, fig. IX), but without any indication for extraction in the prehistory. On the other hand, numerous mineral springs related to the Neogene tectonic processes exist in the western Rhodopes. A large group of springs flows near Velingrad, and another one close to the northern periphery of the mountains, which draws its waters from the Upper Thracian Plain. Marble, limestone, and volcanic tuff are currently extracted from the area as another type of natural resource used in construction (Michev et al. 1980, 414).

Different rock formations participate in the geological structure of the entire Struma basin. Along the upper stream, the majority of rocks consist of old metamorphic rocks, mainly gneiss, schist and amphibolite. In the highlands, sandstone, and more importantly, limestone, dolomite, and conglomerate are more common (Galabov et al. 1977, 404-410). The area is also highly seismic, causing the appearance of numerous high-temperature water springs.

Climate Thrace is climatically transitional, which predefines living conditions as different from those in the Aegean and continental Europe. Immediately north of the Aegean Sea lies ‘Europe beyond the olive trees’ (Leshtakov 2006, 142; Yaranov 1940, 10-12), which hosts specific climatic characteristics defined by a mix of mountains, plateaus, valleys, and plains, cut by several large rivers Vardar (gr. Άξιος), Maritsa, Struma, and Mesta. A significant climatic barrier for the area was and still is the Balkan Mountains, a substantial single factor that affects local microclimatic conditions by preventing the northern currents penetrating the areas south of the mountain range. Its elevation defines the existence of a ‘north’ and a ‘south’ in the Balkans, where the south is warmed much earlier in the year than the north. In the neighbouring area just north of the Balkan Mountains, the climate is entirely continental, consisting of cold, wet winters (with temperatures as low as -30 degrees Celsius) and cooler summers. The south is much warmer in the summer and reaches 38 degrees Celsius, even in the area of the Rhodope Mountains, which, in turn, stop the dry southern winds. As a result, a trans-continental climate prevails in the space between the Balkan Mountains and the Rhodopes, where the corresponding Black Sea coast differs only through its warmer winters. The Black Sea has a mitigating effect on the climate of the shoreline, where autumn is more temperate than spring and the winter months have average temperatures above 0 degrees Celsius. On the south coast, there is also an interweaving of the Black Sea with Mediterranean climatic influence. The Black Sea strongly influences and has a climatic impact on the weather in the Strandzha region and the Ergene Plain (Ayanov 1938, 10, 21).

Besides the variety of geological settings distributed across the study area, there are some exploitable ore deposits and natural mineral sources. Although consisting of mostly alluvial terraces, the Sub-Balkan Fields, for example, contain deposits of copper, manganese and pyrites (Galabov et al. 1977, 229-231, 223-226). With similar geomorphology and sporadic copper deposits, the Upper Thracian Plain also has the distinct advantage offered by mineral water springs located near Plovdiv, Stara Zagora and Simeonovgrad (Borislavov 1999, 38). As a highly karstic area, the Upper Struma Valley also contains hot mineral sources (Michev et al. 1980, 19). The Sakar area is deficient in minerals, except for some auriferous river streams. In contrast, the Strandzha are known to contain iron and copper deposits, with evidence of exploitation of some of the copper ores in the area during the fifth century AD (Konyarov 1953, 18; Raychevski 1986, 318). During the 1970s, there were more than fifty copper ore deposits and separate ore sections registered in this area. (Chernykh 1978, 61-62). Near the Slivarovo village, there are also gold deposits (Archibald 1998, 23), without evidence for extraction in the past. Several copper-rich ore deposits appear in the low mountainous areas of the Black Sea coast (Borislavov 1999). The complicated geological processes in the eastern Rhodopes are the reason for the formation of specific ores and minerals. Along with the abundance of copper ores, there are also isolated deposits of obsidian, near Dzhebel. There are iron ores accumulated in the region of Haskovo, and silver and gold deposits around Madzharovo, Zvezdel, and Sedefche (Nikolaev at al. 1976). More recently, gold mines have been excavated near Krumovgrad with evidence of exploitation during the second millennium BC (Popov and Jockenhövel 2011; Popov et al. 2011a; 2011b; 2015; 2017; Popov and Nikov 2014). Furthermore, most of the rivers and streams all over the Rhodope Mountains are auriferous (Archibald 1998, 21-22, fig. IX).

The Upper Thracian Plain also has typically mild winters favourable for sustainable agriculture. The relatively low relief of the eastern Rhodope Mountains sharply increases the influence of the zonal hydro-climatic factors by its differentiation as a separate geographical unit (Galabov et al. 1977, 209-219). The sub-region is extensively open to the east and southeast in the direction of the Mediterranean. To the north, the mountains’ slopes steeply descend towards the Upper Thracian Plain. The steep slopes and the frequent heavy winter rainfalls, contribute to the intensity of the present-day erosion processes. South of the Rhodopes, the climate is slightly wetter and milder than it is further south in the Aegean, however, as in the

In the western part of these mountains there are numerous locations with iron and gold ore deposits, as well as copper, lead-zinc and silver (Archibald 1998,

6

Introduction and background rest of modern Greece, the summer rainfall is significantly low (Archibald 1998, 18).

was suitable for oak forests during all prehistoric periods, as well as olive cultivation from the second millennium BC onwards (Nikolova 1999).

The climate along the Struma Valley changes from continental to Mediterranean in the direction of north to south, and it is favourable for a variety of agricultural uses. Although it allows the intrusion of northern Aegean air currents along with olive trees and citruses in large part of the area, in the highlands, similar to the western Rhodope range, the climate acquires an entirely mountainous character and precludes the cultivation of Mediterranean crops (see Popova 2009; Popova and Bozhilova 1998).

Natural routes The mixed set of landscapes of relatively small valleys and vast plains divided by mountain chains, which lies between the northern Aegean coast and the Balkan Mountains, has had a profound impact on the subsistence systems on the extent of interaction between the cultural groups of the area, throughout the ages. The segmentation described above naturally encourages the occurrence of local as well as interregional routes. Since the early prehistoric times, the region has been a crossroad of human communication and occupation, where east meets west and north meets south. It is the link between Anatolia and mainland Greece as well as the bridge between the whole Aegean world and the farmlands of the Lower Danube (Elster and Renfrew 2003, 6). Placed strategically at the crossroads of Europe and Asia, the area has proven both a tempting object of conquest and a passageway. The straits of Bosporus and Dardanelles have routinely played a dual role as both a dividing corridor between the Mediterranean and the Black Sea and a connecting link between the two continents. All this has preconditioned the existence of local as well as interregional routes, with the region being culturally and climatically transitional between the Mediterranean and central Europe.

Soils and vegetation The entire Upper Thracian Plain, and to an extent the SubBalkan fields, are known for the quality of the arable land. The most fertile black chernozems and mollisols, similar to those in Ukraine, Moldavia, and the Carpathians, can be found along the valleys of Maritsa and Tundzha, as well as in the Strandzha Mountains. Unlike the fertile valleys of the rivers in the Upper Thracian Plain, the valleys of the Struma basin are stripped of fertile soils. Most represented are alluvial and cinnamon forest soils and only by exception are there chernozems (Minchev 1980). Alluvial deposits cover almost entirely the lower valleys of Struma, Maritsa, Mesta and the Ergene basin due to their marshy appearance. Large areas of the eastern Rhodopes are also alluvial, mostly along the river valleys dissecting the landscape, but cinnamon forest soils and occasional mollisols are covering the highlands in the northern part of the mountains (Shishkov and Kolev 2014). The soil coverage in the western Rhodopes consists mainly of brown forest and cinnamon forest soils (Galabov et al. 1977, 165-405).

We learn from historical sources that both main and auxiliary roads existed in the region in the first millennium BC (Brownson 1998, Xen. Anab. 6.3; Jones 1924, Str. 7.7.4), some of which might be considered as likely to have been functional still earlier. It is known that the main Roman arteries crossing the Balkan Peninsula were already long-established, convenient routes. Little is known about the Via Pontica, but its trajectory can generally be traced from north to south, along the west coast of the Black Sea. The Via Egnatia crossed the northern Aegean coast and connected the Adriatic coast with the Bosporus. The Via Diagonalis also started from the Bosporus area but travelled towards the Danube via the valleys of Maritsa and Morava, passing through modern Serbia (Popovic 2010). Hellenistic sources also mention distances to Philippi and Amphipolis from different places in the northern Aegean (Collart 1935; Koukouli-Chryssanthaki 2001). Thucydides describes a thirteen-day trip between Byzantion (Istanbul) and Vitosha Mountain (near Sofia, 2.97.2). An inscription from Pistiros further offers information about a road linking Upper Thrace and Thasos through Maroneia and Apollonia Pontica (Valeva et al. 2015).

The range of rock compositions in the Sakar region consists of a variety of soils, amongst which the cinnamonforest soils dominate. The most unequivocal evidence of trans-Mediterranean climate in the Stranzdha region is the presence of podzolised soils, a relic soil type, formed in a once warm and humid environment (Borislavov 1999, 48). The soil coverage consists of brown forest soils in the western part and secondary brown forest soils in the eastern part of the sub-area (Galabov et al. 1977, 209-219). There is not much known about the vegetation of the study area in the prehistory, but there is some research to provide insight. The paleobotanical record in the Sub-Balkan Fields indicates that in the past the region was covered by deciduous forests, mainly oak (Borislavov 1999, 61). There is evidence that in the past woodlands covered the Sakar Mountains, even if today it is almost completely deforested. It seems that dense forests covered the slopes of the Strandzha range until the beginning of the twentieth century, making it highly impenetrable (Ayanov 1938, 10, 21). An opinion exists that the climate in the Drama plain

Natural routes link the Upper Thracian Plain and the northern Aegean through the valleys of Mesta, Maritsa and the Black Sea littoral via the Straits (Batakliev 1942; Delchev 1965, 10; Leshtakov 2006, 144; Stuard 1997; Nikolova 1999). A maritime route between the south and the north has been proposed to explain the distribution of

7

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 1.3. Map of the region indicating natural routes following main river arteries (black dotted lines) and mountain passes (red dotted lines), associated with present-day locations.

oxhide ingots and stone anchors (Velkov 1972; Leshtakov 2007a). Imported objects exist from throughout the first millennium BC, tracing the entire area of the western Rhodopes to the Upper Thracian Plain (Archibald 1998, 14), complemented by some second millennium BC Aegean weaponry finds (Bonev 1988; Leshtakov 2011; Panayotov 1980; 1981). This evidence suggests that communication arteries through the western Rhodope passes existed with or without continuity during both the first and the second millennia BC.

flocks to winter in the Aegean lowlands (Batakliev 1942). Migrant artisans, mainly tailors and masons, exploited the same routes when transferring from the Pirin and the Rhodope Mountains to the Aegean Sea in the late nineteenth century. There are several other paths in this area, long used by transhumant herdsmen, the nomads of the Balkans (Wace and Thompson 1912). In the eastern Rhodopes, the direction of the river system changes from north-south to east-west. This change provides relatively convenient connections in the eastwest direction; the major tributaries of the river Arda cross the mountain range here in a north-south direction. A convenient pass is a depression located along the river Suyutliyka in the middle of the valley of Ardino. It is a natural link between the modern towns of Kardzhali and Haskovo, connected with the south along the river valley of Varbitsa and through the pass of Makaza leading to Komotini. The lowest and most suitable passage in the eastern Rhodopes is indeed Makaza, which joins Kardzhali with Komotini (Batakliev 1942, 20). Here, together with the Avren Pass, one meets conditions to cross over the eastern part of the mountains and to connect the Upper Thracian Plain with the Aegean. The natural longitudinal road in the eastern Rhodopes follows the valley of Arda. This road is known to be used by the tailors (abadzhii), who would come down from Smolyan to Edirne (Batakliev 1942, 199). Further west, the Axios Valley is considered the main route joining south with north (Theocharis 1971), leaving the Struma Valley a secondary choice (Elster and Renfrew 2003).

Natural routes in the heart of the Rhodope Mountains can be traced mainly through the major river valleys (Fig. 1.3). The Mesta Valley in the western Rhodopes appears to be relatively easy to pass until the area south of Nevrokopi where it becomes too narrow; one can reach the Drama Plain at this spot through the Zarnevski Pass. The Mesta Valley is a convenient way to cross the mountain since from its upper course the route continues northwards through Yundola, along with the valley of the river Yadenitsa to Belovo where it enters the Upper Thracian Plain. Another natural connection exists between the modern towns of Smolyan and Xanthi and from there, along the valley of the Chepelarska River, one could easily approach the area of Assenovgrad to the east and the Maritsa Valley. There is also a natural connection between the towns of Nevrokopi and Serres. The river valley of Matnitsa serves to link the plain of Batak with Chepino. Some scholars describe these routes as utilised by the seasonal migrants, the Yurutsi, who summer in the Rhodopes and lead their

8

Introduction and background The Maritsa River constitutes the main drainage network, which was navigable by small rafts in the past. There is evidence for the transport of goods between the northern Aegean coast and the Thracian hinterland along the Maritsa (Archibald 1998, 13). Many also consider the valley of Maritsa the most convenient route from Asia Minor to the central parts of the Balkan Peninsula and, from there, to central Europe. The economic and strategic importance of this valley as a route has made an impact on the concentration and location of settlements in prehistory. Along its full length, this route crosses other paths between north and south through the Rhodopes and the Balkan Mountains and thus acts as a hub of different influences and goods approaching from different directions. Leshtakov offers a method for reconstruction of the meridional pathways between the Upper Thracian Plain and the Aegean in prehistory (Leshtakov 2006). He linked reconstructions of the paleoenvironment, geomorphology and palaeoclimatology, complemented by data about certain shepherd groups (Yurutsi, Karakachani, Vlachi and Bulgari) and their seasonal migrations, ethnographic sources and local history. The author argues that the entire Rhodope Mountains could be crossed over, in a south-north direction, in less than ten days (Leshtakov 2006, 144). Concerning ethnographic examples from the nineteenth century, Leshtakov suggests that people were able to sail along the Maritsa River from the area of modern Pazardzhik to the delta in 5-6 days, with the return journey taking less than a month (Delchev 1965; Stuard 1997). According to Nikolova (1999, 24), the western and highest part of the Rhodope range was not a severe barrier to contacts and interaction between the north and the south. The Rhodope passes were the usual way to connect western Upper Thrace with the northern Aegean. Theocharis argues, however, that prehistoric inter-communications between the Aegean Thrace and inland Thrace were problematic because of the Rhodope mountain barrier, which obstructs movement from south to north and vice-versa.

Accordingly, the Mesta valley can hardly be described as a pass, since it is mostly hilly and mountainous with relatively high elevation. However, the Theocharis saw the value of the river system – the Èvros, Strymon and their main tributaries, namely the Tundzha and Arda – and links it to the still more critical route leading from Axios to Anatolia (Theocharis 1971, 144). Nevertheless, even if one would typically assume that the ancient routes followed the most convenient and easily passable locations, having desirable physical and geographical features (Theodossiev 2000, 16), it is essential to bear in mind that the cultural and potentially cosmological nature of human choices across the landscape. ‘The knowledge of the ‘natural’ routes of exchange over mountain-passes, through river valleys or over the sea is usually taken as granted, thereby unconsciously transferring today’s such a familiar geographical impression of the world to the distant past …. Moreover, M. H. Helms has reminded us that notions of space and distance are culturally created.’ (Maran 2007, 4; Helms 1988). Several aspects emerge from the preceding discussion. The region of Thrace was exceptionally suitable for habitation, with water sources, fertile soils, accessible raw materials, generally mild climate, varied relief and vegetation, and a direct connection with the Aegean and the Black Sea. Such geographic and ecological factors often determined a range of prehistoric developments in the Balkans, but it is challenging to isolate, for instance, only ‘lowland agricultural’ or ‘mountain cattle’ groups (Delchev 1965, 7-12). It seems possible that various micro-regions throughout the prehistory have consistently favoured a mixed inter-regional economy (Earle 1997, 65; Nikolova 1999, 23). In particular, I will argue in the following chapters that the period between 1600 and 1000 BC was the time when a degree of ‘connectivity of micro-regions’ existed (Purcell and Horden 2000).

9

2 Change and identity: a theoretical perspective

So far I have outlined and contextualised the main questions that will be discussed in the following chapters, where they will be investigated through an analysis of the variation and distribution of certain elements of the material culture of the southeast Balkans. Built upon this analysis, the study will aim to identify both temporal and spatial changes in social practices and more specifically, to define some specific cultural characteristics of the local LBA. In a broader sense, the focus will be on the spatial variation of local pottery and settlement systems and the importance of the observed patterns for the reconstruction of the past.

together in a shared life, i.e. culture can be learned, thus supplementing instinctive behaviour (Clarke 1968, 83). In this sense, culture is also used as a means for adaptation to the environment, based on knowledge, traditions, and experience. Defined in this framework, culture is a notion that requires the participation of most of the members in a group, where interactions between individuals constitute a larger network of social interaction in that society. Cultures and societies continually transform regardless of their complexity. As Bentley and Maschner (2007, 246) wrote, ‘each generation learns the culture and brings a limited amount of change to it’. Such cultural changes may appear because of new inventions and ideas developed internally or as a result of interaction with a ‘new’ group of people, and thus can also be exogenous to the local culture (Orser 2017, 61). Some of the main triggers for change that have been discussed are environmental cataclysms, demographic growth or technological innovations, and, increasingly, population movements.

Archaeologists have always been concerned with change in material culture, followed by an association with cultural, social, and political causes. Nevertheless, such tendencies have been predominantly directed by the individual’s beliefs and by the unavoidable influence of the social context of their times (Chapman and Hamerow 1997, 2), expressed in often controversial theoretical paradigms. Therefore, the terminology and the implied correlation between material remains and social phenomena need to be discussed to address the extent to which our data can be helpful in the attempt to understand the nature of change.

Demographic growth There is often an emphasis on the role of human populations and their demographic dynamics in cultural change. According to a widespread view, technological advances such as the lithics industry (Shennan 2001) or advances in food production (Cohen 1977; Hassan 1978,), must have accompanied and been accompanied by a demographic rise (see Hassan 1973; Shennan 2001). Some scholars have argued that technological, social and ideological change takes a population out of its homeostatic state and encourages expansion and growth (see Chamberlain 2006, 4). An opposing model, proposed by Binford (1968) and Renfrew (1973), sees cultural change as caused by population growth rather than vice versa. Studies have treated fluctuations in the size of a population as though the availability of resources determined them. Thus, a society should reach an upper quantifiable demographic limit after the exhaustion of vital supplies. The problem of growth, then, is solved by innovations, technological advance, change in land use or extensive exchange practices.

Culture and culture change Within most disciplines that study humans, including archaeology, some general issues have shaped core research agendas. Arguably we could summarise them into two main questions: ‘what is culture?’ and ‘why culture changes?’ (Johnson 2010, 68; Malinowski 1945). Although the term ‘culture’ may appear straightforward in our modern perception, many studies have proven that culture is a very complex and challenging concept, resulting in a plethora of definitions (see Reide 2011). One group of meanings focus on the established behaviour of a person, who is considered ‘cultured’ or ‘civilised’ regarding their manners, while another group of meanings refers to observed differences between groups of people, allowing them to define each-other as foreigners (Orser 2017, 56). Following the latter, culture could be seen as a system uniting members of a society based on their belief system, knowledge, customs, art, but also the physical remnants of their actions (see Renfrew 1984). This composite nature of culture includes elements from practices such as textile production or ceramic production, to mythology and dance, and defines the ways of being accepted by the members of the group. Culture belongs to a group of people brought

Intensification in subsistence agriculture was often thought to be related to population increase, but the direction of cause and effect is also disputed (Shennan et al. 2013). For example, Snodgrass attributed the rise of the Greek city-states to demographic increase, which he considered a primary factor leading to intensive agriculture (Snodgrass 11

Late Bronze Age Social Landscapes of the Southeast Balkans 1980). Already in the late nineteenth century, change was ascribed to pastoralism, giving way to farming as population densities increased (Nilsson 1868 in Trigger 2006, 225). The so-called ‘oasis’ theory, according to which postglacial drought in the Middle East forced people to cluster around water sources and innovate to feed higher population densities (see Childe 1928), was also based on this concept. There was also an argument that cultural development inevitably resulted in population growth, whereas ‘primitive’ people declined in numbers or remained static (Lubbock 1865 in Trigger 2006). Finally, many authors have noted that culture change may itself result in demographic change, and therefore the demographic signature of migration or related processes may remain unclear (Adams et al., 1978; Anthony, 1990; 1992). Either way, demographic growth has been seen by many as a prime driver of prehistoric culture change and was very prevalent in models tracing the agricultural spread from the Near East to Europe (Ammerman and Cavalli-Sforza 1973; Chamberlain 2006; Zubrow 1989).

the region sometime during the late fourth millennium BC. It also corresponded to considerable demographic growth in the Alps and north-central Europe. Another commonly invoked climatic episode takes place towards the end of the third millennium BC, also known as the ‘4.2k BP event’ (Höflmayer 2014), as part of which Şahoğlu (2005) amongst others has suggested the climate-induced collapse of the extensive Anatolian Early Bronze Age networks. A dramatic cultural shift that has been ascribed to rapid climate change is the destruction and abandonment of many eastern Mediterranean urban centres at the end of the Bronze Age (Dickinson 2010), with Kaniewski et al. (2010) arguing that a ‘centuries-long megadrought’ (Drake 2012, 1862) was responsible for the collapse of the Bronze Age palatial systems. The comprehensive analysis by Drake shows evidence for changes in the entire Northern Hemisphere during the LBA collapse (Drake 2012, 1864). Drake envisaged this climatic cataclysm to have occurred between 1315 and 1050 BC, after which the entire Aegean appears more sparsely populated for more than three centuries (Desborough 1964). The arrival of presumably ethnically diverse ‘Sea Peoples’ to Egypt and the eastern Mediterranean is, according to this logic, associated with the relocation of west-central Mediterranean and Aegean groups to other areas, and argued to be visible in changes such as local pottery from the Levant and other regions being executed in an Aegean-influenced style (Drake 2012, 1863).

Climate change as a forcing factor Environmental factors such as climate change have also often been considered as a cause of culture change (see Höflmayer 2014; Wiener 2014). Some of the earlier prehistoric transformations on the Balkans have been connected to climatic cataclysms, and similar explanations have been proposed to episodes in the second half of the second millennium BC. While far more dramatic climate change events can be documented for early geologic epochs of the earth’s history, serious if smaller-scale climate changes have also been recorded for the later Holocene. While there can be little argument that largerscale climatic shifts had an impact on hominid and human evolution, it is far more debated whether these later, smaller-scale climatic events were still important enough to have had a primary causative role in culture change. For instance, some researchers have connected early farming and herding practices in the Near East and changes in the social organisation with contemporaneous climate conditions, further arguing for the existence of repetitive global cooling anomalies known as Rapid Climate Change events (e.g. Weninger et al. 2009). Some of these anomalies are often pointed to as one of the possible reasons for dramatic shifts in the material culture, alongside economic and military explanations.

That a climatic or environmental factor played a role in such LBA ‘catastrophes’ was proposed for the first time by Carpenter (1966). However, Kaniewski et al. (2010), were the first to claim that climate change was a significant cause of the Bronze Age collapse, demonstrated through a three to four degrees Celsius decline in the Ionian Sea region and a one to two degrees Celsius in the Adriatic. More climatic shifts with different significance can also be seen in a larger span between 1694 BC and 1197 BC (Rosen 2007; Staubwasser and Weiss 2006). The general conclusion is that between 1350 BC and 1124 BC temperatures dropped significantly. In contrast to this view embracing the significance of smaller-scale environmental catastrophes, a more critical view of the impact of climatic shifts on past societies argues that the effect of climate on culture can vary and depends on the ‘rapidity, magnitude, duration, and frequency’ of such events and indeed of the resilience of specific cultures (Broodbank 2013, 50). Whether or not climate events were responsible for the collapse of the Aegean and Anatolian Bronze Age societies, I believe that such fluctuations have impacted the regional demographics and possibly caused some widerscale population shifts affecting the southeast Balkans.

Bar-Matthews et al. (1998; 2003) and Meller et al. (2015) discuss major climatic events in the Holocene, where earlier episode associates with events around 3150 BC and a later low precipitation crisis or dryness and drought occurring in the Aegean and Anatolia ca. 1200 (Bryce 1998; Dickinson 2006; Wiener 2014). There is a hypothesis which considers the entire Bronze Age as influenced by a climatic episode that led to the highest registered level of the world’s oceans after 3500 BC (Todorova 2007). Such an event seems to have coincided with the gradual depopulation of Anatolia and the Balkans and the end of the Chalcolithic cultures in

Metals and Metallschock Innovation is another factor often associated with change. Amongst other innovations, the adoption of new metals for mass production of tools and weapons is probably the

12

Change and identity: a theoretical perspective most significant, so much so that it is used to label entire chronological periods. Schachermeyr (1955) suggested the term metallschock, later developed by Branigan and Renfrew in the context of the EBA II period in the Aegean, as a way of describing a sharp increase in metallurgical activities, but also to stress the elevated ideological importance of metal in the Aegean in general (Papadatos 2007, 154); a phenomenon that has also been described as an ‘explosion’ of metallurgy (Kayafa et al. 2000, 39). More generally, metallurgy has been argued to have triggered an Aegean-wide increase in cultural interactions, because metal was the first ‘commodity worth trading’ (Renfrew 1972, 455). Renfrew also considered the rise in metallurgy in the Aegean as a part of an ‘international spirit’, consisting of intensive interaction between regions and people exchanging raw material, technological skills, knowledge, ready-made objects, and abstract ideas (Renfrew 1972, 451-455).

of copper metallurgy seems to have collapsed and been replaced entirely by the use of arsenical bronze in the early fourth millennium BC (Vaysov 1993). According to Chernykh, the Circum-Pontic province lasted until the midsecond millennium when diverse regional metallurgical traditions from the European and Caucasian traditions came to replace it (Chernykh 1992). The European metallurgical province, however, was an interaction area already present before the beginning of the second millennium BC and is often considered a driving factor in economic growth, social complexity, and elite stratification. Important for this discussion is that of Bulgarian bronzes, the EBA ones are mostly arsenical bronze, or ‘pure’ copper and metallurgy only changed dramatically in the LBA with a large number of tin bronzes (Chernykh 1978). By contrast, regular use of tin bronzes in the Carpathian Basin started during the Romanian Middle Bronze Age (MBA) (Pare 2010, 14-15). It should be noted that tin bronze artefacts appear south of the Danube only with the emergence of scrap hoards during the LBA, following a Central European practice from the first half of the second millennium BC. The pattern of hoards changed ca. 15001300 BC, when scrap metal and oxhide ingots began to circulate regularly (Dietrich 2014). With the increase in the amounts of copper, tin, and bronze available, a lowlevel short-distance exchange possibly facilitated the longdistance trade mechanisms (Primas and Pernicka 1998). Whether or not the adoption of bronze was unavoidable as a techno-evolutionary imperative, the abundance of bronze artefacts must testify for the existence of long-distance relationships at least in pursuit of tin towards Afghanistan and on the Iberian Peninsula (Harrison 2011, 10). It is worth noting that the sudden appearance of tin bronzes, in fact, – is a common phenomenon in EBA II in the Aegean in general, which suggests a pre-existing network and knowledge on the procurement ways of obtaining at least raw material.

For the Bronze Age, in particular, the defining aspect implied by this label is the development or adoption of bronze technology as a dominant source for a wide range of products all over Europe, where the most critical and complex point is the transition from ‘pure’ copper to alloyed bronze. The spread of bronze technology in a strictly diffusionist way was long considered to be from the Near East towards Anatolia, then to the southeast, central, and finally northwest Europe (see Sherratt 1993). Diffusionist models dominated many of the early studies and usually involve travelling agents from the Near East in search of copper and tin. Recently, most of the Carpathian and Central European cultures have been re-dated via radiocarbon dates to much earlier than previously thought with a transitional period of metal use argued to exist between the Chalcolithic and the Bronze Age going back as far as 3500 BC (for example Glina III – Schneckenberg B). Such an early date for metallurgical innovation could be accounted for, at least partially, by the independent development of bronze production within Europe and by possible broader employment of European ‘tin-specialists’ from central and western Europe, for example as part of the Bell Beaker phenomenon (Pare 2000). Furthermore, the distinctive development of an earlier copper smelting tradition from as early as ca. 5000 BC as a part of the Vinča culture, along with the early development of metallurgy within the Chalcolithic complex of KodzhadermenGumelnitsa-Karanovo VI (see Dzhanfezova 2012), indicate the potentially independent origins of Balkan metallurgy (Radivojević and Rehren 2016, 205).

This type of evidence signifies that weighed metal was possibly used as a means of payment and corresponded to the rise of the Tumulus culture and the end of tell settlements in the Carpathian Basin (David 1998; Pare 2010). Nevertheless, the hoards and the majority of bronze finds remained concentrated into the north of the Balkan Mountains, while the southeastern part of the Balkan peninsula stayed outside of the zone of intensive distribution. There is a hypothesis postulating that the regulated exchange system in the region collapsed due to the increased amount of tin bronze circulation and was replaced by weighed fragmented bronze. After 1600 BC, this new system was adopted across the entire European Metallurgical Province, and in the Urnfield period a large number of scrap hoards had already been accumulated between the Carpathians and the Atlantic (see discussion in Pare 2000 and Sherratt 2000). While these processes might have affected the development of communities in Thrace during the second millennium BC, the region was somehow left out of the bronze circulation zone, possibly

The ‘metallurgical provinces’ (arguably a substitute term for an archaeological culture defined by pottery), first established by Chernykh (1992), play an important part in the discussion of bronze metallurgy in the Balkans. Chernykh saw a significant change in southeastern European metal production at the end of the Chalcolithic, which he described as the replacement of a CarpathoBalkan metallurgical province by the Circum-Pontic one (Chernykh 1992; Pernicka 1990). This early ‘boom’

13

Late Bronze Age Social Landscapes of the Southeast Balkans not participating intensively in regional trade or processes related to the extraction and exchange of raw materials. Some Aegean types of bronze artefacts appeared during the sixteenth and fifteenth centuries BC, but their number remained insignificant. During the final stages of the LBA, some new bronze tool and weapon types, originating in the northern Black Sea and central Europe emerged in Thrace and remained in use until the adoption of iron as raw material (Leshtakov 2011, 48).

where it continues to be. It is true that understanding human movement and mobility in the past is essential but challenging. It is also clear that in the latter part of the twentieth century, migration as an explanation of cultural change in archaeology became unfashionable, even, one might say, stigmatised. Like other aspects of the culture-historical approach in post-Second World War western Europe and America, migration studies also became obsolete. It would be fair to say that the lack of research on the causality behind, and the mechanisms involved in assumed migrations in earlier migration studies played an important role in this, as well as the dearth of comprehensive studies on how migration could be expressed in the material record. The re-emergence of the population-movement theme came indeed with the attempt to explain some of the actual mechanisms of movement (see Anthony 1990; 1992; 1997; 2007; Heather 2009). Essential for migration studies was the proposition that migrations need to be considered not as events, but as social strategies, where a substantial effect on the decision has a combination between ‘push’ and ‘pull’ factors (Anthony 1990, 905; 1997, 22). The decision can be individual or collective, socio-economic or not rational at all (Lee 1966, 51). Antony suggested that we should think of the possibility of an initial wave of migration, followed by a ‘counterstream’ moving back to the perceived homeland (Anthony 1990, 908). This initial migration of an individual, family or a small group is exploratory since people need to investigate the conditions in the target destination (Wiseman and Roseman 1979, 330331). Return migration is also a significant aspect in many migration streams, and at least some of the ‘imported’ goods traditionally considered as trade objects should instead be identified as a product of such a back and forth movement, especially when addressing a limited number of foreign commodities (Anthony 1990).

Cultural diffusion and population movement The concept of diffusion defines the transfer of physical mannerisms from one culture to another (Childe 1950). Hypotheses of cultural diffusion are often central, if controversial when addressing culture change. Diffusion has frequently been deployed to account for the spread of cultures across the world. The idea of cultural cores and peripheries, where innovations appear over extended periods in particular areas and then diffuse towards the related fringes, is based on Montelius interpretation of cultural development (see Trigger 2006). The basic idea of diffusionism is that external ‘influences’, originating from areas outside of a region result in culture change. In this sense, change is entirely exogenous to the system. Thus, within the diffusionist framework, the spatial distribution of specific material culture features is considered a result of contact with an outside area (Renfrew and Cherry 1986, 7). While this contact might cause the diffusion of ideas between groups, in many instances migrations are also used as a theoretical tool, often assuming a replacement of one group of people and their culture by another. An alternative concept explaining culture change, is that of peer polity interaction, developed by Renfrew and Cherry (1986), in which polities of similar standing interact with each other through warfare, trade, gift-exchange, elite emulation, etc., mutually transforming their cultures.

Another critical issue is ‘who moves?’. Is it the extended family, the village or an entire group? How do we distinguish long-distance movement from short-distance, which could have potentially different consequences and traces in the archaeological record? In this sense, there are also different types of movements associated with different conditions, where, besides the oft-invoked population pressure, there could also be economic factors, forced or conflict-induced migration due to invasion, - and reasons concerning culture-specific values and belief systems, environmental stress or a combination thereof. It should also be borne in mind that population movement at any scale does not start as a mass-migration, and usually, an entire ‘culture’ does not migrate, but only specific, goalorientated sub-groups (Anthony 1990, 899-900, 908).

It is worth emphasising that the concept of diffusion covers a range of social, economic and cultural transfers with potentially different impacts. We should, therefore, recognise that the approach involves complex social processes with different actors. Moreover, diffusion cannot be an explanation for change on its own, but it is one among many partaking mechanisms (see Kristinasen 2005). It is also important to acknowledge and investigate the process of transfer since the object does not travel without the agency of people. The movement of individuals or migration of groups has long been employed to answer some of archaeology’s biggest questions and thus has been a focal point of archaeological theory for decades. Many studies have used migration to explain change in the prehistory, especially in eastern Europe, where the heritage of the culturehistorical approach (see below) was rarely challenged, unlike in western European and American archaeologies,

Considering migrations as a process, if visible in the archaeological record, their paths should be recognisable. When interpreting migration through the material record there are three key points: i) migrants do not keep their material culture in their new environment exactly as it was in their previous location; ii) cultural changes must

14

Change and identity: a theoretical perspective be expected due to the impact of local societies, as total population replacement is an anomaly in human history iii) The material evidence in the destination area must be later than in the area of origin (Collett 1987, 115). One good example of such processes is provided by the gradual mixture of material culture between the agricultural towns of the Cucuteni-Tripolye culture and the horsedomesticating Yamnaya culture from the Pontic-Caspian steppes. After their initial mutual expansion, CucuteniTripolye cultural forms appear to be eventually replaced by Yamnaya, with the latter argued to have defined the Bronze Age development of much of the rest of Europe (see Anthony 1990; 2007; Klejn et al. 2017; Kristiansen et al. 2017).

approach are rooted in the last decades of the nineteenth century when emerging archaeological research in central and eastern Europe was as perhaps expected, influenced by western European evolutionary perspectives. Once scholars of ancient studies rejected the idea of cultural evolutionism, a central internal drive for change lost its popularity, which promoted other possible explanations, especially external factors such as the diffusion of ideas and migration of people (see Slobodin 1978). To explain why and how the culture-historical approach became the most popular vein of thought in Bulgarian archaeology, I need to highlight two points: a) the lack of emphasis given to independent cultural development, and b) the belief that most inventions like pottery or bronze technology are unlikely to have been discovered more than once in different areas. This left diffusion and migration the most likely explanation of cultural change (Trigger 2006, 218).

Although reviving all culture-historical migration theories, recent developments in bioarchaeology, ancient DNA and strontium isotope analysis make it possible, at least in principle, to connect different individuals genetically or to identify particular people as coming from specific geochemically-defined regions (Boric and Price 2013; Cavalli-Sforza et al.,1994; van Dommelen 2014, 479; Kristiansen 1998; Kristiansen and Larsson 2005;). These new lines of research render it more possible to demonstrate the plausibility of past migrations scientifically. This re-emphasis should encourage further analysis of the motivations behind people’s movement, the mechanisms involved, the likely composition, scale and character of the moving groups and the actual consequences (van Dommelen 2014, 480). With population movement put back on the table as a topic, more weight has concentrated on the exact mechanisms involved, such as demic diffusion, the so-called folk migration, leapfrog colonisation, acculturation of local population or hybridisation (see Hakenbeck 2008; Stockhammer 2012). As an example of hybrid society, Boric and Price (2013) state that, in the Danube Gorges during the Late Mesolithic/Early Neolithic transition, there is a well-documented mixed population from more than one place of origin, supplemented by cultural hybridity visible in the material culture, as well as some innovative settlement structures and foreign burial rites. They also argue that a route through the Bosporus, linking Anatolia and the Balkans, must have been a central axis for the dispersal of Neolithic groups from Asia and that during the early period of prehistory, dispersal must have happened rather rapidly, for about two centuries, covering vast territories (Boric and Price 2013, 3302). While remaining cautious, the same potential holds for this region in later periods.

One of the reasons for the stigmatisation of culturehistorical archaeology is the fact that its development coincided with the rise of nationalism, promoting ethnicity as ‘the most important factor shaping human history’ (Trigger 2006, 211). Nationalism as a concept integrates a sense of national identity and the sharing of a common homeland, but not always explicitly of shared ethnicity. A change in that perception, however, occurred when most European nation-states came to identify themselves with a specific ethnicity, based on a language, a common culture, and a perceived shared history. This trend created strong links between the politics of northern, central and eastern Europe on the one hand, and prehistoric archaeology which was building cultural chronologies and tracing the perceived deep roots of their countries’ pasts, on the other (Fischer and Kristiansen 2002). This was a troublesome period for the discipline of archaeology, where nationalist and racist agendas determined the archaeological discourse. This forever tainted the legitimacy of the culture-historical approach in post-war War western Europe and North America. In eastern Europe, archaeological theory followed a drastically different path and culture-historical archaeology never lost its importance. For example, after 1945 when eastern Europe was recognised as a part of the Soviet sphere of influence, an International Congress of Slavic Archaeology was founded to encourage closer relations among Slavic nations and, thus, research on Slavic ethnogenesis received substantial resources. This, however, did not restrict research only to chasing Slavic origins, but also encouraged the opening up of questions concerning the time of the arrival of the ‘Indo-Europeans’ in Europe, as well as a debate on the origin of the Thracians in Bulgarian and Soviet studies (see Fol 1997; Tacheva 1987).

Culture-historical archaeology Many of the issues described above have been at the centre of various theoretical paradigms throughout the history of archaeological thought. Traditionally, archaeological research in eastern Europe and more specifically Bulgaria has been strongly related to the culture-historical approach, seeking to answer the questions ‘when’, ‘where’, and ‘what’ has happened. The origins of the regional culture-historical

Historical and ethnic issues of this sort, along with the ideals of increasing nationalism, encouraged the tracing of the spatial distributions of artefact types. The resulting archaeological methods were later understandably

15

Late Bronze Age Social Landscapes of the Southeast Balkans rejected as useless, being strongly influenced by the need to demonstrate the longevity of the nation and a longestablished shared identity (Kohl and Fawcett 1995; Trigger 2007, 215). Since eastern European archaeology has long espoused culture-historical theory, it became difficult to address issues of variability and spatial patterning outside of this traditional paradigm. Some major spatiotemporal schemes were designed based on culture-historically defined changes over long periods. However, despite its many shortcomings, culture-historical archaeology facilitates the archaeologist with a valuable toolset, namely, a method to detect change in material culture across space and time. This, in turn, allows us to define distinct entities chronologically or spatially, i.e. archaeological cultures. What culture-historical archaeology fails to address, as is well known, is how and why changes occurred, but it is difficult to deny its success in describing diachronic patterns, which is a central archaeological enterprise, regardless of one’s theoretical leanings (Shennan 2002, 11).

and time (Roberts and Vander Linden 2011). This same need has also prompted the adoption of related terms such as ‘traditions’ (Osborne 2008) ‘cultural groups’ (Willey and Phillips 1958), ‘horizons’, ‘techno-complexes’, and ‘zones’ (Cunliffe 2005; Roberts and Vander Linden 2011), each of them serving a similar role to the notion of archaeological culture, often used interchangeably or with minor embellishments. Despite its practical effectiveness, the main critique of the concept addresses the inequality between material culture and society and postulates that ethnicity in all its complexity cannot be identified based solely on groups of artefacts (Hodder 1978). Childe argued that ‘homemade’ pottery, ornaments and burial rites tend to be related to local tastes and traditions and thus are most useful for identifying ethnic groups; weaponry and technologically advanced artefacts tend to diffuse rapidly because of either trade or copying (Binford 1983a, 400; Childe 1925; 1929, 248;). This idea counts artefacts as expressions of cultural norms and these norms define the meaning of culture and the alignment between archaeological culture and human culture (Johnson 2010, 35-36). The frequent association of archaeological cultures with ethnolinguistic groups has often been pointed out as one of the most damaging influences of culture-historical archaeology on both methodological frameworks and subsequent interpretations. In more recent research and as a reaction, archaeological assemblages or archaeological cultures were somewhat distanced from the identification of social, cultural or ethnic affiliation, provoking a heated debate on the ability of archaeology to trace and recognise clear ethnic identity (Hu 2013, 372). Therefore, the term gained a reputation as a profoundly flawed anachronism, particularly in the Anglo-American literature (Roberts and Vander Linden 2011).

One of the primary outcomes that traditional culturehistorical archaeology left on the map is the complex variations of cultures across the world and specifically in Europe, as a prominent result of cultural intermingling (Harris 1968, 382-392). An essential step for European archaeology was the modification of this idea via the more specific and contingent concept of an ‘archaeological culture’ that was still often equated with perceived tribes or other ethnolinguistic groups, but perhaps now more equivocally (see Roberts and Vander Linden 2011). Cultural archaeology and archaeological culture The concept of archaeological culture is central for culture-historical archaeology. Despite the bias caused by the perceived need throughout the twentieth century to contribute to the shaping of national identities, the practical generation of Europe’s archaeological cultures, from a time-space systematics point of view, was largely successful, so that many of these constructs are still in use today. Due to the inevitable necessity to outline regional differences in archaeological assemblages, the majority of European prehistory still largely depends on the notion of ‘archaeological culture’. Borrowed from the fields of cultural anthropology and human geography (Trigger 1989, 529), in the archaeological context the term applies to an assemblage of one or more elements of material culture – usually pottery, lithics, and metal artefacts,– and/or settlement and burial types, house organisation, and rituals that continuously recur together (Harris 1994, but see also Veit 1989 and Lyman et al. 1997). An archaeological culture is also largely accepted because of its ability to accommodate new data, although without acknowledgement (Adams and Adams 1991). The universal function of the concept of archaeological culture is to allow treatment of similar material assemblages as interconnected units and to allow tracing of archaeologically identifiable patterns through space

One of the weaknesses of ‘archaeological cultures’ is that they have been treated as ‘entities’ equivalent to ‘tribes’, ‘societies’, and ‘ethnic groups’ (Shennan 1994, 11). It has been argued that such entities should be replaced with social networks, instead of societies, often overlapping and of varying scales, and related to each other through different mechanisms (see Mann 1986). Accordingly, in this study, as I use quantitative methods to study the variation distribution of different artefacts, I often discover partial overlaps and cross-cutting patterns, which disassociate ‘cultures’ from entities with restricted bounds (Shennan 1994, 13). Nevertheless, social practices and interpretative paradigms vary geographically, and the patterns of such practices change over time – variations which are archaeologically universal. Although the concept of archaeological cultures has often been demonised, it has overall robustly withstood critique and remains one of the main instruments of some archaeologies, particularly in the European prehistory. Thus, in countries with a resilient culture-historical tradition, such as Bulgaria, archaeological cultures and

16

Change and identity: a theoretical perspective cultural groups have become an accepted and rarely questioned notion. In this sense, it will be at present insufficient to analyse and interpret archaeological datasets from the study area without the employment of the concept of archaeological culture. What we need to do is not to discard the term entirely, but to re-assess and use it with caution as Roberts and Vander Linden (2011) already pointed out.

can be examined on microscopic and petrographic levels. Because ceramic production is an additive process, a pot embodies many of the choices made in the production sequence. In all this, the process of decision-making, in choosing specific techniques and not others, is of vital importance for the social aspect of our reconstruction of the past. ‘Indeed, pottery producers […] select particular courses of action based on their particular knowledge of and position within collective structures and social institutions’ (Knappett 2002, 171). Thus, pots are a choice medium for the study of technical systems, subsistence, and social interaction.

Following a course between the two positions described above, certain discussions in this study will directly or indirectly address issues concerning ethnicity and identity. These, however, need to be considered in the outlined context. Although I agree with Stark (1999) that the search for ethnicity in an archaeological context is often unproductive and plagued by modern conceptualisation, I believe the attempt to identify spatial patterns based on artefact variability in non-state societies is a reasonable approach towards understanding different levels of social boundaries.

In archaeology, we traditionally perceive social structures and activities as consisting of political, economic, and cultural components (see Knapp 1988). While these three aspects are all parts of any social phenomenon, it is difficult to analyse them together. A reliable way to understand and distinguish these processes has been defined by Knappett (2002) as ‘material relations’ (economy), ‘social relations’ (politics) and ‘social meanings’ (culture). The need for distinction comes when differentiating between processes such as pottery production, adjudged predominantly economic, and pottery consumption, which would instead be considered political or cultural, as would the use of different pottery types from cooking ware to luxurious feasting ensembles, sometimes used to explain social differentiation in early complex societies (see examples in Renfrew 1972). Using pottery, we can potentially detect aspects of cultural and social identity to illuminate networks, distribution paths, trade, migration, culture change, exchange, and innovation (see RodríguezAlegría and Graff 2012; Spataro and Villing 2009; 2015). In this respect, it is also important to stress that ceramic vessels can participate in complex processes and travel considerable distances, both as technological trends and as individual objects, containers, and commodities.

Pots as politics Regardless of whether they find ‘archaeological culture’ as a useful concept or not, archaeologists handle past societies, their material culture, and the changes in them, through the observable material record. Archaeological pottery and lithics are often the most robust elements of the material record, owing to their durability. Considerable attention has been paid to pottery as a means to study technology, establish chronology or detect cultural similarities and differences. Often the study of utilitarian pottery has been recognised as a tool for investigating social, economic and political aspects of the past. One of the main issues with this approach is ‘equating pots with people’, an argument discussed on various occasions (see Hodder 1991; Knappett 2002; Politis 2003; Trigger 2006). In this regard, a crucial issue to be addressed is the social role of artefacts, how we reconstruct aspects of past social structure from artefacts alone, and the problems caused by the analytical gap between material culture and social structure. In particular, fragmentary pottery is often considered incapable of answering ideological or political questions. That said, the production and distribution of pottery happens through the agency of people and thus transfers information within the past and from the past to the present; the condition of the ceramic vessels on their own can be a ‘witness’ of such transfers of information and the processes involved to produce fragmentary remains. In this sense, material culture emphasises the constitutive practice of artefact manufacture, use, and discard. (Chilton, 1999, 1).

Conversely, tradition and ideological values can stimulate the maintenance of a style for centuries despite innovations. As traditional behaviours often link to identity and a variety of social and cultural factors, they influence aspects of pottery production from fabric, shaping, and decoration to firing (Gosselain 2011; Gosselain and Livingstone Smith 2000; Livingstone Smith 2000). Besides, cultural and political factors can play a significant role and they, naturally, also include the movement of people, which may lead to the introduction of new foods, drinks, and serving practices along with pottery shapes and potentially new pottery techniques (see Boileau and Whitley 2010). Beyond the conservative perspective on pottery in human history, it is still possible to argue that ceramic style can be an attribute with which to explore long-term cultural traditions and more extensive spatial interactions. One of the main reasons why archaeologists favour pottery is the fact that it is usually the most abundant preserved artefact type, present at most European post-Mesolithic sites, usually stylistically datable and receptive to

To elaborate, the first observable evidence of human involvement in the pottery cycle is seen in traces of production processes, macroscopically visible mostly in forming techniques, paste preparation, surface finishing, decoration and firing; further details related to provenance

17

Late Bronze Age Social Landscapes of the Southeast Balkans different analyses concerning its morphology, technology, fabric, and decoration. A combination of at least some of these elements can associate with collective groups of people reproducing cultural traditions via a range of learning and sharing mechanisms. For example, based on fineware pottery alone, ‘changes in the strength of intersite relationships over time’ are detectable (Morgan and Whitelaw 1991, 79). Variations in fabric, style and design can be indicators of local or foreign production and can potentially associate with group identity and traditions.

groups has also hampered the study of distributional patterning in the material record (Cordell and Yannie 1991; MacEachern 1998; Shennan 1989; Stark 1998, 2). Although many archaeologists have equated stylistic boundaries with ethnicity and still do so, it has been long acknowledged that there is a significant problem with such interpretation (Cordell and Yannie 1991; Hodder 1979; Shennan 1989). The concept of ‘ethnicity’ in fact is a modern ‘evanescent situational construct’ (Shennan 1994, 13) that cannot directly reflect the past reality of especially non-state societies (Shennan 1989; Stark 1998, 10). Though such societies are commonly known to exhibit extensive cultural diversity, consistent spatiotemporal patterns in the material record can be often identified, which legitimates the search for cultural groups. Persistent questions related to long-term change in the social processes hinge on identifying such groups (Stark 1999, 25, 26).

Style and social boundaries While we have agreed so far that pots do not directly equal people, it goes without saying that we all use material culture in our research. According to some, material culture patterning is the primary domain of archaeological interpretation (Goodby 1998). Archaeologists often accept that such patterning forms ‘cultural boundaries’ which reflect past social boundaries (Parkinson 2006, 33); that there is an association between stylistic artefact variation and a certain level of social organisation, such as ‘peoples’ or ‘societies’ (Maceachern 1998, 107). Such population groups and their identification have been a persistent objective throughout the history of the field. The identification of group boundaries is usually approached by tracing distributions of key artefact types and other practices that leave physical traces (Stark 1999, 25). In European prehistoric archaeology, cultural groups and complexes and other frequently used constructs are defined primarily based on variation in ceramic shapes and types.

The strong processual and following post-processual critique on the forced association between material culture variation and ethnicity by culture-historical archaeologists resulted in serious adverse effects on modern archaeologists’ willingness and ability to understand the artefact world. Thus, the relationship between material culture and human activity was not entirely rejected, but somewhat ignored (Conkey 1989, 17). Such a limited approach towards material culture theory has also alienated researchers interested in material culture studies, besides the purely technological side of artefact analyses (Stark 1999, 25). A significant exception to this general neglect of material culture was the proliferation of archaeological theory related to style (see Conkey and Hastorf 1990). Focusing on the active role of style in artefact variability, Wobst (1977) and others explored the role of information exchange and social messaging (e.g., Braun and Plog 1982; Conkey 1978; DeBoer and Moore 1982; Hodder 1977, 1979; Lathrap 1983; Wiessner 1983).

Although the nature of social boundaries in an archaeological context is complex and considered by many to be highly contextually based (Conkey 1990; Hodder 1979, Wiessner 1983), it is my conviction that the search for those boundaries in material culture patterning is indeed ‘a productive avenue for research’ (Stark 1998, 9). I believe that this practice is especially useful when employed in the study of non-state societies, where social boundaries are flexible or ill-defined (Parkinson 2002; 2006). Combining analyses of variation in material culture with the study of settlement patterns makes the detection of boundaries more possible and strengthens the interpretation (Parsons 1972; Parkinson 2006).

Another vein of critique considers style as deprived of meaning and lacking serious foundations (see Wobst 1977). Nevertheless, the study of material culture variation and the identification of social boundaries remains a persistent archaeological interest in many different parts of the world. The value of style comes with the understanding that it reflects continuously established norms or ‘socially informed actions’ within a group displaying the ‘way things are always done’, which is considered to be resistant to change (Stark 1998, 6; Wiessner 1984, 161, 195). Thus, the study of style becomes a compelling asset in our attempts to fill the gap between the people and the record of the material culture that survived them (Parkinson 2006, 36) or in the understanding of the connection between people and the objects created by them (Chilton 1999, 2). Accordingly, style changes through time to be replaced by new ‘types’ that are either derived from earlier versions (i.e. transform types) or not associated with another type (i.e. independent types) within a ‘single cultural assemblage’ (Clarke 1968, 211). The interpretation of change related to transformed types differs significantly from the one

The study of style, and specifically technological style, and its validity as a conceptual tool for measuring social boundaries has been long debated. Traditionally, and by assumption, style has been treated as the primary apparatus for distinguishing social groups, a view which was criticised by Binford (1965) and the ‘New Archaeology’ as inadequate to identify cultural areas based on classified material. Nevertheless, tracing regional and interregional modelling based on typological relationships associated with households, population groups, regional systems or culture areas, has been the goal of many archaeologists (Dobres 1999; Stark 1998; Trigger 1989). The critique that posits that variation cannot be equated with past population

18

Change and identity: a theoretical perspective associated with the appearance of a new, independent set of types (Shennan 2004, 6).

significant consideration is how a ‘region’, which ideally would reflect a space of defined human behaviour in the past, is delineated in practice. Indeed, archaeological research is regularly restricted spatially by modern political borders or other practical constraints (Parsons 2004, 8). It is also common to initiate a study based on topographic boundaries while considering the patterning of material culture to define analytical regions, ‘often equating a spatially contiguous distribution of distinctive artefacts with a sociopolitical or sociocultural group’ (Kantner 2005; 2008, 42). Thus, the research questions, the methodology, and the tools applied depend mainly on the choice of a region and its boundaries (Johnson 1977, 499). Nevertheless, although it is nearly impossible to suppress the modern perception of an area, the controlled awareness of the restrictions caused by that perception can enhance many aspects of the research.

The identification of social boundaries comes alongside with the interpretation of homogeneity and divergence. Beyond the general modelling of geographical distributions based on stylistic attributes, stands uniformity which is often associated with communication density across space, on the one hand, and variation, which equates to discontinuity in an information exchange or sociocultural isolation, on the other (Wobst 1977). One could argue that such variation, however, might be a product of chronological inconsistency in our record. While we cannot rule out chronology as a significant factor in stylistic discrepancies, chronology alone would not be enough to explain patterning, strongly dependent on long established and conservative traditions of knowledge and practice, such as the ones associated with the production of pottery (Parkinson 2006, 52). Wobst (1977) pointed out that some artefacts correlate with socio-cultural boundaries better than others. Such studies often use pottery as a category of material culture which participates in information exchange in as many different contexts as possible.

During and after the 1990s archaeological approaches to archaeological regionalism have been affected by rapid progress in computer technology and its potential for complex spatial analysis. The adoption of Geographical Information Systems (GIS), which only became widely available during that period (Gaffney et al. 1996; Lake et al. 1998), allowed more thorough regional analysis and understanding of the restrictions involved, and encouraged better applications of tools specifically for archaeological needs (Conolly and Lake 2006; Kvamme 1989; Maschner 1996; Wheatley and Gillings 2002). Within the theoretical framework described earlier, with the help of GIS and a combination of methods to be described in the following chapter, this study has the potential to cast the character of the LBA in the southeastern corner of the Balkans in a fresh light and potentially to help reconstruct some of the processes that affected the region during the second millennium BC.

Physical environments and human landscapes Last, but not least, this theoretical review in the present context of the current study calls for recognition of some aspects to do with our perception of past landscapes and with the relationship between people and the environment – a central subject in the field of cultural ecology. Cultural ecology focuses explicitly on the complex relationship between people and the environment on a regional scale (Shennan 2002; 2015). It is sometimes argued that people and places could be seen to be connected through the creation of material identities (Chapman 1988; Chapman 2013, 183, 185; Tuan 1977). Such social landscapes are, according to Chapman, created by groups of people within a structure cohesive enough to leave material markers; otherwise, both the structure and the place would remain invisible (Chapman 2013).

Although necessary in such a research environment, culture-historical archaeology is not enough to approach more pressing topics. On the other hand, it cannot be ignored altogether either, especially while processualist studies regularly employ typologies as well as radiocarbon dating. Similarly, while cultural and demic diffusion cannot be taken as the sole explanation for change, they should be considered at least part of the cause in many European pre- and non-state societies. Naturally, different theories apply to various processes and their interpretation. I believe that a combination of approaches would be more likely to answer questions of distribution and change. Such a multifaceted approach towards cultural change, taking into account population movement, diffusion of ideas, internal cultural evolution and broader economic and political contexts influencing the need for and facilitating the use of innovations, is perhaps the only meaningful way to address a cultural system and its dynamics.

A central tenet of cultural ecology holds that adaptation to the environment provides an impetus for culture change (Steward 1955). In this, the level of human adaptation to social and physical environments is believed to enable a human population to survive in that environment, implicating cultural variation concerning geographic areas with different ecologically-based resources. Accordingly, the majority of archaeological studies are considered ‘region-sensitive’ or ‘site-sensitive’, regardless of the research topic: they implicitly or explicitly have a study area beyond which their conclusions are not meant to hold, it is often hard to choose and precisely define a study region and, even harder to justify that choice. A

19

3 Data and method

This study aims to address the LBA in the southeast Balkans in an attempt to trace spatial variability visible in material culture and to reassess some traditional cultural ideas. Because of the patchiness of archaeological research and as a result of the various challenges overshadowing most twentieth-century archaeology in the region, the character of the evidence varies significantly and therefore requires specific attention.

and far-from-perfect datasets. Because of the substantial progress in the field of archaeology, adopting, adapting and developing scientific techniques in dealing, at least partially, with such patchy datasets, it becomes feasible and also important to initiate such a study in a region that has long suffered from either pure ‘sherdology’ (Özdoğan 2003) or politically-driven avoidance of this problematic area altogether.

There are two main datasets, which can be traced over the region in line with the research questions outlined above. The first type is a relatively large sample of pottery. Most of the sites dated to the LBA have been identified through the presence of specific ceramic types, traditionally associated with the second half of the second millennium BC. Such a dataset hinders an in-depth analysis of many individual findspots but allows examination of the spatial variation in pottery distribution. The second type comprises site location data. While only geographic coordinates can be assigned to some of the ceramic material, other sites have been partially or fully excavated and based on these or other investigative methods interpreted typically as ‘settlements’, ‘sanctuaries’ or ‘cemeteries’. This research explores the distribution of such types and their relationship with the environment and combines the result with the outcome of a spatial analysis of selected pottery types aiming to address potential patterns in the regional sociocultural and economic landscape.

Analysis of pottery data Pottery is often treated as one of the most robust elements of the archaeological record. As previously discussed in chapter two, great attention has been traditionally paid to ceramics as a means to develop chronological sequences or to define archaeological cultures and culture groups. The debate on the meaning of material style and technological choice in this respect has troubled the various theoretical schools of thought in past decades (see Dietler and Herbich 1989; Dobres and Hoffman 1994; Gosselain 1994; Hegmon 1998; Latour and Lemonier 1994; Stark 1999; Stark et al. 1998;). In contrast, a different trend in archaeology examines pottery with the help of various quantitative methods and scientific tools particularly in the study of technology and provenance, but also adding a positivist note to the interpretation of the material record overall (see Binford 1983b; Renfrew 1980). The analysis of pottery data conducted in the current research is divided into two parts: i) by following a traditional course, the first part of the study groups a sample of the dataset amenable to stylistic categorisation into a structured typological scheme; ii) based on the developed classification, the second part models the distribution of spatially sensitive pottery types and amalgamates the result aiming to identify potential cumulative zones, while reducing uncertainty and thus increasing the accuracy in the outcome.

Although useful, there are many problems associated with the two-pronged dataset just outlined. The majority of the pottery is in a very fragmentary state, which obstructs a more comprehensive definition and understanding of the artefacts’ shape, style, and function. Moreover, the studied data is an aggregate of not only published ceramics but also some raw unpublished assemblages. This mixture leads to a bias in the description and interpretation derived from the variety of research and fieldwork priorities. Additionally, the site records included in this study are based on an amalgamation of exploratory approaches such as intensive or extensive survey campaigns, excavations of different character and extent, along with the locations of isolated chance finds. Such bias can, and certainly has, resulted in some artificial gaps in the archaeological record that need to be taken into account.

Classification and Typology As discussed in the previous chapter, typologies and classifications often form the core of the archaeological interpretation (Chilton 1999, 44). Typologies do serve a specific purpose and are thus designed according to that purpose. It has been argued that there are no right or wrong ways to devise a classification, and as long as the outcome relates to the questions, it does not need further justification (Adams and Adams 1991, 4-5, 8). Nevertheless, an extensive ‘typological debate’ challenges the essential epistemological difference between the

Bearing in mind these limitations, this study intends to synthesise and analyse data obtained from almost a century of archaeological work and is not alone in using uneven 21

Late Bronze Age Social Landscapes of the Southeast Balkans empiricist model of ‘best typology possible’ versus the positivist typologies constructed to assist specific research questions (Hill and Evans 1972). According to the former, typologies should include all elements recognisable in the assemblage, which often results in extensively large and overly complex classifications. This approach can be unnecessarily challenging, obstructing the analysis and bringing too much ambiguity to the outcome. The positivists’ typologies, however, although sometimes accused of being biased or self-serving, handle data in a manner directed towards individual research, which allows for the selection of a set of criteria relevant to the problems of the particular study, while omitting background noise in the data.

four levels: group, category, type, and variant. The group distinguishes between fine and course ceramics, while the shape refers to the general morphology of the vessel. The types are different interpretations of the same shape, based on specific rim or base manufacturing, silhouette, and size. A variant is usually used to address the presence or absence of handles or attachments and other secondary details. Technological markers such as different techniques used for similar types of decoration are also traced across the sample, as are the differences in the surface treatment as well as some generic colour tones. Unique one-off examples of a vessel type are typically excluded from the spatial analysis, but incorporated into the final interpretation of origin and function.

The study of LBA ceramics from the southeast Balkans is strongly dependent on the research questions, requiring a selection of criteria that can tackle issues of spatial character. Such criteria include functional and morphological markers and some preserved macroscopically observable technological indicators. The selected sample of local ceramics consists of those elements exhibiting the least uncertainty in the recording and the interpretation of the diagnostic characteristics. As a result, the original database of over 20 000 LBA pottery sherds and whole vessels was sub-sampled into a highly diagnostic set of 3094 records coming from over a hundred sites. The level of uncertainty was also measured based on the preservation, presence or absence of highly distinctive morphological features, and clearly identifiable decoration techniques. The ceramic fragments with uncertainty higher than 60 per cent were excluded from the analysis. Although downsampled as a whole, it can be argued that this is the first large-sample holistic treatment of such a dataset for this region.

The main limitation in the pottery analysis is mostly related to perceived gaps in the data. First, many subregions lack archaeological record. Such a deficiency may have resulted from the disproportionate amounts of archaeological exploration across the area combined with methodological discrepancies where work has been done, which makes comparison difficult. Second, pottery data was not always available for study, because of record loss or lack of context and provenance information along with some issues of access. Consequently, some of the pottery characteristics analysed here, combine my observations and those of other scholars, especially concerning published technological descriptions. Such shortcomings have inevitably caused a bias to be embedded in the analysis. One attempt to reduce the effects of such amalgamation of sources of interpretation was to involve predominantly morphologically-based criteria in the classification.

The classification scheme used here is based on nomenclature developed by Georgy Nekhrizov to record LBA and EIA pottery data from southern Bulgaria (Nekhrizov 2008b, 121-122). The main (sorry) criteria allow the description of some technological characteristics as well as the morphology and the decoration of each artefact. The first level of classification is based on general technological features, separating coarse from fine and semi-fineware as a group, which is also often referred to as a distinction between cooking ware, on the one hand, and tableware, on the other (see Leshtakov 1988; Nekhrizov 2005b, 23-31). To refer to coarse and fine ceramics here seems more applicable since some types discovered only in burial contexts, for example, can be assigned to neither table nor cooking ware, but could share the technological characteristics of either. In this respect, the names traditionally given to most categories or classes are often functional, based on the resemblance of contemporary or ancient shapes but are also technomorphologically rather than functionally defined. Such an approach has been selected mostly to avoid overly detailed descriptions and to invoke immediate association without any intentional implication of function. That said, the typological scheme described in chapter seven includes

One of the challenges of the study of local LBA pottery lies in the fact that the majority of the associated pottery is highly fragmentary and good stratigraphy or stable chronological frameworks are currently non-existent, all of which inhibits thorough statistical analysis. That said, there is a significant body of accumulated material, allowing us to trace distribution patterns across the area. One formal analytical approach capable of handling such data is to examine the spatial distribution of different ceramic types by constructing ‘relative risk’ surfaces (see Bevan 2012). This method maps the spatially varying ratio of the probability of an event occurring (Bevan 2012; Hazelton and Davies 2009; Kelsall and Diggle 1995), identifying the extent to which different types are spatially correlated and whether their combined distribution forms clusters of preferred pottery shapes. The idea is to assess whether spatial variation in the incidence of two or more types of events, in this case, ceramic types, is the same (Kelsall and Diggle 1995, 2335). Alternatively, the identification of areas of unusually high or low incidence of one of the types within the study region identifies possible ‘risk factors’. In other words, the analysis results in raster surfaces with an estimate of the spatial variation of each type across the

Regional zoning and spatial clustering

22

Data and method area, based on the presence or absence of each type and their relationship with the other types within the category.

algorithm with a maximum number of iterations allowed set to 100. K-means is another unsupervised clustering technique, used here to identify groups in the data without defined categories such as the aggregates produced by the relative risk, which are then split and regrouped by the RF.

The next step examines the cumulative distribution of different, seemingly unrelated pottery types, to test for a more extensive regional pattern, beyond the plotting of single distributions. The method is more experimental but extends the idea of relative risk surfaces to a multivariate case, taking the ratio of the kernel density of the observed cases to the population at risk (Bevan 2012, 500). An unsupervised clustering technique, identifying whether there is a hidden structure within the data regardless of its categorisation, is applied on a stack of relative-risk surfaces of all pottery types as if they were a multi-band raster image. The sample is then classified based on each pixel across these different bands using Random Forests (RF) classification developed by Breiman (2001). A random forest is a machine learning technique based on decision trees. A decision tree on its own, with its branches and leaves, can be employed to classify a sample based on its internal variation but is limited by the tree’s set of rules used in the classification. The forest, on the other hand, aggregates a set of decision trees and randomises the procedure via a selection of different subsets (Tang et al. 2018). Such a random selection produces an assembly of trees that can identify intricate patterns and can increase the predictive performance of the classifier (Denisko and Hoffman 2018, 1691). The technique is attractive also because it handles mixed variable types, is independent of the variations with the individual variables and is not affected by outlying observations; further, it controls a large number of variables due to its ‘intrinsic variable selection’, which assesses the weight of each variable based on its dependency of other variables (Shi and Horvath 2006).

The analysis was performed using R (https://www.rproject.org/), employing packages rgdal (Bivand et al. 2019) handling spatial data interoperability, maptools (Bivand and Lewin-Koh 2019), including various methods dealing with spatial data, spatstat (Baddeley et al. 2019) for point processing, raster (Hijmans 2019) to manipulate raster datasets, and randomForest (Liaw and Wiener 2018). The applied script adapts an original script for relative risk mapping (Bevan 2015). Site analysis ‘Site’ has been interpreted and used in different ways throughout by archaeologists and has been subjected to significant criticism (see Binford 1983a; Clarke 1972; Dunnell and Dancey 1983; Schiffer 1972, 1976, 1983;). Nevertheless, the continuation of this debate addressing a deeper understanding of the term has no application in the current study. An archaeological site here is defined mainly based on the presence or absence of archaeological material (see Anderson 1984; Wagstaff 1991) while acknowledging that this approach does not directly handle any extensive use of the site catchment area and does not address individual site boundaries. Besides their original definition as LBA findspots, some of the sites are also able to be subjected or have been already to qualitative assessment, based on the simple distinction between sites with indications of settlement or ritual activities and others indicating burial practices. Such evaluation will be carried out within the first part of the site analysis aiming to facilitate the following semi-quantitative assessment of potential occupational changes in landscape choice across the whole region while identifying and discussing perceived gaps in the archaeological distributions.

The main disadvantage of the technique is that it is strongly dependent on subsampling. Without significant subsampling, the results may be inconsistent because of the violation of diversity (Tang et al. 2018). Furthermore, RF involves many ‘driving forces’ obstructing the analysis and interpretation. It has been pointed out that the mathematical properties of the technique, along with the mechanism behind a ‘true’ RF are still mostly unknown and not fully understood (Biau 2012, 1064). Finally, a large number of trees may result in slow computing, but a more accurate prediction requires more trees. It should also be kept in mind that the technique is predictive and not descriptive (Donges 2018).

Site definition and exploratory site-based analyses The site data collection strategy was formulated based on the presence of characteristic local LBA pottery. The sample consists of 251 sites from Bulgaria and 110 relevant sites from northern Greece. Although a significant amount has been published in both Bulgarian and Greek archaeological literature, almost nothing is known about the European part of Turkey, which is the easternmost section of the study area. There have been several intensive survey campaigns in Turkish Thrace. However, it was concluded that the entire second millennium BC was underrepresented (see Özdoğan 1986; 2001; 2002; 2003). The current study also involved the personal inspection of the collected material from that area by the author thanks to the generous collaboration enabled by Mehmet Özdoğan,

The RF applied to the distribution of LBA ceramics involves the relative risk factors of individual pottery types employed as variables used for splitting. The number of trees to grow is set to 1000, with a depth of five, and the subset of 500 trees, to ensure that every input row gets predicted enough number of times securing a robust mean. Proximity is also calculated to replace missing data, to locate outliers, and to facilitate the visual output of the data. The result is then clustered using a k-means

23

Late Bronze Age Social Landscapes of the Southeast Balkans to address potential terminological discrepancies, especially in a territory with an abundance of EIA sites. During this inspection, no material comparable to the LBA incised ware from south Bulgaria, and northern Greece was identified. Consequently, no archaeological sites from Turkish Thrace can be included in the study as currently formulated.

into account. However, there are only isolated examples of each, and thus, it is difficult at this stage to incorporate them in any formal locational modelling. As a complement to the pottery distribution analysis, a regional zoning approach using ‘relative risk’ surfaces will also be applied to the site types (as for the pottery distribution), identifying the way in which different site types persist across the study area and how this relates to independent evidence arising from the pottery-type zoning. More precisely, the combined ratio-based surfaces of the site evidence will be subjected to an unsupervised Random Forest clustering and compared with the results from a similar approach to pottery to explore their relationship.

A substantial number of the sites from Bulgaria were initially discovered and recorded as a part of the ongoing project ‘Archaeological Map of Bulgaria’ (AMB, bg. ‘Археологическа Карта на България’ or АКБ), initiated in 1990 under Mieczysław Domaradzki and later Georgy Nekhrizov. As part of the strategy of the project, the results of some survey campaigns and other forms of site registration were entered in a standardised way into a database that allowed for the extraction and querying of some essential criteria. The most useful aspect of this study of site attributes was information related to the geographic location, functional site interpretation and identification of each site’s successive phases. Some of the sites have been published in a more detailed fashion, which has allowed further refinement of the record in terms of the definition of the function of the site.

Statistical location modelling Taking into account the risks associated with an uncritical drift into geographical determinism (Gaffney and Van Leusen 1996), the study of LBA site patterns in the southeast Balkans engages with technologies such as Geographical Information Systems (GIS), which whilst modelling Cartesian spaces, can also be employed in various ways to understand more extensive social landscapes (Wheatley 1993). Besides addressing relationships between people and places, GIS and various techniques of spatial analysis can also be used to track changes over time. Central to such GIS-led aims is the effort to detect if there is a pattern, and the attempt to identify how this pattern came into existence. The present research develops a spatial analytical component that builds upon assessment of ceramic distributions and focuses on three major objectives: i) qualitative evaluation of variability in the character of the LBA sites across the whole region, ii) semi-quantitative analysis of the spatial distribution of those sites, and iii) quantitative assessment of differences across space and of perceived gaps in the archaeological evidence. This strand of research falls under the umbrella of formal, computationally-led, site location modelling. Location modelling more generally addresses questions of external factors influencing the relative density or intensity of a discernible pattern of locations in space. In archaeology, the analysis of site distributions has commonly become known as ‘site predictive modelling’, ‘site location modelling’ or often just ‘predictive modelling’ (see Allen et al. 1990; Bevan et al. 2013; Crema and Eve 2014; Hudak et al. 2002; Judge and Sebastian 1988; van Leusen and Kamermans 2005; van Leusen et al. 2005; Lock and Stančič 1995; Kohler and Parker 1986; Kvamme 1989, 1990, 1995, 2006; Verhagen 2007; Verhagen and Whitely 2012). A predictive model, in turn, refers to an expression of a probabilistic relationship between social site location behaviour and pre-existing environmental conditions (Verhagen and Whitley 2012, 71), assuming that human spatial behaviour is predictable and can be revealed via statistical models. In data-driven models, some parameters or variables such as availability of geological resources, elevation, distance to water, etc., are considered to be partaking in the decision when

Equivalent information concerning sites from northern Greece has been obtained mostly thanks to the work of David French, who undertook careful cataloguing and mapping of the majority of the known archaeological sites in the area (French 1967). Other information was derived from publications, primarily based on the works of Chaido Koukouli-Chryssanthaki (1980; 1982; 1992; 2001; 2008), Stelios Andreou (2001; 2003; 2009; Andreou and Psaraki 2007), Kostas Kotsakis (Andreou and Kotsakis 1996; Andreou et al. 1996; 2001), Demetrios Grammenos, Dimitrios Matsas, Kenneth Wardle and Diamantis Triantaphyllos (1973; 1980; 1990). Based on this empirical foundation, the LBA sites were further divided into three main categories: i) settlements, ii) burial sites or cemeteries, and iii) sanctuaries or cult sites. The first two site categories have been defined based on scholarly interpretation resulting from excavations or based on other visible markers in cases of looting. The term ‘sanctuaries’, however, is traditionally assigned by Bulgarian archaeologists to rocky, steep environments with no perceptible leeway for continuous occupation (see Domaradzki 2002). Very few of these sites have produced features or artefacts, which could be related to strategically selected places for religious practice (see Leshtakov 2008). Nevertheless, this category seems to have specific characteristics, different from other archaeological sites and therefore, deserves attention regardless of its functional definition. A significant amount of surface pottery has originated from such findspots, encouraging their involvement in the analysis as a separate category, provisionally termed ‘sanctuaries’. There are a few other types of sites with relevant material, i.e. caves, mines, niches, well deposits, and enclosures that have been taken

24

Data and method choosing settlement location (Duncan and Beckman 2000; Kvamme 1983, 1985, 1990; Verhagen 2018; Warren and Asch 2000). In the study of LBA sites from the southeast Balkans, a similar data-driven approach is applied, because of the difficulties related to objectively obtaining ‘cultural’ variables that might be more appropriate to explain or identify behaviour. Such difficulty concerns ‘the scarcity of relevant data and the lack of quantifiable theoretical models’ (Verhagen 2018, 2). Nevertheless, the aim of the model at the end is not to predict site locations in areas with no current archaeological recording, but to facilitate the explanation of potential preferential factors in the perceived site clusters.

commonly assumed that if a site is not reported at that location, then it is, in fact, a non-site. Consequently, the developed global model is based on the assumption that random spots remain a good indicator for general trends and dependencies in environmentally-based correlations. One of the popular techniques employed by the predictive modelling method is logistic regression analysis (Conolly and Lake 2006; Menard 2001) or multivariate logistic regression when multiple predictors are involved. The approach allows forming a relationship between a dependent variable, site and non-site locations, and a set of independent variables. The multiple logistic regression applied here employs a set of independent variables or predictors to explain the variation in site/non-site locations across different cells of a given landscape. Similar to the more straightforward linear regression, logistic regression gives each estimator a coefficient b that measures each contribution to the variation in the dependent variable. The advantage of logistic regression over linear is that the independent variables, or predictors, can be either a mix of variable types, nominal, ordinal, interval or ratio (Conolly and Lake 2006, 183). However, the dependent variable can be only binary, i.e. handling only two values, 0 and 1, as described above.

Predictive modelling is also the subject of a substantial body of critique. The main objection concerns the level of representativeness in the data and the bias in the archaeological record, which is considered to be incapable of reflecting past human activity. The choice of environmental variables, being based on modern data that are not necessarily related to the past has been another reason for rejection, along with the scarcity or the complete lack of sociocultural variables (van Leusen and Kamermans 2005). In the model used for actual site location prediction, low-probability zones resulting from a predictive model were often ignored when surveying, thus creating a further bias instead of contributing to the evaluation of the model (Wheatley 2004). Furthermore, the location of pre-existing sites may have also influenced the human use of space (Conolly and Lake 2006, 180).

Other limitations of the predictive modelling methodology include issues of potential interdependence and significance amongst the predictor variables. A univariate exploration of each variable followed by a stepwise model selection based on Akaike Information Criterion (AIC) (see Anderson et al. 1994; Burnham and Anderson 2002) has been undertaken prior to the global modelling in an attempt to isolate the insignificant predictors and to detect potential confounders. The AIC is then employed to minimise the loss of information in a given model by identifying which model best approximates the observed data, via sequentially reducing the number of predictors and considering how much their absence from the model lessens the effectiveness of the prediction. Although suitable for archaeological research, the informationcriterion has been rarely employed (see the discussion in Crema and Eve 2014, 272). The critical issue with this logistic regression analysis is the possibility of integrating too many variables within the sample. Reducing their number, however, could result in another bias if we select the ones that might be of importance to one sub-region but do not play a role in the prehistory of another. ‘The best model is thus the one that provides the highest amount of information that we are interested in, with the lowest level of complexity and number of assumptions’ (Crema and Eve 2014, 272). Therefore, in this study, the information criterion is applied after the general fitted model has been produced, to assist in omitting redundant variables, after which the alternative models can be compared to see which is more likely to have created the observed record.

A further concern when addressing spatial patterns of site locations is the choice of their representation by a point, which is considered a simplification of a real-world entity that could otherwise be abstracted as a more complex 2d polygon footprint, a fuzzily bounded distribution of onsite artefacts or complex 3d topographic or architectural substances. The degree to which such abstraction is a problem depends on the research question requirements for the chosen analysis (Bevan et al. 2013). For a macroregional study such as this one, where the dichotomous nature of the dependent variable requires input data for site presences on one hand, and site absences on the other (put merely, grid cells on the digital elevation model and related surfaces, which are either 1 for site presence or 0 for absence), a point abstraction for each site seems appropriate. The former should consist of known archaeological locations, while the site absence category should ideally include places where archaeological sites are directly-confirmed as absent (non-sites). However, in this particular study, not all regions have been explicitly surveyed (as is true of many or most country-scale datasets), so it should be noted that the non-site sample is created using a random sampling method. The use of randomly generated non-site samples has been addressed by Kvamme (1992) and subsequently employed broadly. Because archaeological sites are considerably rare occurrences across the landscape, and it is very likely that a randomly generated non-site is, in fact, a non-site, it is

Finally, a complementary analysis of covariance (ANCOVA) compares the mean values of the dependent

25

Late Bronze Age Social Landscapes of the Southeast Balkans variable and the predictors and thus controls the effects of the covariance, in case covariates are present. As a part of ANCOVA F-test of significance is applied to test for effects, where the variance between groups is divided by the variance within groups (Huitema 2011; Randolph and Myers 2013). The approach, however, comes with many assumptions, and it is therefore employed here mainly as a corrective to the conventional logistic regression (see Menard 2001). The reconstruction that follows is based on the most reliable model comparing the global estimate, the stepwise information-criterion, and the analysis of covariance.

would not be the same as those relevant when choosing a place to bury the deceased. A model that discriminates consistently between settlements and other site types would support a much more accurate interpretation. The dataset under study, although abundant with archaeological findspots, fails to represent the variety of site types evenly, some of which are dominating, while others are underrepresented to the extent that they are too few to be analysed statistically. Nevertheless, an intensively surveyed area in the eastern Rhodope Mountains provides a sufficient basis to integrate site categories into a location model. Built into a case-study, this micro-region is also used to explore the implications of a change to a smaller scale of interpretation. This subset of analyses also experiments with a non-parametric multinomial modelling, allowing the prediction of outcomes of a categorically distributed dependent variable. Consequently, all site types from the area were differentiated but examined within the same model, which made the non-site subset unnecessary, omitting the uncertainty coming from the random non-site locations. This model was explored on a smaller scale as a potentially useful technique for future research.

Beyond the identification of sites and non-sites, our understandings of geographical and cultural patterns and processes are highly dependent on the scale (Harris 2006, 39; Lock and Molyneaux 2006). Ignoring the importance of scale and its complexities might lead to misrepresentation of the past in a variety of ways (Lock and Molyneaux 2006, 1). Changing the size and positions of the chosen spatial units, such as the sub-regional blocks used in chapter six, can result in the disappearance or emergence of statistical patterns (Harris 2006; Openshaw 1984, 18). A core problem is the choice of an appropriate scale or scales of analysis, and the sensitivity of the outcome based on that scale. One of the branches of spatial analysis in this study will, therefore, address the influence of scale on inferences across the study area and the effect that spatial heterogeneity (localised variability that undermines global conclusions) or spatial non-stationarity (see Fotheringham et al. 2000) have on the dataset. The concept of local spatial heterogeneity can be seen to relate to both natural and social properties such as the uneven distribution of resources like drinking water or access to communication routes. Similarly, arable land is not available everywhere or can already be occupied by others; ores and minerals also have limited availability. This divergence is therefore employed to reveal and interpret people’s locational choices across the landscape concerning their social, cultural or economic needs.

A crucial and well-known methodological aspect of site location modelling is that correlations between site locations and the environment do not automatically imply a causal relationship (Kohler and Parker 1986, 400, 401). Nevertheless, used carefully and with attention to issues of co-dependency among variables, location modelling can still facilitate the identification of spatial patterning. The combination of approaches described here will be synthesised in the penultimate chapter of this study, attempting to trace directions of influence and cultural change across the regional LBA. Although it cannot serve the purposes of any site-scale analysis, the methodology discussed here was specifically devised to handle a regional approach to the study of material culture and to trace large-scale spatial patterns. In this sense, it has the potential to handle the current dataset and its shortcomings. Before moving forward with the analyses, in the following chapter, I will outline the century-long history behind the generation of such a dataset.

It seems plausible to assume that different types of sites result from differently motivated choices about location. The criteria operating in the selection of settlement ground

26

4 Archaeological overview and prior research

Research on the LBA in the southeast Balkans can be traced back to the end of the nineteenth century and since then has continuously developed. Although this area remained outside the main scope of western archaeologists, it inspired local research and led to the evolution of regional archaeology. The proximity of the study area to the southern Aegean and Anatolia (Fig. 4.1), as well as its location on the way to continental Europe, has provoked discussion about migration, ethnicity, trade, and exchange. The location of this European corner spanning among three nation-states, however, has significantly obstructed the examination of such questions. This chapter offers an overview of the empirical record of archaeological finds and traces the historical development of archaeological investigation within the study area.

the European part of Turkey began in the 1930s, but the work in the area was mostly stalled after Atatürk’s death (Atakuman 2010) and was only partially restored during the 1980s and 1990s. Archaeology is born The earliest systematic archaeological interest in the southeast Balkans was associated with the recognition of tell sites (gr. τούμπα) in northern Greece at the turn of the twentieth century. A significant contribution was the distinction between the settlement sites and the burial mounds, resembling similar artificial landforms (Traeger 1901; 1902). A first attempt to classify archaeological material was published by the German Archaeological Institute in Athens (Struck 1908), based on surface collection from many tell sites. At the same time, the Bronze Age was also identified in Bulgaria with the publication of a few chance finds (Popov 1913) and the initiation of the excavations at tell Sveti Kirilovo in 1914 (Katsarov 1914). The first documented interest in the function, the date and the origin of the megalithic monuments in the Sakar also occurred at the beginning of the twentieth century (Bonchev 1901). Rafail Popov continued with the publication of isolated Bronze Age and Iron Age finds until the beginning of the First World War (Popov 1913), while his later works were dedicated to broader issues concerning the entire ‘Stone Age’ and the ‘Metal Age’ (Popov 1930). During the war, the archivist Léon Rey compiled a detailed index of tell sites located along the frontline in northern Greece (Rey 1921) - an effort which laid the foundation for post-war archaeological work in the area.

Prior research The majority of the relevant archaeological material comes from research conducted in the territory of Bulgaria, for which five separate periods of development embracing changing research agendas and varying intensity of investigation can be outlined. The first stage can be defined by the transition between antiquarianism and modern archaeology stretching from the end of the nineteenth century and to the beginning of the First World War. The inter-war period was marked by a growing body of archaeological research that laid the foundations of twentieth-century archaeology. Another phase of significant progress began slowly in the 1950s, following the intellectual vacuum caused by the Second World War. The 1980s and 1990s were years of the more rapid archaeological investigation while the beginning of the twenty-first century marked a further development of the discipline both locally while engaging with the progress of archaeological thought worldwide.

Archaeology after the First World War After the First World War, a new generation of scholars emerged in Bulgaria. After the initial investigations at tell Sveti Kirilovo, other tell sites along the middle streams of the rivers Maritsa and Tundzha attracted attention. The tell sites in northern Greece were also a subject of interest at the time when Walter Heurtley intended to study one site in each part of Macedonia to get a more comprehensive idea of the prehistory of the region (Heurtley 1939; Heurtley and Hutchinson 1926). In 1939 he published a compilation of his work, which began in 1924 and continued during most of the 1920s and 1930s (see Heurtley 1939). Excavations at the tell sites at Olynthos and Dikili Tash also started in the 1920s, (Mylonas 1929).

Relevant research on northern Greece does not entirely match the divisions of scholarship on the Bulgarian LBA, but it follows a similar trajectory. Early on, archaeological interest was mostly orientated towards the region of Macedonia, but some peripheral attention was also paid to the northern Aegean hinterland and western Thrace. The interwar period strongly defined the archaeological agenda in northern Greece, when an extensive range of research took place. After the end of the Second World War, archaeological research in the area picked up again from the early 1970s onwards, progressively increasing during the 1980s and 1990s. Archaeological works in 27

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 4.1. Map of the area with main places mentioned in the text: 1. Istanbul; 2. Vize; 3. Tekir Dağ; 4. Lüleburgaz; 5. Kırklareli; 6. Edirne; 7. Troy; 8. Eceabat; 9. Mihalich; 10. Drama (Merdzhumekya); 11. Chokoba; 12. Bikovo; 13. Ovcharitsa; 14. Nova Zagora; 15. Ezero; 16. Asenovets; 17. Karanovo; 18. Konstantsia (Asara); 19. Sveti Kirilovo; 20. Svilengrad; 21. Kush Kaya; 22. Gluhite Kamani; 23. Stambolovo; 24. Perperikon; 25. Ada Tepe; 26. Zvezdel; 27. Tatul; 28. Plazishte; 29. Stomantsi; 30. Dragoyna; 31. Dikili Tash; 32. Startsevo; 33. Zimnicea; 34. Razkopanitsa; 35. Plovdiv; 36. Trigrad; 37. Batak; 38. Potamoi; 39. Aggista; 40. Exohi; 41. Koprivlen; 42. Bresto; 43. Faia Petra; 44. Olynthos; 45. Sandanski; 46. Kamenska Chuka; 47. Orsoya; 48. Assiros; 49. Toumba Thessaloniki; 50. Kastanas; 51. Baley.

Some of the artefacts collected from the surface of the tells in Bulgaria inspired comparison with southern Aegean and Anatolian artefacts. By the 1920s, many bronzes find had already been donated to the Bulgarian National Archaeological Museum, which encouraged the archaeology professor Bogdan Filov to seek contacts between ‘Thrace and Mycenae’ (Filov 1920; 1937/1938). Simultaneously Gavril Katsarov started a discussion based on the origin of the EBA societies in Bulgaria and argued that by the sixteenth century BC, Thrace underwent a parallel development to the ‘Mycenaean world’ (Katsarov 1926, 14). At that time, more metal finds and hoards associated with the LBA were discovered, which continued to provoke Bulgarian researchers to seek connections between the sites in Bulgaria, and Troy and Mycenae (Popov 1930). A trajectory of parallelism also followed Casson’s Macedonia, Thrace and Illyria, which provided some of the very first examples of and insights into the prehistory of those areas (Casson 1926).

Vassil Mikov published all known archaeological finds and suggested several locations with potential LBA settlements (Mikov 1933). Mikov also collected material and began some small-scale excavations in the Rhodope and Sakar Mountains (Mikov 1933; 1935; 1940/1941/1942). Apart from his efforts, the period between 1933 and the end of the Second World War generally lacked archaeological research, but this was also the time of the first excavations of an LBA settlement in Bulgaria (Tsonchev 1939, 12-14). In contrast, this period was flourishing for the Turkish archaeology. From 1933 onwards, significant steps were taken towards the developing of the discipline with a few exploratory excavations and surveys concentrated in the European part of Turkey. The main motivation was to establishing prehistoric ethnic connections between Turks and Europeans, as a strategy defined during the First Turkish History Congress in 1932 (Atakuman 2010; 2012). The ‘Marmara Region Research Project’ was a significant regional research initiative, which involved surface collections as well as a few excavations in the area of Kırklareli (Atakuman 2008, 224). In 1936, on behalf of the Turkish History Institute, Arif Müfid Mansel recorded

The publication of LBA chance finds in Bulgaria continued in the following years (see Petkov 1933, 375). In 1933, 28

Archaeological overview and prior research and investigated some burial mounds in the areas of Vize, Kırklareli and Lüleburgaz, which constituted the first systematic research in this area, although not yielding any identifiable LBA material (Delemen 2001). After the death of Mustafa Kemal in 1938 and the political aftermath of the Second World War, the Turkish History Thesis was withdrawn, which pushed the discipline of archaeology into a long-term vacuum (see Atakuman 2010). During the war, Turkish archaeology went through a methodological and ideological transformation, which resulted in prioritising the studies of Anatolia, while distancing from other more remote areas (Atakuman 2008, 230).

1963; Peykov 1971) and published short inventories of some LBA and EIA settlements and isolated finds (Detev 1960; 1966; 1974). Away from the Aegean-Anatolian connection, Detev found similarities between some ceramic finds from tell Razkopanitsa and others from sites in Transylvania and considered the existence of an established link between the LBA societies on both sides of the Danube (Detev 1981). Later on, based mostly on the well deposit from Plovdiv, Katincharov suggested an indigenous development of Thrace, independent from the southern Aegean, Anatolia or central Europe. Katincharov also argued that this isolated development had started already during the MBA (Katincharov 1975, 3). This debate continued in the 1970s and the 1980s during and after the initial excavations at the settlements at Nova Zagora and Asenovets (see Kanchev 1974; 1982; 1984; 1991; Kanchev and Kancheva 1990; Katincharov 1972). A significant contribution to the discipline during this period was the first more elaborate periodisation of the Bronze Age from southern Bulgaria (Katincharov 1974; 1982), which stimulated further discussion about the connections between Bulgaria and other territories in southeastern Europe (Katincharov 1979; 1982b). A link between the southern Bulgarian LBA and the tell sites in the region of Macedonia was first suggested after the discovery of a pottery assemblage during the construction of a dam in the western part of the Rhodope Mountains (Tsonchev and Milchev 1970).

Archaeology after the Second World War After the Second World War, in the years between 1940 and 1950, very little archaeological research was conducted. Some projects of infrastructural development enabled the intensification of surface collections and rescue excavations. The first year of planned archaeological research was 1946 when Petar Detev began the excavations of tell ‘Razkopanitsa’, which lasted until 1966 and was eventually published in 1981 (Detev 1981). Razkopanitsa later became an emblematic site of the LBA in southern Bulgaria. During the 1950s the quantity of chance bronze finds increased, which, while searching for synchronisations, reintroduced the discussion of possible trade connections with Mycenae (Detev 1959; Koychev 1950; 1959, 95; Milchev 1955). Joseph Wiesner was the first western scholar interested in the LBA of southern Bulgaria. Wiesner spent some time investigating the relationships between Bulgaria and the southern Aegean, arguing that the contacts between these two areas were always strong (Wiesner 1963, 45-46).

The interest in the Thracian ethnogenesis increased significantly in the 1970s, which encouraged further research in the Rhodope (see Aladzhov 1969, 1997; Balkanski 1975; Domaradzki et al. 1976), Strandzha and Sakar Mountains, targeting the discovery and the documentation of megalithic sites as a part of the research strategy of the newly founded Thracology Institute in Sofia (Delev 1982; Fol 1982; Venedikov et al. 1976). The expeditions sought to trace the origins and the distribution of such monuments and to link them with some potential Thracian groups inhabiting the area as early as in the LBA. The ‘thracologist’ Alexander Fol focused on the reconstruction of the ‘political history of the Thracians from the end of the second and the beginning of the first millennium BC’ (Fol 1972, 88). Along with the uncertainty and the compulsory interpretation attached to this process, it also led to the recording of some LBA sites such as Gluhite Kamani, which continues to be excavated today (Delev 1982; Fol 1982; Nekhrizov et al. 2018; Venedikov et al. 1976).

In the early 1960s, Bulgarian archaeologists and historians initiated the excavation of many Bronze Age tell sites and open-air settlements, amongst which were Karanovo, Ezero, Mihalich, Razkopanitsa, Bikovo, Kazanlak, Nova Zagora and Asenovets. The period was, however, marked by the introduction of the discussion about the Thracian ethnogenesis, which later became a popular topic in Bulgarian history and archaeology. The first organised surface collections, which began in the late 1960s were also based on the ‘hunt’ for the Thracians. Interest in the archaeology of the end of the Bronze Age and the following EIA also increased significantly, when researchers were encouraged to investigate the characteristics of the ‘early Thracian society’ (Chichikova 1968; Georgiev 2014; Kanchev 1974; Todorova 1972; Velkov 1972). The discovery of the well deposit from Plovdiv (Detev 1964), which later became an eponym for an entire cultural horizon known as ‘Plovdiv-Zimnicea’, was also discovered during this period. The area around Plovdiv in southern Bulgaria was also investigated during the 1960s and the 1970s, when Petar Detev began excavations at the LBA settlement at Nebet Tepe (Botusharova 1963; Detev

The 1960s and the 1970s were also when western scholars also became more interested in the archaeology of the southeast Balkans. During the 1960s David French visited a large number of prehistoric sites in Macedonia as part of his doctoral research and published a detailed index of the surface collection and the environmental characteristics of each site (French 1967). At the same time, Bernhard Hänsel contributed significantly to the understanding of the LBA and EIA in Bulgaria by connecting the majority of

29

Late Bronze Age Social Landscapes of the Southeast Balkans the finds held in museums at that time with other contexts from Romania (Hänsel 1976). His research focus later shifted to excavations in Macedonia and more specifically tell Kastanas, where he addressed the potential regional connections from a southern perspective (Hänsel 1976; 1982a; 1982b; 1989).

At the end of the 1970s and during the 1980s some new sites were also excavated in the northern Greek part of the study area. Demetrios Grammenos worked at a burial mound cemetery in the district of Nevrokopi (Grammenos 1979), which inspired some further research into the LBA of northern Greece (Grammenos 1982). Chaido Koukouli-Chryssanthaki undertook research on the island of Thasos, which resulted in a monograph that was an essential contribution to the regional LBA and EIA (see Koukouli-Chryssanthaki 1980; 1992). KoukouliChryssanthaki also excavated a settlement at Aggista, published later as a short article in a Bulgarian periodical (Koukouli–Chrysantaki 1982). Between 1986 and 1996, a Greek-French research programme directed by Chaido Koukouli-Chryssanthaki and René Treuil continued the work at Dikili Tash, although little attention was paid to the later phases of the Bronze Age (KoukouliChryssanthaki 2008; Treuil 1992). The excavations and the publication of the internal pottery sequence from tell Kastanas and tell Assiros played a significant role in the regional LBA archaeology (see Hochstetter 1984; Wardle 1982; 1986; 1997; 1997a; 2005; Wardle et al. 2007). This study coincided with the excavations on Thasos, which in combination offered a solid ground for comparison between the LBA assemblages discovered in northern Greece on the one hand, and southern Bulgaria, on the other (see Grammenos 1982; Koukouli-Chryssanthaki 1982; 1992).

From 1961 to 1975, Jean Deshayes and Dimitrios Theocharis initiated the first systematic excavations at Dikili Tash under the auspices of the French School at Athens and the Archaeological Society of Athens. Their primary objective was to determine the stratigraphic and chronological sequence of the Neolithic and the Bronze Age in the region, which was poorly understood at that point. Since the 1960s, a few other excavations have also explored some Bronze and Iron Age settlements (Elster and Renfrew 2003; Renfrew et al. 1986; Romiopoulou 1971). At the same time, Ken Wardle, who focused mostly on the western Macedonian Bronze Age began the excavations of toumba Assiros (Wardle 1977). The ‘Archaeological Enlightenment’ Towards the end of the 1970s and during the 1980s and the 1990s, the number of investigations increased dramatically in all three countries. During that period, the first excavations in the Rhodopes (Kissyov 1988, 55; 1998, 2123; Kulov 1991, 73; Panayotov 1981;), Sakar (Aladzhov and Balabanyan 1984, 185-235; Venedikov et al. 1982, 37-43), and Strandzha Mountains as well as investigations along the Black Sea coast (Orachev 1988, 353-362; Porozhanov 1988; 1989) took place. Archaeological work in the Upper Thracian Plain also intensified (Gergova 1986, 13; Kanchev 1984, 134-159).

This was also a period when numerous surface surveys took place. Kostas Kotsakis and Stelios Andreou completed a survey in the Langadas basin in central Macedonia, which was another study partially related to the study area (see Andreou et al. 1996; 2001). Andreou, Fotiadis and Kotsakis published a synthetic study of some LBA sites from northern Greece (Andreou et al. 1996). The number of surface survey campaigns and testpit excavations in the Sakar and the Rhodope Mountains increased from the start of the 1990s, breaking from the existing pattern of investigations only into dolmens, resulting in the discovery of a few LBA settlements (Stoyanov and Nikov 1997). Research in the Rhodope Mountains has been especially intensive since the late 1980s, beginning with several surface collections and test-pit excavations of LBA sites (Bonev 1988; Domaradzki 1986; Domarazki 1995; 1999; Georgieva 1991; Gergova 1990; Gotsev 1990; 1994; 1995; Kissyov 1988; 1988; Kulov 1991; Kulov 1999; Panayotov and Valcheva 1989; Shalganova and Gotzev 1995, 334335; Valchanova 1999, 150-156). Noteworthy are the test-pit excavations at Tsrancha, Alada, and Bosilkovo (Domaradzki 1986; 1986a; Georgieva 1982; Leshtakov 1990; Raduncheva 1998). Following these surveys, Tsvetana Dremsizova-Nelchinova, who worked in the 1980s mostly in the eastern Rhodope range, excavated and published the LBA site of Visherad (DremsizovaNelchinova 1984). The first detailed articles about LBA pottery from the eastern Rhodopes also appeared during this period (Leshtakov 1990; Nekhrizov 1995).

Since the start of the 1980s, a significant body of planned research began along the Struma Valley and in the Rhodope, Strandzha, and Sakar Mountains. Because of intensive infrastructure development initiated in the 1990s, a substantial amount of newly discovered sites required rescue excavations. The project ‘Maritsa-Iztok’ (Georgieva 1991, 91-104), the ‘Trakiya’ motorway and the ‘Plovdiv-Svilengrad’ railway were the main sources of archaeological data. A large number of occupations from the end of the second millennium and the beginning of the Early Iron Age were discovered along the course of ‘Maritsa-Iztok’ (see Kancheva-Ruseva 1991; Nikov 1994; Savatinov 1995). One of the preplanned excavations in the area, independent of the infrastructure projects, were those at tell Drama-Merdzhumekya (Fol et al. 1990; Lichardus et al. 2001; 2003; 2004). What became clear after a couple of decades of excavation of tell sites in the Upper Thracian Plain was that the majority of the LBA material was discovered at the very top of the tells without preserved stratigraphy or a recognised structure (Lichardus et al. 2001). Georgieva (1985) first reported LBA material from tell Dyadovo, and also discovered and excavated Plazishte in the area of the eastern Rhodopes (Georgieva 1985).

30

Archaeological overview and prior research Surface collections were reinitiated in the European part of Turkey in 1993 when an international team lead by Hermann Parzinger surveyed a large portion of the area. Besides the documentation of many Neolithic and Chalcolithic sites, one of the most significant results of this project was the discovery and the partial excavation of an EBA site by Kanlıgeçit (Özdoğan 2005; Özdoğan and Parzinger 2012). The general outcome was the realisation that there is a significant lack of archaeological material dated after the end of the third millennium BC and that no new settlements in the vicinity of Kırklareli were established during the second millennium BC (Özdoğan 2001, 62). In 1980, during the Kırklareli survey campaign, a rescue excavation at ‘Taşlıcabayır’ produced fifty-one fluted vessels initially associated with the LBA, but later related to Troy VIIb2 and the Balkan EIA (see Özdoğan 1987; 2001). Between 1980 and 1985 the team surveyed the European parts of the project, including the districts of Istanbul, Edirne, Kırklareli, Tekirdağ, Gelibolu and Ecebat, resulting in the discovery of 300 prehistoric sites, 100 sites of later periods, 800 tumuli and 90 megalithic structures (Özdoğan 2001,284). After the study seasons, which took place between 1989 and 1997 (see Özdoğan 1984; 1985; 1987; 1993; 1998), it was confirmed that there was no site definitively dated in the second millennium BC (Özdoğan 2001,287).

and to feed data into the system. After Domaradzki’s death in 1998, the project continued under the leadership of Georgy Nekhrizov (Domaradzki 1994; Nekhrizov 2014). As part of the AMB project, Nekhrizov in collaboration with Georgy Kulov conducted numerous intensive as well as extensive surface surveys, which resulted in the discovery of many archaeological sites such as Stomantsi, Zvezdel, and Perperikon in the eastern Rhodopes (Kulov 1991; 1999; Kulov et al. 1990; Ovcharov and Kulov 1990). Substantial efforts were made towards developing a structured chronology and periodisation of the Bronze Age (Panayotov 1975; 1981; 1985; 1988; 1989). Panayotov compiled all known bronze weapons, which he dated to the second half of the second millennium BC and based on their synchronisation with neighbouring regions defined several chronological horizons (Panayotov 1978; 1980; 1985). Herman Parzinger later attempted to redefine the periodisation of the Bulgarian prehistory (Parzinger 1993). The most significant contribution during this period was the introduction of radiocarbon dating to the Bulgarian archaeology, which became a part of the work of Yavor Boyadzhiev (see Boyadzhiev 1986; 1994; 1995; 2003). Boyadzhiev made significant contributions to the chronological understanding of the prehistory in Bulgaria and promoted the adoption of such scientific methods to the local archaeological society, although the number of relevant samples was not sufficient to facilitate the development of the chronological scheme of the LBA.

The search for the Thracian ethnogenesis continued and intensified particularly after 1981 with the rise of the national identity propaganda, initiated by the project ‘1300 Years Bulgaria’. Different studies focused on proving the longevity of the ‘Thracian ethnicity’, targeting specifically the LBA and the EIA archaeology, based on the work of the Russian archaeologist Tatyana Zlatkovskaya, who studied the ethnic processes in Thrace at the end of the second and beginning of the first millennium BC (Zlatkovskaya 1964). The search for the Thracians in the 1980s also resulted in the deliberate chronological, but also the spatial distribution of topics and areas of interest among archaeologists. Kostadin Kissyov’s work in the western Rhodopes as a part of this division resulted in the discovery and the excavation of numerous LBA burial mounds (Kissyov 1985; 1988a; 1990a; 1991; 1993a). Kissyov also attempted a thorough definition of the settlements’ characteristics in the middle part of the Rhodope Mountains (1988b; 1990b; 1998).

Discussion of interregional connections was not ignored in this period. In the western part of the Rhodope Mountains, Hristina Valchanova excavated an LBA burial mound near Trigrad, which she associated directly with Mycenae (Valchanova 1982; 1986; 1999). Towards the end of the 1980s the material related to the end of the Bronze Age had increased in quantity and variability, which motivated Alexandar Bonev to publish the synthesis ‘Thrace and the Aegean World in the Second Half of the Second Millennium’, incorporating a large number of the discoveries to date (Bonev 1988). Bonev was the first to attempt to analyse and interpret on a larger scale the LBA dynamics in the area, in the context of the southern Aegean and the Carpathian area (Bonev 1988). A few years later, he attempted to connect southern Bulgaria with Troy (Bonev 2003), as also had Plamen Georgiev (1981). In the same direction was Bonev’s work based on the discovery of the Valchitran treasure, in which he discussed the origin and the date of deposition of the find and defined the cultural horizon called ‘Mycenae-Valchitran-Borodino’ (Bonev 1983; 1990).

The 1990s were marked by the establishment and development of the Bulgarian information system known as the ‘Archaeological Map of Bulgaria’ (AMB), designed and initially administered by Mecislav Domaradzki. The project aimed to record all known archaeological sites in Bulgaria in a standardised system with established criteria and a map-based approach. Furthermore, Domaradzki intended to survey and potentially discover new sites in the same way and to encourage other researchers to contribute to creating a complete database of known Bulgarian archaeological sites. The entire decade was devoted primarily to the attempt to cover large areas of the country

A new contribution to the problems of the LBA in Bulgaria was the work of Lam Thi My Dzung, which was based on the definition of the Baley-Orsoya and RazkopanitsaAsenovets cultural groups (Dzung 1982; 1984; 1989). As a broader study, Ana Yotsova (1989) focused on the settlement system in the LBA and EIA. Similarly, the Yugoslavian archaeologist Milutin Garašanin discussed

31

Late Bronze Age Social Landscapes of the Southeast Balkans some of the problems related to the chronology and the history of the ‘southern Thracians’ comparing them to the ‘southern Macedonians’ (Garašanin 1983). In this general discussion, Diana Gergova, while focusing on the specific characteristics of the LBA burial customs from Bulgaria, also tried to define some general cultural features of the local LBA (Gergova 1995; Gergova-Domaradzka 1989). The original work of Evgeniy Chernykh on the metallurgy and the metallurgical provinces covering the Bulgarian lands in the LBA was essential for the ‘international’ debate was (Chernykh 1971; 1978). Finally, Jan Bouzek, addressed the regional context of the Thracian LBA and focused on issues of migration (Bouzek 1973) and the relations between the Aegean, Anatolia and Europe, with specific emphasis on the Balkans (Bouzek 1985; 1994; 1996).

the unpublished PhD dissertation of Borislav Borislavov concerning the transitional Period between the LBA and the EIA in southern Bulgaria (Borislavov 1999). A study devoted to the LBA and EIA pottery from the Rhodope Mountains was the topic of a PhD dissertation defended by Georgy Nekhrizov in 2005 (Nekhrizov 2005b). It was based on a large number of sites, among which the first excavations at Ada Tepe. Nekhrizov’s work in the area of the eastern Rhodopes continues until today with excavations at Gluhite Kamani and Stambolovo along with numerous surface collections. The work at Ada Tepe continued with the excavations of the gold mines and the associated LBA settlement structures located on the slopes of the Ada Tepe peak (see Jockenhövel and Popov 2007; 2008; 2012; Popov and Jockenhövel 2011; Popov and Nikov 2018; Popov et al. 2011a; 2011b; 2015; 2017). Simultaneously Popov began excavations at another LBA and EIA settlement at Kush Kaya also located in the eastern Rhodope (Popov 2009; 2016; Popov and Iliev 2007; 2008). The work of Krassimir Leshtakov continued as a part of a larger project excavating the multi-period sites at Perperikon and Tatul, which, amongst others, yielded a volume of representative pottery dated to different stages of the LBA. The settlement at Dragoyna located on the outskirts of the Rhodope area has been excavated since 2007 (Bozhinova 2007a; Bozhinova 2014; Bozhinova and Andonova 2008; 2009; 2010;). The medieval fortress at Lyutitsa was also the subject of an archaeological investigation, which resulted in the discovery of an earlier settlement from the end of the second millennium BC, published Yana Dimitrova (2009). In the western part of the Rhodopes, a group of LBA burial mounds was destroyed by looters in the early 2000s and later excavated by a team led by Borislav Borislavov (Borislavov 2008; Borislavov and Ivanova 2007; 2008; Delchev et al. 2005; Leshtakov 2008). Borislavov also focused his research on an LBA and EIA sanctuary in the Sakar, near Dositeevo (Borislavov 1999; 2007).

Archaeology in the twenty-first century A vast number of archaeological sites were discovered and partially excavated as a result of the increasing development of infrastructure in Bulgaria, beginning with the late 1990s. This type of work intensified after the acceptance of Bulgaria into the European Union, and it is still ongoing as a part of different European programmes for regional development. A large number of sites were explored only through surface collections, while others were also excavated. More material became available because of the increased looting, which became common in Bulgaria after the collapse of the totalitarian regime in 1989. This emphasis on archaeology attempting to save the majority of endangered sites resulted in more excavations, more survey campaigns and finally in an increase in the number of recorded sites of all periods. However, this boom in archaeological fieldwork reduced the time, and the possibility for detailed documentation and comprehensive publication, which was often was reduced to the required annual reports. Nevertheless, several significant works are influencing the development of local archaeological research. Krassimir Leshtakov’s doctoral thesis written in the 1990s and published a decade later, compared many southern Bulgarian Bronze Age sites chronologically, building upon a ‘comparative stratigraphy’. Although the main focus in his work was based on EBA dissemination, Leshtakov also incorporated some LBA pottery and attempted a reconstruction of the settlement system as a comparison to the patterns of earlier Bronze Age periods (see Leshtakov 2006).

Outside of the Rhodope Mountains fieldwork continued along the construction zone of the ‘Trakiya’ motorway in the Upper Thracian Plain and the ‘Struma’ motorway along the Struma Valley in southwestern Bulgaria. Several surface collections associated with the early phase of the investigations resulted in the discovery of numerous LBA settlements in both areas (see Athanassov et al. 2014; Leshtakov 2010). Several LBA or multiperiod settlements were discovered as a result of the works associated with ‘Trakiya’ such as the sites near Chokoba, Zavoy, and Shihanov Bryag (see Bakardzhiev 2010; Ignatov and Kancheva-Ruseva 2007; Kancheva-Ruseva et al. 2008; Leshtakov 2010; 2011). Besides the documentation of LBA sites along the valley of Struma, the excavation of two significant settlements, Kamenska Chuka (Stefanovich and Bankoff 1998) and Bresto (Athanassov et al. 2013; 2014; 2016a; 2016b; 2018), have opened a discussion about the connectivity in and from the Central Balkans (see Athanassov 2011; Athanssov and Krauß 2015). Some

After Alexander Bonev’s (1988) synthesis of the LBA in Bulgaria, there was no other comprehensive research on the subject until the posthumous publication of a manuscript he was working on before his death (see Bonev 2003). Further contribution to the discipline and more specifically to the problems of the LBA were some general articles by Krassimir Leshtakov addressing the entire Bronze Age in ‘Upper Thrace’ (Leshtakov 2000; 2002; 2006), as well as

32

Archaeological overview and prior research internal and regional dynamics are discussed by the author of this book in a summary publication preceding the final results of the current study (Nenova 2018). There are also recent contributions in the fields of the mortuary practices (Leshtakov 2011) and the chronology of the entire Bronze Age in the area (Leshtakov and Tsirtsoni 2016). The growing body of LBA ceramic material has also provoked the publication of various assemblages and started a dialogue based on different pottery aspects (see Budakova 2017; Horejs 2015; Leshtakov 2017; Nikov 2016) and a detailed typology of the bronze military artefacts from the entirety of Bulgaria in a broader regional context (Leshtakov 2011; 2015).

the Austrian Archaeological Institute in Vienna (Haag et al. 2016). Some general works also discussed the ‘early contacts’ between Thrace on the one hand, and the Aegean and Anatolia, on the other (Leshtakov 2007; 2009; Pavúk 2017; Stoyanov 2004; 2004).

In the part of the study area belonging to northern Greece, a substantial contribution was the discovery, the excavation, and the publication of a tumulus cemetery at Faia Petra near Serres (Valla 2007). Also Of some relevance to the current analysis are the ongoing excavations and the study of material, at Toumba Thessalonikis. A number of articles and a few dissertations on the pottery from Toumba shed light on the central Macedonian handmade pottery tradition and innovation with some respect being paid also to neighbouring areas (Andreou 2001; 2003; 2009; Andreou and Kotsakis 1996; Andreou and Psaraki 2007; Andreou et al. 2001; Kiriatzi and Andreou 2016; Kiriatzi et al. 1997; 2000; Psaraki 2004). The work associated with the later periods at Dikili Tash was also recently renewed (Treuil 1992).1 In 2007 Barbara Horejs published a study on the handmade pottery from Agios Mamas (Olynthus), which became a significant reference publication for the local LBA pottery. Particularly valuable for current research is Horejs’ analysis of the pottery set in a broader regional context reaching the Lower Danube (see Horejs 2007).

Problems of interpretation

The result of the last twenty years of archaeological work in the area is that there is a considerable body of sites and material evidence, which continues to grow and opens questions of regional importance. With more publications becoming internationally available, topics of characterisation, connectivity, and segregation have set a new environment for even more vibrant discussions.

Besides the challenges caused by the modern nation-state partition of a geographically united study area, there are other methodological and theoretical issues. One of the central problems is the absence of an excavated stratified LBA settlement. Most of the material is studied and interpreted on a micro-regional level, while the local chronology has been derived from synchronised sites in adjacent territories. The attempt to create parallel sequences resulted in the creation of broad cultural horizons with even more significant spatial and temporal gaps. The hope for an undiscovered ‘stratified’ site, analogous to the tell sites in Macedonia, which upon discovery would solve many of the problems of the local LBA, has limited the motivation for regional or in-depth analysis facing nearly a century of research. A second obstacle is the shortage of radiocarbon dates for this period, which would potentially allow for the tracing of some chronologically distinct patterns and material sequences. Furthermore, the strong culture-historical tradition of the Bulgarian archaeology has the creation of numerous typologies as a final goal of the archaeological research and as such gets trapped into ‘sherdology’ instead of contributing to or introducing a more in-depth discussion. Material analogies have been used mostly to define or refine local chronology or to argue that trade mechanisms integrated the local LBA societies into wider regional patterns. This overall deficit of thematic research has usually resulted in short publications which, although numerous, are often inaccessible to the broader archaeological scholarship, because of the language barrier. Consequently, and in a global context, the map of the southeast Balkans is habitually left blank on the background of general patterns and distributions suggesting connections between the southern Aegean and Anatolia on the one hand, and the Carpathian Basin and central Europe, on the other.

The argument about interconnectivity with the southern Aegean and Macedonia continued after the excavations of the settlement at Koprivlen and the cemetery at Sandanski, based on the discovery of a few imported pottery fragments (Alexandrov 1998; 2002; 2005; Alexandrov et al. 2007). The discussion was supported by the results from the excavations at Bresto and Kamenska Chuka, which produced some of the few scientifically based dates for the LBA in the area and allowed the timeframe of the potential interaction to be narrowed (Athanassov 2011; Athanassov and Krauβ 2015; Athanassov et al. 2013; Jordanova and Kovacheva 1998; Stefanovich and Bankoff 1998;). A few imported sherds of Mycenaean type were also found at Dragoyna, securing another chronological fixed point and extending the suggested contact to areas beyond southwestern Bulgaria (Bozhinova et al. 2013; Jung et al. 2017). Furthermore, the site at Ada Tepe initiated a dispute, based on the discovery of LBA exploitation of the Ada Tepe gold mines, which inspired a conference dedicated to ‘The First Gold’ organised by

Another limitation of most LBA studies in the area is working with a combination of primary artefacts and also secondary, legacy data from old excavations, which often lacks essential contextual information. The varying degree and type of investigation is also a challenge. As

1 The pottery from Dikili Tash is not included in the overall analysis provided here.

33

Late Bronze Age Social Landscapes of the Southeast Balkans previously mentioned, between the Balkan Mountains and the Aegean, there is no substantially excavated settlement which could offer sufficient contextual data. The most common investigations have been test-pit excavations, but most of the site data come from surface collections and surveys of varying intensity, bringing further uncertainty in the interpretation. This is closely related to the criteria used to define a site as a settlement or a sanctuary and the fact that not many of those have been explored via systematic excavations; a large number have been registered and identified as cult places based on different external indicators, like location and the existence of rock cuts (Borislavov 1999, 154). It is also difficult to distinguish the so-called ‘peak sanctuaries’ from settlements located on peaks (see Bozhinova 2018). Through the increasing amount of archaeological material and observations, the function of many sites interpreted as settlements has been reconsidered (Kissyov 1990b, 67). Most emblematic are the examples from the sites near Startsevo (Valchanova 1986, 67), Plazishte (Georgieva 1985, 66), and Nebettepe (Detev 1960, 51). The widespread opinion, prevalent in the 1970s, concerning the sudden appearance of the ‘Thracian fortresses’ in the EIA, has been revised (Balkanski 1977, 131-134; Chichikova 1976, 15-17; Panayotov 1977, 4758; Petrova 1975, 89-98), and they were later identified as sanctuaries deriving from the LBA (Kissyov 1990b, 6474). After that, the origin, the function, and the distribution of these sites became one of the main subjects of discussion in the related schools of archaeology in Bulgaria.

of archaeological sites but also erase the stratigraphic information associated with such thin-layered occupations. The absence of LBA sites from the European part of Turkey today seems to be a matter of interpretation instead of archaeological recovery or terminology. Furthermore, the problem of recovery also exists in the border area along mostly the lower riversides of Maritsa but also creating a wide corridor on the Greek side of the GreekBulgarian border. Equally barren is the Turkish-Bulgarian border zone. Politically sensitive areas like these have resulted in a significant gap in the archaeological record and continue to discourage archaeological field work today. The limited research in the Greek part of Thrace has never systematically approached the border with Bulgaria. Nevertheless, unlike the Turkish portion of the study area, in northern Greece, there are a relatively large number of sites, discovered mostly during surface surveys, but the general level of investigation is not comparable to that in Bulgaria. Last, but not least, the part of the Greek-Bulgarian border that cuts the Rhodope Mountains end-to-end further obstructs the possibility for research and addressing questions of the perceived change between the mountainous and the coastal cultural environments. In ending this chapter with a set of problems arising out of existing research priorities and data gaps, the goal of the remaining part of the study is not to postpone interpretation or neglect existing, although incomplete, datasets but, despite the complex conditions of this cross-border archaeology, to propose one possibility for meaningful examination. Based on the vast body of accumulated data and keeping in mind all the shortcomings discussed in this chapter the following chapters will seek to present and discuss some aspects of the LBA from the southeast Balkans in greater details.

The extent of site preservation is also a concern. Very often, the LBA layer is very close to the surface and therefore was damaged during ploughing or was affected by erosion. Tell sites and open-air settlements, if not affected by disturbances from later periods, have been largely destroyed by modern agricultural activities or military trenches. Such destructions increase the visibility

34

5 Site characterisation and issues of preservation

Settlement and landscape approaches in archaeology have become increasingly important as a means to recognise, understand and explain cultural behaviour at any regional or sub-regional scale. A central aspect of this research involves the characterisation of individual LBA archaeological sites, building up towards a spatial modelling of the observed variation in site characteristics and potential landscape patterning.

1990; Leshtakov 2009; 2010; 2011; 2012; Nekhrizov 1995; 1996; 2003a; 2003b; 2008; Nekhrizov and Iliev 2007; 2008; Nekrizov and Ivanova 2008; Nekrizov and Mikov 2001; 2002; Nekhrizov and Tsvetkova 2010; Nikov et al. 2016; Popov 2007; 2008; 2009; Popov and Iliev 2007; 2008; Popov and Nikov 2014). A more synthetic study covering a significant body of site-scale data was completed as a part of a PhD thesis discussing mortuary practices in the western Rhodopes region (Kissyov 1993a). Other equally important works were the attempts to summarise the existing LBA record from the eastern Balkans, albeit with a predominantly ceramic focus (Bonev 1998; 2003; Hänsel 1976; Horejs 2007). The nearly complete site record involved in the current study contains a large amount of unpublished survey data and information stored in the AKB repository database, providing limited methodological information. It is thus hard to identify which data gaps are real gaps in the social landscape and which are the result of bias based on visibility, areal discrimination or research agendas.

Archaeological recovery A key point to stress as part of this analysis is that there has been significant variation in the intensity of archaeological investigation across different parts of the study area, which has also led to large gaps in the data. For example, for the region around Nova Zagora, there are a number of well-known and partially investigated settlements as a result of several long-term infrastructure projects during the 1990s and 2000s, but in the area enclosed by Burgas, the Strandzha Mountains, the Balkan Mountains, and the valley of Tundzha, there is almost no information (Fig. 5.1). Archaeological research along the lower Maritsa Valley was always problematic, because of its strategic geopolitical value, standing inbetween Bulgaria, Greece, and Turkey. The southwestern Black Sea coast is also lacking LBA sites, traditionally explained by the suspected submergence of settlements. Work in northern Greece during the twentieth century has concentrated more on Chalkidiki and the administrative district of Central Macedonia offering well-stratified sites and a variety of imported goods, while considerably less attention was paid to the area currently referred to as western Thrace. Furthermore, the militarised protection of national borders has resulted in two wide ‘belts’ with no associated archaeological investigations. More detailed information comes from the publications of the cemeteries at Faia Petra (Valla 2000; 2002; 2007), Exohi and Potamoi (Grammenos 1979), Kastri and Kentria on Thasos (Koukouli-Chryssanthaki 1992), as well as the preliminary report on the settlement at Stathmos Aggista (Koukouli-Chryssanthaki 1980), the upper layer at Dikili Tash (Koukouli-Chryssanthaki 2008) and the work of David French (1967). Previous research on the Bulgarian side has involved the publication of single sites or single finds and has mainly been disseminated via annual fieldwork reports (see Borislavov 2007; Borislavov and Doumanov 2008; Borislavov and Ivanova 2007; 2008; Bozhinova 2007a; Bozhinova and Andonova 2008; 2009; 2010; Kissyov 1993a; 1993b; 1995; 1999; Leshtakov

One area, however, which has been intensively and thoroughly surveyed in the last few decades is the European part of Turkey, which led to the inclusion of this sub-region in the general analysis. An investigation of the available sites and material from Turkish Thrace was conducted during 2012 by the author of this book, following the discovery of a large number of archaeological sites during the 1980s and 1990s (see Özdoğan 1993; 2002; 2003). Nothing was previously published in terms of pottery for comparison, apart from the information provided by Mehmet Özdoğan that the second millennium BC is not represented in Turkish Thrace (Özdoğan 2003). Along the Thracian coast of the Marmara Sea there is a homogenous distribution of ‘Anatolian’ type of earlier Bronze Age site formations compared to Troy I and Troy II-V, mostly on the Gelibolu Peninsula, characterised by ‘small, fortified settlements, the use of potter’s wheel, the emergence of an elite ruling class, increased number of status objects, technologically developed metallurgy, etc.’ (Özdoğan 2003, 108, 111). It seems that there is no visible evidence of occupation along the European Marmara coast all the way until the Bosporus and including the entirety of inner Thrace during the second millennium BC. This gap appeared artificial, because of the modern political boundaries seemingly dividing the LBA (Fig. 5.2). However, it was confirmed after a study season performed with the kind permission and collaboration of Mehmet Özdoğan at the University of Istanbul that none of the 35

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 5.1. Map of the area showing the main places mentioned in the text.

datable pottery finds could be confidently associated with the comparable LBA from the rest of the study area and was predominantly of earlier EBA or later EIA character. A discussion of this apparent absence will be offered later in this book, as a part of a broader regional and microregional reconstruction.

the Bronze Age, social stratification became more pronounced with consequences affecting the development of the settlement system (Popov 2015, 110). Nevertheless, in-depth information about the internal structure of settlements is almost non-existent because of looting, agricultural destruction or lack of excavation. Therefore, the only possible reading of the settlement patterning in Thrace during the LBA at the moment largely depends on a more general typology leading towards an assessment of micro-regional synchronous and potentially internally diachronic variation. Such general classification involves three distinct types. The first group comprises of settlement mounds, often referred to as tells, tell sites or toumbas (n=76). These are multi-layered settlements usually manifested as an artificial convex landform visible above ground, consisting of accumulated anthropogenic debris, such as mudbrick residue. Many of these sites typically began life in earlier prehistory, usually containing an Early Neolithic component and being continuously inhabited or with interruptions until the end of the EBA. Often, there is evidence of LBA occupation in the upper or topmost levels of such tell sites, although typically lacking stratigraphy and sometimes mixed with later material.

Site classification The site data for this research consists of all identifiable site locations across the study area. This database comprises 361 sites, of which 251 come from Bulgaria and 110 from northern Greece. Two distinct site categories predominate and are therefore subjectable to a more formal examination: i) settlements (n=222) and ii) cemeteries (n=42). A third group known in the Bulgarian literature as ‘sanctuaries’ (n=36) is also included. A number of rarely occurring site types, which because of their scarcity could not take part in the statistical analysis, are nevertheless described in the sub-sections to follow. Settlements The tradition to study settlement distribution across various periods in southern Bulgaria and northern Greece is almost a century old, but there is no work dedicated to LBA settlement patterns, beyond isolated micro-regional studies. According to some, during the late stages of

Besides tells, there is also a second group of typical settlements defined in the literature as ‘open’ settlements (n=120). This type is traditionally described as a settlement 36

Site characterisation and issues of preservation

Figure 5.2. Distribution of LBA sites in western and northern Thrace, and Macedonia in comparison to the distribution of sites in eastern Thrace.

according to which there are three types of cemeteries: i) ‘flat’ cemeteries, consisting of buried underground pits; ii) single burial mounds, and iii) cemeteries of clustered tumuli. The most common type is the burial mound found either isolated in the landscape or one part of a group of several mounds, which presumably accumulated over time.

with a shorter life span (usually embracing one or two chronological phases, although multi-period examples exist), that has not resulted in any noticeable mound formation. A third type of settlement worth distinguishing covers settlements with evidence for monumental stone architecture (n=14), otherwise unusual for the area. Their isolated distribution, as demonstrated in chapter six, may suggest specific functional characteristics and time-span of existence (see Stefanovich and Bankoff 1998; Stefanovich and Kulov 2005).

The construction of the mound follows the generic ‘kurgan’ form, with a small pit in the ground or directly lying on the surface covered by an earthen pile. Preserved evidence of ritual indicates mostly cremation in an urn within the tumulus, but inhumations are also known (see Kissyov 1993a; 2009). Sometimes burial mounds cover built or carved structures known as ‘dolmens’ or ‘cistgraves’ (Fig. 5.7), which constitute another type of burial, whose earliest use in the area can be traced back to the LBA (n=6). All structures of this type are constructed out of local grey gneiss or schist. Sometimes it is difficult to distinguish between types if the structure was looted and the cover or the front stone was destroyed. Although the dolmens in Thrace are mostly dated to the EIA (see a discussion in Nekhrizov 2015), there is a suggestion that their use started sometime during the LBA. Nevertheless, despite the LBA material associated with some of these structures from the eastern Rhodopes and Sakar, the

Cemeteries The characterisation of specific features of burial is another complicated task in the current context. Many sites have been recognised as burials or cemeteries and documented as such based on looting activities, which allowed the locations to be identified. However, because a result of the looting, often a large portion of the burials and their belongings were already gone, which prevents an interpretation beyond a general dating of the structures. Nevertheless, a few trends can be categorised and further discussed. The general classification used is based on burial method, i.e. inhumation or cremation, and burial structure,

37

Late Bronze Age Social Landscapes of the Southeast Balkans Sanctuaries

possibility of interpreting them with confidence is limited, because, as stable and visible constructions, they have been both reused and robbed on multiple occasions.

The site category labelled as sanctuaries, as discussed above, is problematic. Nevertheless, their seemingly isolated distribution suggests a potentially meaningful pattern regardless of what the precise site function might be (Fig. 5.3).

The category of flat cemeteries is usually associated with inhumation and only rarely with cremation. The inhumation ritual is similar to that attested during the EBA. The body is laid out in either a flexed or straight position with varying grave orientations. The depth of the pit varies between 0.60 and 1.20 m (Kanchev and Kancheva 1990). Apart from the concentration of flat cemeteries with inhumations in the Upper Thracian Plain, there is also a flat cemetery with eight excavated inhumations in southwestern Bulgaria, near Sandanski. Concerning the ritual in the burial mounds in the Upper Thracian Plain, there are a few tumuli covering one or more inhumations in the centre of the tumulus, and in one case, there are three skeletons buried together. Aside from these isolated examples, LBA tumuli are mostly associated with cremation. In both inhumation and cremation burials, grave goods are quite rare and typically consist of a vessel or two. The distribution of tumuli in the western Rhodopes, where most of the burial mounds are considerably richer in comparison to the other sub-regions, containing bronze weaponry, gold, silver, bronze or amber jewels is noteworthy.

These sites although traditionally interpreted as ‘sanctuaries’ or ‘peak sanctuaries’ (Domaradzki 1986a; Kissyov 1990b, 64-74; Raduncheva 1998), have also sometimes been referred to as habitation sites (KoukouliChryssanthaki 1992, 820). The scholarly trend of isolating this group became popular amongst researchers designating all peak-top sites in the area as places of cult practice, where some of the Bulgarian archaeological institutions undertook several expeditions to investigate such ‘sanctuaries’ along with megalithic sites in the Rhodopes, Sakar and Strandzha mountains (Delev 1982; 1984; Mikov 1933, 144; Venedikov 1976, 32; Venedikov et al. 1982, 173). The main reason for their interpretation as cult sites was the rocky nature of the steep peaks where, despite the discovery of significant amounts of LBA pottery, continuous habitation would have been physically challenging. No evidence of cult practices, however, have been found or excavated at any of these sites, except for at Perperikon (Leshtakov 2007b).

There is rarely evidence of a stable grave construction with the tumuli, but if there is one, it is usually indicated by a pile of stones covering the grave or sometimes a rectangular or circular stone feature surrounding the deceased. The main feature, however, is the urn, which is usually a ceramic vessel, covered by another ceramic vessel. Sometimes the urn is placed directly on the ground or within the stone ‘circles’. The mounds are relatively small, from 3 to 10 m in diameter and 0.45 to 1.50 m in height and grouped in clusters from 2 to 20 burial mounds (Kissyov 1989, 43). Often, the initial, lowermost layer of the mound consists of small to medium-sized cobbles, covered with a mixed layer of soil and pebbles. In a few cases, there are stone markers on top of the mounds or secondary burials in the slopes.

Other categories In addition to these three main categories, there are also several other smaller site types dated to the LBA, such as rock niches, enclosure ditches, mines, stone circles, and caves. Of the mines, only two recently discovered sites can be related to mining activities exploited in the prehistory. The mine site by Sedefche was used to exploit silver deposits and was working as recently as the 1930s. Some LBA pottery was discovered in front of its entrance, which suggests possible exploitation in the distant past as well. More recently excavated is a gold mine on the northern slope of Ada Tepe in the eastern Rhodopes where evidence for Bronze Age activity is more clear-cut. The site was initially dated in the fourteenth to thirteenth centuries BC (Popov 2011, 136), but the start of its use was later thought to be not later than the mid-second millennium BC (Jockenhövel and Popov 2008: 254-257).

In the Rhodope Mountains, the tumulus burial structure appeared for a first time during the LBA and was never used before in this micro-region. There is, however, a significant level of topographic continuity combining the existing cemeteries with later EIA and Roman burials, adding up to 200 tumuli within the same tumulus cluster. Despite this continuity in using the same funerary landscape and burial structure, the burial ritual changes, abandoning LBA cremation at the start of the EIA to be replaced almost entirely by inhumation (Kissyov 1991; 1993a). In contrast, during the same transition in the Upper Thracian Plain, inhumation is replaced by cremation in urns from another ceramic tradition, the knobbed ware. What such shifts in the funerary rituals might mean in terms of general patterning, will be discussed in the following chapters.

Rock niches, which sometimes have steps and cut marks, are thought to appear first in the LBA (Leshtakov 2002; Hristov 2001). They have often been interpreted as sanctuaries and referred to as ‘rock sanctuaries’ (Fol 1980; Hristov 2001; Venedikov 1982;). However, there is no consensus concerning the time of their cutting and their function, even if in some cases LBA pottery was scattered at the foot of rock outcrops with niches (Bozhinova 2007a; 2007b). There is also an isolated example of an enclosure filled with LBA pottery that is not related to any other structures

38

Site characterisation and issues of preservation

Figure 5.3. Distribution of sanctuaries across the study area.

properly prevents including them in any more formal investigation.

(Hristova and Ivanov 2010, 123-126). It was interpreted as a ‘ritual complex from the second millennium BC’ (Hristova and Ivanov 2010, 123), and is an elliptical structure, carved into the bedrock from three to eight meters in depth, reminiscent of an enclosure of central European character.

Site distribution The significant difference observed between the number of sites within some categories can be explained by uncertainty born of varying investigation in the various sub-regions, as well as the deliberate targeting of certain site types based on differing research agendas (Fig. 5.4). For example, in the area of the western Rhodopes, the most commonly investigated type is the cemetery due to the high number of burial mounds visible above ground, which were therefore easily targeted by both looters and archaeologists. In fact, most of the archaeological artefacts from this part of the mountains come from emergency and rescue excavations or are simply chance finds. Similarly, more attention has been paid to the also ‘visible’ tell sites in the Upper Thracian Plain, investigated mostly for their earlier prehistoric occupation during the Neolithic, Chalcolithic or the EBA, while few flat cemeteries and open-air settlements were discovered during the infrastructure projects described in chapter four. The burial mounds in this area are usually of later (IA and Roman) or earlier (EBA/MBA) character (see Alexandrov 2011), but isolated examples

During the renewed ‘Struma’ expedition in 2007, two circular stone structures without any clear context or function were associated with LBA pottery. Caves are also places with uncertain Bronze Age use, and only a limited number contain LBA material. For example, in the karst-rich western Rhodopes, only two out of some thirty identified caves with archaeological material contain LBA pottery (Yagodinska and Dolna Karanska caves). Those that were occupied during the LBA were probably used as temporary or seasonal settlements (Kissyov 1990b), although the religious character of cave use has been emphasised for contemporary contexts in the southern Aegean and should not be excluded as a possibility here either (Rutkowski 1986). No excavations have been undertaken, so it is difficult to offer greater certainty and arguments exist supporting both theories (Gotsev 1995, 49; Rutkowski 1986, 306; Teoklieva and Balabanov 1981, 22). The small sample of these more unusual site categories and our limited ability to characterise them

39

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 5.4. Levels of archaeological investigation across the study area.

The physical link between the mounds in the latter two regions is interrupted along the lower Struma river, where almost no sites have been discovered so far. It seems that the lower Struma constitutes a gap between the settlement mounds on both sides of the river. It is also possible that the area was generally avoided as a result of regular alluviation events, caused by changes in the coastal landscape between the fourteenth and the eleventh centuries BC when several gulfs and river estuaries became lagoons and inlets (Orachev 1988, 12). Such processes may have caused the displacement of a number of settlements further inland, or alternatively, they could have significantly obstructed the surface visibility of sites potentially in the area. Similar processes might explain the absence of sites along the Black Sea coast and the lower Maritsa watercourse. According to some, earthquakes might have also helped to cause the disappearance of at least some settlements along the coast in particular (Stiros 1996; Stiros and Papageorgiou 1991, 263-276, Fig. 5).

of LBA tumuli have also been excavated. Similarly, in the areas of northern Greece, LBA material comes almost exclusively from settlement mounds and the surface collection of artefacts from a few open-air settlements. More frequently investigated is the area of the eastern Rhodopes, which in addition to rescue excavations has been the subject of a long-term intensive survey thanks to the efforts of Georgy Nekhrizov and the work for the Archaeological Map of Bulgaria. More recently, the Valley of Struma has been subject to intensive surface collections and terrain survey campaigns. However, the results of these surveys are still largely unavailable but could potentially shed more light on the Struma microregion. Taking into account all this uncertainty, the informal spatial distribution of the categories defined above can be described as follows. Open-Air settlements appear more frequently between the eastern part of the Upper Thracian Plain, the eastern Rhodopes and the middle/upper Struma watercourse. Settlement mounds cluster into two groups in the Upper Thracian Plain and the western Thrace, connecting to the toumbas in Chalkidiki and its hinterland (setting aside the large concentration of settlement mounds further in the district of Macedonia), while they are absent in this period from the Rhodope Mountains and the Struma Valley.

An interesting distribution can be observed across the Upper Thracian Plain in comparison to the area of Chalkidiki and its hinterland. The settlements in both cases are primarily tells, which were the dominant type of settlement during the Bronze Age in northern Greece (Andreou 2001, 163;

40

Site characterisation and issues of preservation

Figure 5.5. Distribution of tell sites across the study area.

Wardle 1997, 518). The main difference is that the tells in Greece have a long BA tradition and solid evidence of habitation, while the ones in the Upper Thracian Plain have only sparse evidence on top of earlier prehistoric mounds. Nevertheless, only in these two micro-regions there are tell sites exhibiting a certain level of LBA occupation. On the other hand, sites like Dikili Tash and the rest of the settlements in the Greek part of the study area, do not have the typical signs of the LBA from the toumbas in central and eastern Macedonia (Hochstetter 1984; Wardle 1986). As a settlement tradition, they resemble those in the Upper Thracian Plain, founded during the early Neolithic and Chalcolithic times (Fig. 5.5).

the Late Bronze Age in the eastern Rhodopes (Popov and Jockenhövel 2011).

One view is that that the settlement pattern in the Upper Thracian Plain follows the distribution of the earlier tell sites, already abandoned for a few centuries and reinhabited sometime before the end of the Late Bronze Age (Leshtakov 2006, 167). At the same time, human occupation was also present on isolated peaks and hills. Another view about the settlement tradition in the study area is the possibility that in the eastern Rhodopes, Sakar, and Strandzha Mountains gold and silver ore deposits were exploited at some point during the Bronze Age, around which a settlement network developed. This hypothesis was recently supported by the excavated gold mine from

Likewise, the number of known cemeteries is still insufficient for a more in-depth study. Initially, academic interest in LBA funerary rites was restricted to the investigation of dolmen structures in Sakar and the eastern Rhodopes (Bonchev 1901; Mikov 1942; Shkorpil 1926). During this early period, there were only three excavations of burial deposits, one single burial in the lower slopes of Dzendem Tepe (Detev 1963), a mound cemetery near the Batak dam (Tsonchev and Milchev 1970), and an open-air cemetery near Dolno Sahrane (Getov 1965). This situation changed at the start of the 1970s when investigations took place in different regions, and a more intensive collection of

The number of settlements with stone architecture is not yet sufficient to be tested formally, but their concentration in the middle Struma and Mesta valleys is clear (Fig. 5.6). Their dominant position is supported by the presence of stone walls for which a reconstructed height of up to 5-6 m has been proposed (Stefanovich and Bankoff 1998, 273279), as well as the lack of intensive long-term habitation, which has encouraged them to be interpreted as primarily defensive in character (Stefanovich and Bankoff 1998, 273-279).

41

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 5.6. Distribution of sites with stone architecture across the study area.

worth reiterating that a number of cemeteries, especially burial mounds, are known because LBA material has been discovered on the surface, but no excavations have yet taken place.

new data began. The majority of information was collected via rescue excavations and, at the same time, the number of sites registered via surface collections also increased (Delev 1982; 1986; Gergova-Domaradska 1989; Kissyov 1990b; 1993). Despite this intensification, the number of flat cemeteries discovered remained small, partially because of the low archaeological visibility typical for this type of site. To summarise, the general distribution reveals that open-air cemeteries are spread almost exclusively in the Upper Thracian Plain, while single burial mounds with LBA evidence seem to have been discovered mostly in the eastern Rhodope area. The clusters of burial mounds, on the other hand, appear distributed across the western part of the Rhodope Mountains (Fig. 5.7; Kissyov 1990a; 1993; 1999).

As mentioned above, there are also some dolmens and cist-graves which have been discovered under low burial mounds in the eastern Rhodopes and Sakar. There are no dolmens beyond the Pilashevo–Bezvodno–Zagrazhden– Zlatograd line (Kulov 1991; Venedikov et al. 1982), which is usually considered to be the division between the western and eastern Rhodopes. It is worth clarifying that the overall distribution of prehistoric dolmens across the Thracian region is wider than this and also includes areas in the Strandzha Mountains and the Edirne region (Fig. 5.1; see Iliev 2008). Nonetheless, all of those structures are dated to the EIA roughly between the eleventh and eighth century BC (Erdoğu 2005; Özdoğan 1998; 1999; Özdoğan and Akman 1992;), and materials relevant to the current study have not been discovered. The dolmen structures as a group have been compared with, and sometimes linked indirectly to Caucasian dolmens, whose dates are spread between 2400-1000 BC (Iliev 2008; Joussame 1988, 265). There is a hypothesis that the Thracian megalithic tradition in general possibly originates in the LBA (see Leshtakov 2002), and some have argued it reflects a population

In terms of funerary rituals, cremation appears to be associated strictly with cemeteries of grouped mounds, and therefore its distribution also concentrates in the western part of the Rhodopes. There are very few cases of inhumation present in the eastern and western extremes of the area, but when it is present, it is usually discovered along with burials with cremation (Kissyov 1990a, 43). Flat cemeteries are distributed evenly across the area, but low levels of archaeological investigation do not allow us to link them with specific burial customs at present. It is 42

Site characterisation and issues of preservation

Figure 5.7. Distribution of burial sites across the study area.

any possible preferences in regional locational choices before going into any details of the analysis. Using some of the tools of archaeological predictive modelling and other spatial statistical techniques I will, in what follows attempt, to trace signs of more complex spatial patterning, over a wider period and with respect to topographic, geomorphological and geological characteristics of the landscape.

movement from the southern Russian Steppes towards the Balkan south (Hoddinott 1981). While being mindful of the uncertainty this burial type and its appearance in the area brings, they are included in the spatial analysis of chapter six. Apart from an extensive exploration of the relative spatial pattern of these different site types with respect to one another, a further task in later chapters will be to understand

43

6 Quantifying site distributions

Figure 6.1. Map of the area showing the main places mentioned in the text.

since they do not necessarily imply a direct cause-effect relationship and potentially underemphasise the role of other factors (Kohler and Parker 1986, 400, 401). Correlations, however, can still help to indicate a set of factors influencing people’s past beliefs, behaviour and choices, and how sites of different character are patterned across the landscape.

The second part of the spatial analysis will address site-level data. Although it is common sense that a community’s choice of a place to live is often dependent on cultural, economic and political criteria, the location of a site is also influenced by the geomorphology and general environmental affordances of a region (Perakis and Moysiadis 2010). As described in chapter one, the study area has a diverse topography, local climate differences and natural distribution of resources (Fig. 6.1), which must have in one way or another influenced decision-making about occupation or choices about land use. However, any seemingly obvious relationship between site locations and the landscape can obscure different local and chronologically varying strategies. In this respect, it is necessary to stress that the correlations revealed in what follows need to be treated with caution

With such spatial approaches, there is also a significant dependence on scale. Ignoring the importance of scale and its complexities risks misrepresentation of the past (Lock and Molyneaux 2006, 1). The essence of the problem lies in comparing the relationships identified at two or more different scales (Haggett 1965, 164). In the social sciences, the challenge of scale forces us to think hard about whether an observed pattern is most relevant to 45

Late Bronze Age Social Landscapes of the Southeast Balkans

Figure 6.2. Map of the study area showing the extents of the proposed zones.

an individual, a group, a community or a wider ‘society’ (Mathieu and Scott 2004, 2). In choosing their study areas, archaeologists often ignore the existence of the scalar trap, sometimes called the ‘ecological fallacy’ and related spatial statistical issues such as spatial heterogeneity and spatial non-stationarity. This means that the choice of a study area shape and size can influence the outcome of an analysis (Heywood et al. 2002, 8). Nevertheless, there is no one correct methodology for defining a study area to avoid misleading results. Most areal units are modifiable and depend on the individual research questions and the researcher’s preconceptions (Openshaw 1983).

some models will be juxtaposed to identify the most significant combination of predictors for each area. In total, nine alternative micro-regions will be explored (Tab. 6.1). Some of the areas overlap and thus, the sum of the number of sites from all areas does not amount to the total number of sites in the study area (n=361). The definition of these exploratory zones was first based on the patterns detected in the pottery distributions, combined with the distribution of site types. Overall, there are four main areas of interest – the Upper Thracian Plain, the Rhodope Mountains, the valley of the Struma, the northern Aegean hinterland and, as an external area, and the Chalkidiki peninsula and its hinterland. Although this division seems logical overall, there remain areas where zones overlap or could be sub-divided. For example, the Upper Thracian Plain could be taken as a whole, but its eastern part with the highest density of archaeological sites could also be examined individually. The Rhodope Mountains, traditionally divided into west and east, deserve to be tested both together and split in two. The overall analysis will examine models at different scales emerging from such an investigation (Fig. 6.2).

An understanding of scale is crucial to interpreting archaeological spatial patterns. Important questions involve: how to define the appropriate scale, how consistent the results will be if we change it, whether the patterning is robust and to what degree it matters if we have recognised the right scale. One pitfall is either to lump areas that are better off being treated separately or to split others that show no real sign of different behaviours. Accordingly, one of the tasks of the spatial analysis in this research is to explore the influence of scale in the study area and to acknowledge possible significant relationships that require different interpretations for different sub-regions. So while one of the main tasks in this study is to offer a ‘unified’ archaeological synthesis of the southeast Balkans as a place otherwise torn apart by modern political divisions, it would be somewhat illogical to assume a continuous, consistent distribution within the physical limits of the area as defined in chapter one. The site-type distribution described in chapter five and the pottery distributions already implies a certain level of regionalism, which will be further explored below. Selection of zones and parameters

Table 6.1. Analytical zones used in the location modelling.

According to the methodology described in chapter two, the main statistical framework consists of location modelling based on multivariate logistic regression. Accordingly, 46

Quantifying site distributions

Figure 6.3. Raster maps representing the twelve selected covariates: a) elevation, b) slope, c) aspect, d) distance to large rivers, e) distance to small- and medium-sized water sources, f) distance to marble, g) distance to limestone, h) distance to conglomerate, and i) distance to alluvium.

Proximity analyses are common in prehistoric archaeology, especially when it comes to measuring the distance between different resources and site location. Proximity to water sources is an assumed objective when choosing a settlement location. According to the specifics of the area, two main types of water sources with relevance to human activity can be separated: i) small–medium-sized watercourses characterised by modest hydrographic regimes, possibly targeted mostly for domestic needs, and ii) large river bodies and riverine resources able to serve domestic needs, but often navigable and thus also of use as transport arteries. Instead of regular Euclidean distance, we have used a rough estimate of the cumulative cost of travelling through each cell (using ArcGIS’ isotropic cost distance method). Such an estimate was necessary due to Thrace’s diverse terrain and areas of rugged relief, such as in the Rhodope Mountains and along the Struma Valley, where straight line Euclidean distance does not accurately capture whether a point will be easily accessible. This is often true for locations such as peaks or hilltops. Furthermore, particular landforms, such as ridges, peaks or plains, give the impression of defining different settlement strategies. Location on ridges was defined using fuzzy membership classes where is 0 is a not a ridge, and 1 is a definite ridge with fractional probability in-between. A value is assigned to each pixel on the resulting grid with a spatial resolution 30 m. As a result, the final set of predictors is defined as follows: 1) elevation, 2) slope, 3) aspect, 4) distance to small- and medium-sized water

When confronted with the question of where archaeological sites might be located in the landscape, it is not difficult to suggest the kinds of environmental characteristics traditionally considered important. Whether sedentary or nomadic, societies need access to drinking water and, sometimes, favour naturally fortified landforms (e.g. hilltops or valleys) concerning climatic challenges, social encounters or both. Communities with specific economic needs might have also paid attention to factors such as transport arteries and access to specific raw materials. The most common geological features of the Thracian area consist of soft limestone, marble and conglomerate as well as gneiss and schist deposits and alluvium formations. It is mostly from these basic criteria that a set of independent variables or covariates was created. Cultural variables were not applicable in the current dataset, where the high degree of uncertainty in the in-depth interpretation of different sites would minimise the benefits of such a formal approach. There are two underlying data sources upon which all analysis is based: a 30-meter ASTER global elevation model (https://gdex.cr.usgs.gov/gdex/) and a 1:250 000 geological map of the area. From the elevation model, some further essential information about the terrain was obtained, such as slope and aspect. A map of rainfall was not included, because there is not enough observable variation in precipitation across the region to justify its use.

47

Late Bronze Age Social Landscapes of the Southeast Balkans

Table 6.2. A list of variables used in the multivariate regression models presented in this chapter.

sources, 5) distance to large river bodies, 6) distance to alluvium, 7) distance to conglomerate, 8) distance to limestone, 9) distance to marble, 10) distance to gneiss and schist, 11) location on ridges, and 12) location on peaks (Fig. 6.3; Tab. 6.2). More complex variables such as viewsheds and composite accessibility maps seem unwarranted given the large scale of much of the analysis and the patchy nature of the data. A slightly finer-scale sub-regional study applied to an unusually well-surveyed region in the eastern Rhodopes is included at the end of this chapter.

non-site locations generated in each zone separately to match the exact number of the sites in that zone. Thus, each zonal dataset contains a dependent variable (DV) classified in sites (1) and non-sites (0). The uncertainty of non-site locations is a problem, and a regression model built on one random non-site set may differ from another. A more robust, bootstrapped technique was introduced by Bevan and Conolly (2013, 268-270). Following the logic of their argument in slightly modified form, I have included two additional samples of locations absent of identified archaeological sites, including respectively 100 and 500 data points, which are tested against the observed number of archaeological site locations for each zone in order to explore how sensitive the results are to such changes.

Under the logistic multivariate regression method used here, the dependent variable is dichotomous, made up of both site presences and absences. The former consists of known archaeological locations, while the site absence category should ideally consist of directly confirmed nonsite locations. However, in this particular study, not all regions have been explicitly surveyed, and the non-site sample is created using a random sampling method, with

Before the initial multivariate model is generated, questions of interdependence and significance amongst the possible predictor variables also need to be addressed via diagnostic tests. Due to the varying shape of the frequency distribution of each covariate, an exploratory graphical

48

Quantifying site distributions

Figure 6.4. Histograms of independent variables showing raw and log-transformed distributions.

device (Q-Q plot) comparing the observed and the expected value for each data point, is employed to check the validity of the assumption of a normal distribution. Because the significant skewing impact on the results can also come from potential outliers in the site and non-site locations (see Wheeler and Tiefelsdorf 2005), I applied a Bonferroni’s t-test to calculate a probability value for the extreme observation and calculated a Cook’s distance to identify statistically influential data points (see Kalogirou 2013). Log transformation is also applied when necessary to each independent covariate in an attempt to change the shape of skewed distributions so that they meet the assumptions of parametric statistical tests, such as normal distribution (Fig. 6.4). As a guide to the consistency of the conclusions, all models were built with both the raw and the log-transformed datasets for comparison.

redundant variables. Therefore, and also due to issues of interdependence in the predictors, which can confuse the interpretation and compromise the model itself, the data was also tested using a variance inflation factor (VIF) and correlation matrixes (see Salmerón Gómez et al. 2016). While correlation matrixes compare each pair of variables, VIF quantifies the inflation of the standard errors of the estimated coefficients and can detect linear dependence among three or more variables. Finally, and as an alternative approach, analysis of covariance (ANCOVA), evaluated whether the means of the dependent variable is equal across the set of independent variables. Applying ANCOVA is a way to address the same dataset using a different approach for validity. After the application of these tests, an interpretation regarding site location decisions is suggested based on a comparison of the user-defined global model, the AIC model, and the ANCOVA, comparing raw and log-transformed data. Confidence intervals (CI) are also taken into account as a part of the hypothesis testing. In addition to the single point estimate (the most likely value), confidence intervals comprise the margin of error or the amount of uncertainty around that value. When CIs are reported, one can assess the correctness of the point estimate in the sense that a narrower interval around the sample estimate suggests precision (Garthwaite et al. 2002).

Due to the relatively large number of possible predictor variables, I will explore certain key candidate variables in a univariate way first, and then the significant ones will be combined via stepwise model selection using an Akaike Information Criterion (AIC), in order to minimise the loss of information and to detect which model best approximates the observed data. In this case, AIC is applied as a second option after the general fitted model has been produced, to assist in evaluating statistically 49

Late Bronze Age Social Landscapes of the Southeast Balkans Locational modelling Zone one: the eastern part of the Upper Thracian Plain The site types and pottery distributions examined so far have come to indicate that the Upper Thracian Plain forms a separate unit within the study area. There are a few possible ways in which to, to sub-divide it. The smallest scale embraces the eastern part of the plain, where one can observe a higher density of archaeological sites, albeit still not that many sites. An initial statistical analysis consists of 86 locations combining sites and non-sites. Within this subset, in terms of statistical diagnostics, Bonferonni’s t-test shows no studentized residuals with Bonferonni p < 0.05, meaning that there are no statistically influential data points, which may skew the results (Tab. 6.3). One of the unexcavated site locations was identified as the largest studentized residual or the most deviant point in the distribution. According to the Cook’s distance plots, there are a few unusual places like two sites and one random location (72) in the raw dataset (Fig. 6.5) and only non-site locations (51, 72, 84) from the log-transformed data (Fig. 6.6). The log-function significantly helped the standardisation of the data in this case as is visible on the QQ-plot (Fig. 6.7 and 6.8). Multicollinearity was detected only between elevation and distance to alluvium, which was corrected with the transformation (Tab. 6.4).

Figure 6.6. Cook’s distance plots log-transformed data of zone one.

Figure 6.7. Q-Q plot the raw data of zone one. Table 6.3. Non-constant variance score test of zone one.

Figure 6.8. Q-Q plots of the log-transformed data.

Figure 6.5. Cook’s distance plots of raw data of zone one.

50

Quantifying site distributions In this area, the raw data model based on multivariate regression indicates that the only good predictor of the distribution of sites across the landscape is a negative correlation with distance to alluvium at p 0.1). However, the model of the log-transformed data increased the variance explained to 25.3 per cent (at p < 0.05) and detected a distance to large rivers as a second strong predictor with p < 0.05 (Tab. 6.6).

Table 6.4. Variance inflation factor test for multicollinearity applied to the data of zone one.

As the next step of the multivariate regression analysis, a stepwise AIC comparison was involved in suggesting a better combination of predictors. The already identified significant predictor was now not selected in the AIC model based on the raw data (Tab. 6.7). The AIC model of the log data combines distance to alluvium, and large rivers also suggested by the log data model (Tab. 6.8). The variance explained by this model is about 17 per cent, where p = 0.001633, which makes the model statistically significant. The same combination of predictors is also supported by the ANCOVA of the log data (Tab. 6.9 and 6.10) and by the narrow confidence intervals around the regression values of the relationships between each of these two predictors and the site locations (Tab. 6.12).

Table 6.5. Linear model of the raw data of zone one, where *** p