We would like to express our deepest gratitude to our friend and colleague Ira Schwartz who took the time to review and comment on early iterations of this paper. We would also like to thank the co-organizer of the Delineating the End of a World workshop, Dr. Johnny Samuele Baldi, who also provided encouragement, support, and comments on earlier versions of this article. Also, we would like to thank the scholars who presented at the workshop, and the members of the audience who gave us comments and feedback on our presentation. Finally, our deepest gratitude goes to the three anonymous reviewers for their insightful comments and suggestions which helped improve this paper.
1The developments that took place during the Ubaid and Late Chalcolithic 1 (LC1 hereafter) periods in Greater Mesopotamia have been addressed independently by several scholars (Henrickson and Thuesen 1989; Rothman 2001; Postgate 2002; Stein and Özbal 2007; Carter and Philip 2010a; Stein 2010a; Marro 2012a; McMahon and Crawford 2014; Fisher 2017; Abu Jayyab 2019). The consensus is that the strong interaction networks that emerged across Greater Mesopotamia during the Ubaid 3-4 (5200-4500 BCE) periods gave way to fragmented communities characterized by localized adaptations during LC1 (4500-4200 BCE). Yet the transition between the Ubaid and LC1 periods remains poorly understood (however see work by Baldi 2012b, 2016a; Frangipane 2012; Stein 2012), and has not been examined in a systematic fashion at a sub-regional level (Carter and Philip 2010b; Marro 2012).
2In this work, we explore how interaction networks may have changed across time and space during the Chalcolithic and how spatial and material reconfigurations affected societal developments. We examine stability and change in networks of interaction during the Ubaid 3 (5200-4900 BCE), Ubaid 4 (4900-4500 BCE), and LC 1 (4500-4200 BCE) periods. We use multivariate statistics and formal Social Network Analysis (SNA) to analyze pottery data from several sites across Greater Mesopotamia. The aim is to diachronically examine the strength and directionality of interactions between sites within each period, and how they may have changed over time.
3The Ubaid “culture” was defined based on several characteristics that appeared in southern Mesopotamia. These characteristics included the appearance of ophidian figurines, niched and buttressed tripartite architecture, the practice of skull modification, the use of ear spools and labrets, the development of group cemeteries, the appearance of clay mullers, and the manufacture of painted black-on-buff pottery (Henrickson and Thuesen 1989; Stein and Özbal 2007; Carter and Philip 2010b; Stein 2010a). In 1960 Oates proposed a four-tier division of the period, Ubaid 1-4, based on the sequences from Ur and Eridu (Oates 1960). Some adjustments were made to this chronological scheme, namely the addition of Ubaid 0 noted at the site of ‘Oueili (Lebeau 1987), extending the sequence back to 6500 BCE. During the late 6th millennium BCE, or the Ubaid 3 period (5300/5200 BCE), certain elements of Ubaid material culture began to appear in northern Iraq, northern Syria, and southeastern Turkey, in tandem with the local northern “Halaf” culture (Akkermans and Schwartz 2003: 154-156; Campbell 2007). Elements of Ubaid material assemblages lasted across northern Mesopotamia until the end of the Ubaid 4 (4500 BCE) when the Ubaid cultural manifestations began to give way to more regionalized material cultural expressions of Late Chalcolithic 1 (Rothman 2001; Baldi 2012b, 2016a; Marro 2012; Stein 2012).
4The main explanations for the spread of Ubaid material into neighboring regions were migrations from southern Mesopotamian into northern regions or the acculturation and adaptation of northern groups of Ubaid style material culture (see Baldi 2020: 72). Advocates of migration saw that groups from southern Mesopotamia moved and settled into areas of the north, in a similar pattern to later Uruk migrations (Thuesen 2000; Oates and Oates 2004; Hole 2016). Others saw that the similarity in material culture resulted from the gradual adoption of southern styles by northern groups (Stein 1994, 2010; Breniquet 1996; Frangipane 2001: 322; Stein and Özbal 2007). These processes took place through a combination of trade, the adoption of Ubaid social identity and ceremonial ideology (Stein 1994, 2010a: 34), or intergroup marriage (Breniquet 1996), rather than direct colonization. Baldi (2020) further expanded the process of Ubaid acculturation between southern Mesopotamia and adjacent regions (including the western Levant). He observed an overlap between local ceramic assemblages and Ubaid assemblages at sites in the Levant and northern Iraq and suggested that the Ubaid cultural expansion could have begun as early as Ubaid 1-2 (Baldi 2020: 85; also see Campbell and Fletcher 2010).
5Following a period of roughly 800 years, the loose links across the Ubaid interaction sphere seemed to have weakened after 4500 BCE. This is particularly visible in the gradual disappearance of Ubaid painted pottery styles across Upper Mesopotamia during the Ubaid-LC transition (Stein 2012: 132). The reasons behind this weakening of the Ubaid social networks are still poorly understood. Baldi attributed LC1 regionalization to the emergence of major centers that broke up the uniformity seen in the Ubaid territorial system of villages (Baldi 2012b: 405). Fisher (2017, this volume) observed that increased innovation in different parts of the Near East and the increased social differentiation that resulted from control of these innovations may have contributed to the break we see between the Late Ubaid and LC1 periods.
- 1 Challenging this perceived cohesion, Baldi (2016) identified a total of nine ceramic microregions (...)
6None of the assertions made about the emergence of an Ubaid trans-regional network or indeed its decline and dissolution have been completely tested (see Carter and Philips 2010: 14; Marro 2012: 17). Adding to the ambiguity is the internal bias inherent in cultural historic terms such as Ubaid (see Campbell 2007: 105; Carter and Philip 2010: 13; Pollock 2010), especially when set against chronological time slices such as the Late Chalcolithic 1 (see Rothman 2001). Despite many scholars challenging the notion of an Ubaid cultural homogeneity (Akkermans 1989; Henrickson and Thuesen 1989; Stein and Özbal 2007; Stein 2010a: 34; Baldi 2012b, 2016a: 125; Hole 2016; Marro 2012), the term itself remains an obstacle to analyses.1 In this study, we aim to quantify the flow of different elements of ceramic assemblages across the region (also see Baldi this volume). This approach is intended to develop an understanding of interaction networks beyond the framework of transition from Ubaid to LC1, to a perspective that sees the Ubaid and LC1 as part of a continuous social process.
7To gauge the Ubaid phenomenon, Carter and Philip (2010: 8) recommended disentangling the bundles of elements it consists of. Here we apply that advice by disentangling ceramic assemblages. We see ceramic assemblages as a microcosm of the Ubaid interaction sphere, in which elements were not transmitted in unison but rather in a more selective fashion (Stein and Özbal 2007; Özbal 2010; Stein 2010a; Baldi 2020a). Rather than assuming a priori that interaction spheres during the Ubaid period remained constant or that interactions in the LC1 represented a complete departure from them, our approach focuses on understanding the gradual changes that may have occurred over time.
8Due to the way knowledge is transmitted, craft production and consumption, in general, and pottery production in particular, are deeply embedded within a group’s cultural disposition (Dietler and Herbich 1994; Dobres 2000). Sharing in elements of ceramic production and consumption implies a strong degree of communication and sometimes even face-to-face learning between participants in the manufacturing process (Dietler and Herbich 1994: 465; Gosselain 1998: 78; Dobres 2000: 129; Bowser and Patton 2008: 108; Carter and Philip 2010: 8-10). This comes from the assertion that certain behaviors and technical gestures learnt during the apprenticeship stage become deeply embedded and integrated as motor habits in the craftsperson’s way of doing (see Gosselain 1998: 94-95, 2000, 2015). The participation in these transmission networks is what Lave and Wenger refer to as “communities of practice” (Lave and Wenger 1991; Wenger 1998).
9Knappett and van der Leeuw (2014: 73; also see Eckert 2008: 2) see that some bodies of knowledge will intersect inside large organizations composed of numerous communities of practice. Constant movement and transfer of ideas between sites that share “communities of practice” may develop certain overlapping social structures. As such, examining the distribution of potting traditions across space, and changes within these patterns, can help us understand stability and change in interaction networks from the Ubaid to the LC1. Beyond production, sharing in consumption practices through use of certain vessels or incorporation of different decoration elements, in some cases, may imply the existence of a shared habitus between communities (see Kennedy 2019, this volume). While exploring the context of vessel use is beyond the scope of this study, stronger overlap in ceramic assemblages between sites may suggest stronger underlying structures than between sites that show less overlap.
10By examining changes in the shape and directionality of interaction networks from the early Ubaid, late Ubaid, and LC1, and understanding where connections persist and where they break down across the landscape, we can begin to empirically test some assumptions about the spread of Ubaid material culture into northern Mesopotamia, its development over time, and the circumstances pertaining to the onset of the LC1.
- 2 The coding system for the variables of Ubaid 3-4 and LC1 are independent from one another. This is (...)
11In this work, we look at a dataset consisting of pottery forms and decoration elements across the three periods of Ubaid 3, Ubaid 4, and LC1 (fig. 1).2 We use multivariate statistics (Principal Component Analysis or PCA) and formal Social Network Analysis to explore the ceramic data and determine the strength of social interactions across these periods. For each period we conducted three tests: first for the full dataset (form and decoration elements combined), and then independently for form and for decoration. This three-level analysis allows us to view overall network behaviour, and to assess if form or decoration played a more significant role in the emergent patterns. The combination of PCA and network analysis allows us to test the robustness of each method and explore the network from different perspectives and scales. We can examine the overall structure of the networks at the macro-scale, the role various sites played in these networks at a meso-scale, and which ceramic elements may have governed clustering at a micro-scale.
Fig. 1 – Decoration elements and forms from the Ubaid (left) and LC1 (right) used in the analyses.
- 3 Such finds may have been chance recoveries governed by the size of the excavations carried out and (...)
12The reason behind the use of ceramic data is that it is highly visible in the archaeological record and represents the most common artifacts recovered and reported. Incorporating other elements beyond ceramics3 (e.g., house forms and burial practices), has the potential to bias the dataset. Alternatively, more common vessel forms are likely to still be represented regardless of the size of an excavation. Moreover, ceramics are the main element used to categorize a site as being “Ubaid” even with the absence of other elements in some cases (see Marro 2012). Ceramics were also used to address LC1 regionalization (see Schwartz 2001; Baldi 2012b, 2016). It is important to point out that we do not assume an overlap between the flow of pottery and other fields of material culture (see Carter and Philip 2010: 8), this study is just meant to be heuristic and always open to further refinement.
- 4 We have been asked during the presentation of this paper at 12th ICAANE and by one of the anonymou (...)
13The ceramic data was derived from 107 levels that came from 72 sites located in southern and northern Mesopotamia, southeastern Turkey, western Iran4 and the northern Levant (fig. 2, tab. 1). The areas were further divided into sub-regions with varying boundaries during each period. These sub-regions are from southeast to northwest: southern Mesopotamia, Hamrin/Mandeli, Iraqi Kurdistan, the Tigris Basin, Diyarbakir, the Khabur Basin (upper and lower), the Balikh/Euphrates, and western Syria. These regions were determined by preliminary clusters generated through PCA and on clusters proposed by Baldi (2016).
Fig. 2 – Sites included in the study.
Table 1 – Periodization of sites in the study Ubaid 3, Ubaid 4, and LC1.
Ubaid 3 Sites |
Level |
Date range |
Region |
Reference |
As-Sabiyah |
Periods 3-4 |
5530-4950 |
Gulf |
Carter and Crawford 2010 |
Bahra 1 |
Single Period |
5571-5283, 5297-4949 |
Gulf |
Smogorzewska 2016, 2020 |
Choga Mami |
Well |
|
Hamrin |
Oates 1983 |
Kall Karim |
Surface |
|
Hamrin |
Mazaheri 2018 |
Tell Songor AQ |
Lower Levels |
|
Hamrin |
Jasim 1985 |
Tell Abada |
Level I-II |
|
Hamrin |
Jasim 1985 |
Tell Rashid |
Levels 1-3 |
|
Hamrin |
Jasim 1985 |
Tell Arpachiyah |
Cemetery |
|
Northeast |
Mallowan and Rose 1935 |
Tell Qalinj Agha |
Sounding E-F |
|
Northeast |
Hijara 1973 |
Salat Tepe |
Early Ubaid Levels |
|
Northeast |
Ökse and Gormus 2013 |
Tell Tanjero |
Levels II-III |
|
Northeast |
Saber et al. 2014 |
Tepe Gawra |
XVIII-XVII |
|
Northeast |
Tobler 1950 |
Telul eth Thalathat |
XIV-XII |
|
Northeast |
Shimogama 2019 |
Yarim Tepe III |
Levels IV-II |
|
Northeast |
Merpert and Munchaev 1982 |
Hammam et Turkman |
IVa-b |
5290-4810, 5100-5000 |
Northwest |
Akkermans 1988a, 1988b |
Tell Kosak Shamali |
Levels XII-XIV |
|
Northwest |
Nishiaki and Oguchi 1999 |
Tell ‘Abr |
Level 7-6 |
5200-4950 |
Northwest |
Hammade and Yamazaki 2006; Yamazaki 2010, 2012 |
Tell Mefesh |
Al Ubaid period |
|
Northwest |
Mallowan 1946 |
Tell Zeidan |
Operation 1 |
5150-5000 |
Northwest |
Stein 2009, 2011 |
Al Ubaid |
Single Period |
|
South |
Hall and Woolley 1927 |
Eridu |
TS XI-VIII |
|
South |
Safar et al. 1982 |
Ras Al Amiya |
Levels I-III |
|
South |
Stronach 1961 |
Tel al ‘Uwaily |
X36 Level 9 |
|
South |
Calvet 1987 |
Ur |
Level I |
|
South |
Woolley 1955 |
Hama |
L Early |
|
West |
Thuesen1988, 1989 |
Tell Kurdu |
Amuq C |
|
West |
Braidwood and Braidwood 1966 |
Ubaid 4 Sites |
Level |
Date range |
Region |
Reference |
Yanice Yeni |
YY5 |
4500-4350 |
Diyarbakir |
Bernbeck and Costello 2011; Kennedy 2008 |
Kenan Tepe |
Late Ubaid Levels |
4800-4450 |
Diyarbakir |
Parker and Kennedy 2010; Kennedy 2019 |
Tell Serik |
Surface |
|
Hamrin |
Oates 1968 |
Tell Haizalon |
Single level |
|
Hamrin |
Jasim 1985 |
Tell Madhur |
Levels 1-2 |
|
Hamrin |
Moon and Roaf 1984; Roaf 1989 |
Ibrahim Bayis |
Sounding B |
|
Kurdistan |
El-Amin and Mallowan 1950 |
Kani Shaie |
VII |
5000-4900 |
Kurdistan |
Renette et al. 2021 |
Kudish Saghir |
Trench 16 |
|
Kurdistan |
Starr 1937 |
Qalat Said Ahmadan |
Level 1-2 |
5000-4800 |
Kurdistan |
Tsuneki et al. 2016 |
Ubaid 4 Sites |
Level |
Date range |
Region |
Reference |
Tell Qalinj Agha |
C-D sounding |
|
Kurdistan |
Hijara 1973 |
Surezha |
Various |
4840-4720, 4780-4680 |
Kurdistan |
Stein 2018; Stein and Alizadeh 2014; Stein and Fisher 2020; Alden et al. 2021 |
Tell Halawa |
Levels 6-7 |
|
Kurdistan |
Peyronel et al.2016; Peyronel and Vacca 2015 |
Tell Nader |
Late Ubaid Levels |
|
Kurdistan |
Kopanias ed. 2013 |
Yorghan Tepe |
XII-X |
|
Kurdistan |
Starr 1937 |
Tell Feres |
Levels 10-9B |
4690-4545 |
Northcentral |
Baldi 2015 |
Tell Kashkashuk II |
Cem |
|
Northcentral |
Koizumi 1993 |
Tell Kashkashuk III |
VIIb-VIIA loc.77 |
|
Northcentral |
Sultan 2019 |
Tell Kuran |
B-D |
4826-4552 |
Northcentral |
Hole and Tonoike 2016 |
Tell Leilan |
VIa strat 58 |
|
Northcentral |
Schwartz 1988 |
Tell Mashnaqa |
III |
|
Northcentral |
Hole and Tonoike 2016 |
Mulla Matar |
Ubaid Stratums |
|
Northcentral |
Sürenhagen 1990, 1991 |
Tell Beydar III |
Level 4 |
|
Northcentral |
Suleiman and Nieuwenhuyse 2003 |
Tell Ziyadeh |
EXI-VIII |
4900-4500 |
Northcentral |
Hole and Tonoike 2016 |
Khanijdal |
Levels 1-3 |
|
Northeast |
Wilkinson et al. 1996 |
Salat Tepe |
Late Ubaid Levels |
|
Northeast |
Ökse and Gormus 2013; Koizumi et al. 2016 |
Tell al Hawa |
Surface |
|
Northeast |
Ball et al. 1989 |
Tell al-Imsihli |
cem |
|
Northeast |
Hayani et al. 2011 |
Telul eth Thalathat |
XI-VIIb |
|
Northeast |
Egami 1959; Shimogama 2019 |
Tepe Gawra |
XVI-XIII |
|
Northeast |
Tobler 1950 |
Hammam et Turkman |
IVC |
5000-4600 |
Northwest |
Akkermans 1988a, 1988b |
Tell Kosak Shamali |
Levels A11-B7 |
4940-4713 |
Northwest |
Nishiaki and Oguchi 1999 |
Tell ‘Abr |
Levels 5-3 |
4900-4600 |
Northwest |
Hammade and Yamazaki 2006; Yamazaki 2010, 2012 |
Tell Zeidan |
Various |
4950-4660 |
Northwest |
Stein 2011 |
Eridu |
TS VII-VI, HS VIII-I |
|
South |
Safar et al. 1982; Perkins 1949 |
Tell Zurghul |
Area B |
|
South |
Volpi 2020, |
Telloh |
É poterie peinte |
|
South |
de Genouillac 1934 |
Tell Uqair |
I-VI |
|
South |
Lloyd and Safar 1943 |
Ur |
II-III |
|
South |
Woolley 1955 |
Uruk |
KXVII levels 6-5 |
|
South |
Boehmer 1972 |
Tell al ‘Uwaily |
Ubaid 4 |
|
South |
Lebeau 1983a, 1983b |
Coba Hoyuk |
IVA-B |
|
West |
du Plat Taylor et al. 1950 |
Hama |
L Late |
|
West |
Thuesen1988; Thuesen 1989 |
Tell Kurdu |
Amuq E |
|
West |
Braidwood and Braidwood 1960 |
LC1 Sites |
Level |
Date range |
Region |
Reference |
Salat Tepe |
LC1 Levels |
4500 roughly |
Diyarbakir |
Ökse and Gormus 2013 |
Tell Qalinj Agha |
levels A-B |
|
Iraqi Kurdistan |
Hijara 1973 |
Tell ad-Deim |
XI-VIII |
|
Iraqi Kurdistan |
Tekriti 1960 |
Kani Shaie |
VIb |
4450-4300 |
Iraqi Kurdistan |
Renette et al. 2021 |
Gurga Chiya |
Level 2 |
4530-4340 |
Iraqi Kurdistan |
Wengrow et al. 2017; Carter et al. 2020 |
Begom |
Level 2 |
|
Iraqi Kurdistan |
Mühl and Nieuwenhuyse 2016; Nieuwenhuyse et al. 2016 |
Tell Ziyadeh |
EXVIII-XXII |
4582-4350/4490-4266 |
Northcentral |
Arzt 2001; Hole and Tonoike 2016 |
Tell Beydar III |
Levels 1-3 |
|
Northcentral |
Suleiman and Nieuwenhuyse 2003 |
Tell Mashnaqa |
Various |
4541-4353, 4491-4334 |
Northcentral |
Bizreh 2015; Hole and Tonoike 2016 |
Tell Leilan |
Vib |
|
Northcentral |
Schwartz 1988 |
Tell Kuran |
Late Kuran E |
4500-4250 |
Northcentral |
Hole and Tonoike 2016 |
Tell Kashkashuk III |
VIIA-VIB |
|
Northcentral |
Sultan 2019 |
Tell Feres |
Levels 8-7 |
4513-4353 |
Northcentral |
Baldi 2012a, 2015; Baldi and Abu Jayyab 2012 |
Tell Brak |
CH 18-20 |
|
Northcentral |
Oates 2012 |
Turbe Hoyuk |
LC1 Levels |
|
Northeast |
Sağlamtimur 2013; Sağlamtimur and Özan 2007 |
Tepe Gawra |
XIIA-XII |
|
Northeast |
Rothman 2002; Tobler 1950 |
Telul eth Thalathat |
Level VIIa-V |
|
Northeast |
Egami 1959; Shimogama 2019 |
Tell al Hawa |
Surface and Trench LP |
Northeast |
Ball et al. 1989 |
|
Tell Shelgiyya |
Trenches H-J and PS |
Northeast |
Ball 1997 |
|
Khirbet Derak |
|
|
Northeast |
Breniquet 1996; Forest 1987 |
Hamoukar SE |
Level 4 |
|
Northeast |
Abu Jayyab 2012, 2019; Baldi and Abu Jayyab 2012; Al Quntar and Abu Jayyab 2014 |
Hafuz Mevki |
C4, B4, D4 |
|
Northeast |
Ökse et al. 2015 |
Tell Zeidan |
different stratums |
4350-4230 |
Northwest |
Fisher 2017 |
Tell al ‘Abr |
Level 2 |
|
Northwest |
Hammade and Yamazaki 2006; Yamazaki 2012 |
Tell Kosak Shamali |
Level B5 |
4539-4334 |
Northwest |
Nishiaki and Oguchi 1999 |
Hammam et Turkman |
IVD |
|
Northwest |
Akkermans 1988a, 1988b |
Coba Hoyuk |
IVC-V |
|
Northwest |
du Plat Taylor et al. 1950 |
WS 386 |
TUC Survey |
|
South |
Adams and Nissen 1972 |
WS 218 |
TUC Survey |
|
South |
Adams and Nissen 1972 |
WS 178 |
TUC Survey |
|
South |
Adams and Nissen 1972 |
WS 042 |
TUC Survey |
|
South |
Adams and Nissen 1972 |
WS 022 |
TUC Survey |
|
South |
Adams and Nissen 1972 |
Tell al ‘Uwaily |
S36: C-B |
|
South |
Baldi 2020b; Calvet 1991 |
Uruk |
XII-XVI/II |
|
South |
Sürenhagen 1986 |
Telul al Baqarat |
TB7 |
|
South |
Lippolis et al. 2019 |
Eridu |
temple VI-IV |
South |
Safar et al. 1981; Perkins 1949 |
|
Tabara el Akrad |
Level VII |
|
West |
Hood 1951 |
Tell Afis |
Level 18 |
|
West |
Giannessi 2002, 2012 |
Hama |
L3 |
|
West |
Giannessi 2012; Thuesen 1988 |
14We look at presence/absence data of certain ceramic forms and decorations across three slices of time (5300-4900 BCE, 4900-4500 BCE, 4500-4200 BCE) roughly corresponding to the Ubaid 3, Ubaid 4, and LC1. The choice of using presence/absence data was made due to variation in data collection and publication strategies between the different sites.
15The way elements were selected was based primarily on how visible a form or decoration pattern was in site reports either from illustrations or how frequently it was reported. Based on this approach we started with over 120 elements or variables from each period; these included both decoration patterns and vessel forms. To avoid having extremely common or rare types govern our clusters, the elements collected were further culled by using the criteria that if an element was only present at or absent from three sites it was not selected. By these criteria, we ended with 43 to 55 elements for each period (see tab. 2).
Table 2 – Period by period counts of variables and percentage of variation each PC explains. Counts with (*) indicate some elements that overlapped in form and decoration analysis.
|
Ubaid 3 |
Ubaid 4 |
LC1 |
Number of sites |
26 |
43 |
38 |
Initial Variables |
84 |
87 |
102 |
Threshold Variables |
47 |
41* |
55* |
# of Forms |
19 |
19 |
37 |
# of Dec. elements |
28 |
24 |
21 |
PC1 Comp. |
41.26% |
36% |
41.83% |
PC2 Comp. |
21.61% |
21% |
26.3% |
PC3 Comp. |
13.6% |
14% |
20.48% |
PC1 Forms |
52.25% |
43.83% |
42.56% |
PC2 Forms |
19.19% |
21.5% |
26.8% |
PC3 Forms |
13.45% |
18.36% |
19.88% |
PC1 Dec. |
39.95% |
43.61% |
47.82% |
PC2 Dec. |
23.72% |
19.56% |
26.4% |
PC3 Dec. |
14% |
13.2% |
16.88% |
- 5 For example, a site such as Gurga Chia where the excavators referred to it as Late Ubaid, we chose (...)
16Determining the periodization of each level was the most difficult part of the analysis. To do so, we relied primarily on radiocarbon dates where available (tab. 1). When dates were not available, we used the periodization of the excavators or more recent chronological studies to date a site.5 Problematic sites were not included in the analysis, however, there remain a few sites that we are not completely certain of, highlighted in tab. 1.
17Multivariate analyses allow us to explore the relationship between multiple variables, otherwise undetectable visually in data spreadsheets or databases (Drennen 2009: 300). This is achieved through the reduction of variables in a dataset by combining them and presenting them in a visual format, which allows us to detect patterns and relationships a dataset. In our case, the aim is to access the relationships among sub-regions and sites through space and time based on the make-up of their ceramic assemblages.
- 6 The data at hand was processed using PAST (PAleontological STatistics software package), which is (...)
18The multivariate statistical method selected for this analysis was principal component analysis (PCA). PCA is a useful method for data reduction and for investigating group-specific patterning of data (Schillaci and Gunz 2013: 76), as it has no prior built-in assumptions about the data. It attempts to reduce the number of variables as much as possible without losing important aspects of the patterning in the original dataset (Drennen 2009: 300). PCA does not examine the relationships between the cases but rather the relationships between the variables present across the dataset (Drennen 2009: 300). The main principle is that variables that show strong correlation can be reduced to a single component that accounts for as much of the variance as possible within a multivariate dataset (Baxter 1994: 50, Baxter et al. 2008, Drennen 2009: 300; Hammer 2012: 79). Theoretically, the way PCs are obtained is by the reduction of multidimensional variables to a best-fit straight line running along a multidimensional scatter plot (Drennen 2009: 302). The variance of each PC is called an eigenvalue. The sum of the eigenvalues equals the sum of the variance of all variables in the analysis (Schillaci and Gunz 2013: 76). The first PC will have the highest eigenvalue, accounting for the greatest amount of variance in the dataset.6
- 7 Changes in the designations of the regions was necessitated by the sites included in the analysis (...)
19For our dataset, we are exploring the relationship between regions, and since sites were not always occupied throughout the three phases analyzed, the analysis was run between groups (the determined sub-regions mentioned above) rather than sites and levels. Therefore, the number of PCs generated was determined based on the number of groupings (see Hammer 2012: 78), which correspond to the number of sub-regions analyzed during each period (seven for early Ubaid, eight for Late Ubaid, and six for LC1).7
20In this study, we generated several biplots for each period first with the complete dataset (form and decoration) and then for form and decoration separately. We have two aims from applying PCA to the dataset: first, by generating a biplot for each period and comparing them, we can visually observe the changes in relational space to geographical space based on the relative positions of regions in the biplot. Second, based on the loadings of each PC, we can explore which vessel types were responsible for the observed changes. As each vessel category may indicate a mode of interaction within its own right, the interpretations of some of the observed patterns that we see here remain tentative and only intended to open new debates.
21Network analysis is a quantitative tool that provides us with an exploratory platform with which to detect, interpret and visualize patterns of relationships (Wasserman and Faust 1994; Freeman 2004; Borgatti et al. 2009). Network analysis allows us to consider relations among entities explicitly by applying graph theory to visualize and quantify interactions (Watts and Strogatz 1998; Albert and Barabási 2002). Rather than focus on entities in isolation, we can explicitly study the relationships among actors so that we can fully understand their individual behaviour. In this way, network models and theory can provide a unique perspective in archaeology by facilitating multi-scalar analysis and comparison of large regions composed of many interacting actors (Brughmans 2010, 2013; Knappett 2011, 2013).
22A network consists of a set of vertices or “nodes” and the connections or “edges” between them. In SNA, the nodes are actors, and the edges are relationships between pairs of actors. In this study, nodes represent archaeological sites and edges connect sites with similar ceramic assemblages. Similarity among ceramic assemblages was determined using the Jaccard coefficient. In this case, the Jaccard coefficient is the most appropriate similarity coefficient to use for presence/absence archaeological data because it is not affected by the shared absence of a particular trait (Habiba et al. 2018). This helps to reduce the impact of sampling error due to varied excavation strategies and publication records. In this study, edges are undirected, meaning the flow of interaction between sites is considered symmetric or equal, as the direction of interaction cannot be assumed without provenance studies.
23Edges are weighted using the results of the Jaccard similarity coefficient, a normalized value between 0 (no similarity) and 1 (exact similarity). A greater similarity between two sites means a greater weight for that connection (Peeples and Roberts 2013), and a thicker edge in the network visualization. Connections with extremely low levels of similarity (< 25%) were then systematically removed from the network model until reaching a minimum threshold of network connectivity, such that each site was connected to at least one other site (Golitiko and Fienman 2015). This eliminates background noise and allows us to focus analysis on the strongest and most intense connections.
24Once edges were established between nodes, we applied network measures to quantify both overall network structure and individual node properties. Results of these measures were compared with 1,000 randomly generated networks with the same number of nodes and edges to determine the statistical significance of any observed patterns (Butts 2008; Gjesfjeld and Phillips 2013). Across the three periods we found the results of network clustering and node eigencentrality to be significantly different than random (p ≤ 0.01).
25In our network visualizations, node clustering (modularity) is represented by colour, indicating archaeological sites that have more connections amongst each other than with other sites in the network (Blondel et al. 2008). Node size is proportional to each site’s eigencentrality score. Eigenvector centrality is a measure of node influence within a network, normalized between 0 (no influence), and 1 (most influence). Nodes with high eigenvector centrality have connections to other nodes that are themselves well connected. Finally, nodes are positioned geographically in our network visualizations to facilitate comparisons with geographic features and proximity to other archaeological sites.
26The PCA results from the three periods indicates that the robustness is close to the threshold from the first two PCs, and surpassing it when looking at the first three PCs. The first two PCs account for over 60% of the variation for the Early Ubaid (Ubaid 3) and LC1, while for Late Ubaid (Ubaid 4) the first two PCs account for 56% of the variation (tab. 2).
27The three different periods produced somewhat interesting results. The general pattern that emerges is an early formation of a shared Ubaid ceramic style during Ubaid 3 that gradually breaks apart throughout Ubaid 4 and continues to do so in LC1 (fig. 3).
Fig. 3 – Results of PCA analysis for (a) Ubaid 3, (b) Ubaid 4, and (c) LC1.
28For the Ubaid 3 ceramic interaction spheres we can make four general observations, First, we can see general overlap with differentiation in ceramic assemblages following strong spatial patterning from the southeast to the northwest (fig. 3a). Second, within spatial clusters, the strongest and by far the largest clustering occurs in an area encompassing an arc from southern Mesopotamia along the Zagros piedmont and into northern Mesopotamia. At the center of the graph are sites located in the Hamrin/Mandali subregion. These sites appear to be intermediary between southern and northern Mesopotamia. Specific sites from all three of these subregions show strong overlap (e.g., Tell al-Ubaid, Tell Abada, and Tell Arpachiyah). Third, the Balikh and upper Syrian Euphrates subregions seem to cluster separately, however, this could be misleading since we do not have data from geographically intermediary sites from the Khabur or lower Syrian Euphrates (the stretch from the town of Rawa to the confluence with the Balikh River) included in the analysis. Fourth, edges emerge on either side of the main cluster. Here we see two such peripheries: Ubaid sites along the Persian Gulf on the periphery of southern Mesopotamia, and sites in the west and northwest (Tell Kurdu and Hama) on the periphery of the northwestern cluster (the Euphrates and Balikh). Salat Tepe appears to be distinct from any of the aforementioned clusters. The overall pattern may indicate that a tightly knit “Ubaid” ceramic style emerges across a wide area during the late 6th millennium BCE over the course of three centuries.
29During Ubaid 4 we see increased separation with the emergence of three primary clusters: southern Mesopotamia, northeastern Mesopotamia, and northwestern Mesopotamia (fig. 3b). Northeastern Mesopotamia encompasses several subregions including Iraqi Kurdistan, the Jazira, and to a lesser extent the Hamrin. Northwestern Mesopotamia consists of the Khabur basin triangle, the Balikh, the upper Syrian Euphrates, Diyarbakir, and the west. Southern Mesopotamia remains central judging by the location along PC1 (which explains 30% of the variation), however, a large divide emerges between the northwest and northeast with the site of Tell Kurdu occupying a central role between them. Here the absence of data from a number of regions does not seem to play as important a role as observed for Ubaid 3, as the sites that comprise the separate clusters in the northeast and northwest were contiguous.
30During the LC1 we see further separation and the emergence of new clusters (fig. 3c): southern Mesopotamia, the Shahrizor plain, the Tigris River basin, the upper Khabur, the lower and western Khabur, and areas west of the Balikh River. Here we can note that the western and lower Khabur sites have stronger connections to areas in the west than with sites in the upper eastern Khabur. An interesting observation is that areas in the upper eastern Khabur with the sites of Tell Feres, Tell Leilan, and Tell Brak drift away from the northwest and begin to show closer affinities with northeastern sites (Tigris River basin). The same could be noted for the site of Telul eth Thalathat in the Tigris basin. These four sites become more intermediary between sites in the east and west as they share elements from both regions. Moreover, this centrality may indicate the significance of a communication route that connected the Tigris to the upper Khabur, south of the Sinjar range.
31From this initial exploration of the data, we can argue that in terms of ceramics, LC1 does not represent a dramatic overall break from the Ubaid cultural horizon, but rather the culmination of a gradual distancing that takes place over time. The Ubaid cultural interaction sphere seems to be at its strongest during its inception during Ubaid 3 and begins to show signs of clustering (three clusters) during Ubaid 4 with increase clustering during LC1 (five clusters).
32Comparing the combined form and decoration networks across the three periods (fig. 4) we see the emergence of several strong patterns. Over time we can see that ceramic interaction becomes more clustered, locally focused, and “small world”. Long-distance interaction is increasingly mediated by a small number of “central” sites, acting as bridges between local regions.
Fig. 4 – Comparison of combined ceramic form and decoration networks for all three periods. Node color represents ceramic group and node size is proportional to eigencenrality score. Line thickness is proportional to Jaccard similarity score.
33In the Ubaid 3 we see two large ceramic clusters, one primarily focused on interaction around the upper Euphrates (green), and another larger group along the Tigris (orange). The Tigris group is more clustered than the Euphrates group, maintaining more strong connections amongst themselves than the sites in the Euphrates group. The northern Tigris site of Telul eth-Thalathat is the most central to the network (eigencentrality score = 1), acting as a “bridge” connecting the upper Euphrates and Tigris networks. This could be due to its geographical positioning in the Sinjar area that could have been used as an overland travel route between the river systems. However, bridge sites are not integral to network structure during this period. Many sites maintain long distance connections between ceramic groups, particularly between sites in southern Mesopotamia and the upper Euphrates, likely connected by travel along the Euphrates River.
34In the Ubaid 4 we see the Tigris group break into two distinct ceramic communities, a southern group (orange), and a northeastern group (blue) based around the upper Tigris and Zagros. The southern group is more highly clustered and isolated than in Ubaid 3, maintaining fewer long-distance relationships with the northeastern Tigris and Euphrates groups. However, several outlier sites located along the Orontes and upper Euphrates also belong to the southern ceramic group (orange), including Tell Kurdu, Hama, Tell ‘Abr, and Mulla Matar. Interestingly, these sites maintain more connections to the northeastern Tigris (blue) and Euphrates groups (green), than to the south, but their strongest connections are to southern sites. The northeastern Tigris group includes sites from the upper Tigris and Zagros regions and maintains more and stronger connections with the upper Euphrates group than with the southern group. The locus of interaction for the northwestern group shifts from the Balikh region to the Khabur triangle. Again, we can see that Telul eth Thalathat is the most central to the network (eigencentrality score = 1), facilitating interaction between the Tigris and Euphrates, but maintains more connections to sites in the Khabur.
35In the LC1 we see an intensification of this pattern of increased local interaction, with the complete separation of the south (orange) from the north, in terms of patterns of ceramic consumption. The north also becomes more clustered, with a third ceramic group emerging in the east, focused on interaction along the Zagros (grey). The north is strongly clustered into two regions along the headwaters of the Khabur, with the eastern group (blue) more oriented toward the Anatolian highlands, and the western group (green) more oriented toward the Euphrates. Centrality shifts toward the Khabur region, with Tell Mashnaqa emerging as the most central (eigencengtrality score = 1), likely due to its position at the intersection of the Khabur and the Euphrates, meaning it can maintain strong connections to the Balikh in the west, and the Khabur in the east.
36Comparison of the combined form and decoration networks from each period highlights that the regionalization of ceramic production and consumption patterns seen in the LC1 had already started during the proceeding Ubaid 4. This is especially clear in the dramatic separation of the southern ceramic group beginning in Ubaid 4, with complete separation by the LC1.
37Splitting the dataset into form and decoration in the PCA produces more telling results (fig. 5). The overall pattern that emerges for Ubaid 3 indicates that decoration patterns were shared closer between regions than vessel forms.
Fig. 5 – Graphs of Ubaid 3 PCA Results (a) decoration and (b) form.
38The first two PCs in the decoration analysis account for nearly 64% of the variation, while the first three PCs account for 78% of the variation. In terms of form, the first two PCs account for 71.4% of the variation meeting the robustness test (tab. 2).
- 8 Tell el ’Uwaili is a variation on the spelling of Tell el ‘Oueili.
39It appears that decoration is responsible for creating the closer overall clustering seen during Ubaid 3, perhaps indicative of the emergence of shared aesthetic tastes and/or symbols across the region during this period. Southern Mesopotamia was at the center of this tight clustering (fig. 5a). Out of 28 Ubaid 3 decoration patterns included in the PCA, all the elements were present in southern Mesopotamia, while they were also found commonly across most of the regions. This is primarily responsible for giving the impression of a unified ceramic region across Mesopotamia seen in the overall PCA. The southern Mesopotamian sites of Eridu and Tell el ’Uwaili8, at the center of the graph, were very central in terms of decoration. The northwest subregion is slightly differentiated through the absence of three decoration elements: Xs in reserve (fig. 1O), lines in reserve (fig. 1D), and a ladder/accordion pattern (fig. 1R).
40The picture changes when the analysis was conducted using only forms. Here there seems to be a clear separation between the northwest and southern Mesopotamia (fig. 5b). Out of 19 forms, there are six forms that were completely absent from southern Mesopotamian sites (fig. 1: 1j, 5p, 3j, 5b, 11b, 13b). This absence is what causes the slight regional separation between southern Mesopotamia and other parts of the Near East when forms are viewed independently.
41This implies an overall Ubaid style strongly attested in decoration patterns, and to a lesser extent in forms. The slight variation in forms could indicate that there remained some clear technological and functional boundaries, permeable to decoration, between southern and northwestern Mesopotamia during this timeframe.
42Ceramic data was separated into form and decoration networks to investigate if specific features were creating the patterns observed in the combined networks. For the Ubaid 3, we can see that the form network is more clustered than the decoration network (fig. 6).
Fig. 6 – Ceramic form (left) and decoration (right) networks for Ubaid 3 period.
43In the form network there is more separation along the Tigris, creating two form groups (orange and blue), with the blue group containing the more eastern sites, potentially connected by relationships with the Zagros mountains, and established historical relationships from earlier periods (such as Tell Arpachiya and Choga Mami). Centrality is centered in the south Hamrin area, at Ras al Amiya, and the Tigris group in general (orange) is very strongly connected. This indicates these sites are sharing very strongly in a similar repertoire of forms (potentially eating/cooking similar foods or sharing in patterns of commensality). In comparison, the Euphrates form network (green) is less connected, with each site maintaining slightly more distinct form repertoires from each other than the Tigris groups.
44In contrast, centrality in the decoration network is balanced between the north at Tell Kosak Shamali, and the south at Eridu (both sites have eigencentrality scores of 1). The network is less clustered with sites maintaining more and longer distance connections, sharing more openly in decoration repertories. This suggests that while the way food is cooked, served, and stored may be slightly different between north and south, the visual information it communicates is much more similar.
45Ubaid 4 follows a similar trend to Ubaid 3, albeit with increased separation between the subregions. In terms of decoration, all the regions seemed to have clustered closely, with southern Mesopotamia falling in the center of the graph. The picture is different when we examine forms. Here we see the emergence of three distinct clusters (fig. 7): the first is composed of sites from southern Mesopotamia, the Hamrin, and Iraqi Kurdistan; the second of sites in northern Iraq, and the third of sites in southeastern Turkey and areas to the west of the Khabur River.
Fig. 7 – Graphs of Ubaid 4 PCA Results (a) decoration and (b) form.
46For decoration, the first three PCs account for 83% of the variation. Most decoration elements that were present in the Ubaid 3 continue, with all elements again present in southern Mesopotamia and across the different regions. Two new decoration elements were found exclusively in the Khabur and the Diyarbakir areas; these were horns/eyes (Ai), and circles in reserve (Z; fig. 1). These two patterns represented new symbols on traditional Ubaid forms that occurred alongside traditional Ubaid decoration in these subregions. Another pattern common in the north and northwest, but rare in the south, was the stringed hatched diamond pattern (fig. 1V).
47In Iraqi Kurdistan Dalma incised and impressed wares (fig. 1) and incised decoration patterns (fig. 1U) show affiliation with northern and eastern sites rather than the south. Dalma impressed wares were present in the Hamrin during the previous period, but they seem to become part of the assemblages in the Erbil and Shahrizor plains during the mid 5th millennium BCE. In both the Khabur and Iraqi Kurdistan, elements were shared between those regions and adjacent highland regions. In the first case, there were close relations between the Khabur and the Tur Abdin, and in the second case there appeared to be some connections between Iraqi Kurdistan and the Zagros (see Alden et al. 2021). This increased connection with the highlands could partially explain the separation that is seen in the overall PCA, with sites slightly moving away from the Ubaid interaction sphere and pursuing closer ties with the highlands.
48In terms of vessel forms, the first two PCs account for 66% of the variation while the first three PCs exceed the robustness threshold accounting for 87% of the variation. Here we see the introduction of eight new forms out of the 19 forms that met the criteria for analysis. Common across all regions were holdover types that began during the Early Ubaid, including beaded globular jars (5b), early double rim or internal ledge jars (2p), deep Ubaid bell-shaped beakers (3b), pedestalled bowls (2b), bowls with an internal lip (11b), and neck handled Jars (3j; fig. 1).
49Some forms persisted in southern Mesopotamia but seemed to have been absent from subregions further to the north such as low carinated fine bowls (15b; found also in the eastern region), extended rim carinated bowls (usually decorated; 14b), spouted jars (7p), four lugged jars (6p; also part of the Hamirn assemblages). Iraqi Kurdistan shared some forms with southern Mesopotamia and the Hamrin such as coarse trays, while exclusively sharing Dalma impressed and incised jars with the Hamrin.
50In the Northwest (Upper Syrian Euphrates, Balikh, Khabur) we see the greatest number of new forms, these include Coba bowls, red slipped jars, and internally incised bowls (fig. 1). Double mouth jars (11p) were the only type that seemed to have been unique to the Northeast. Several vessels were shared among regions outside of southern Mesopotamia, including deep urns (9p), globular bowls with a strong incurved rim (9b), and S-shaped beakers (7b).
51Finally, an interesting phenomenon we observed is that some forms that were only found in northern regions during Ubaid 3 appeared in the south during Ubaid 4. This may show that the Ubaid interaction sphere was not as unidirectional as we initially imagined it to be. This bidirectional borrowing was observed by Baldi (2020a: 84) for an earlier timeframe.
- 9 . – For the structural implications of the use of mass-produced bowls see Baldi 2012b and Kennedy 2 (...)
52During the Ubaid 4 we see that potting practices were clearly diverging, especially between the northwestern and southern subregions. This is not unusual but implies that the strong transregional interaction attested during the Ubaid 3 began to give way to more localized interactions and innovations. This is supported by the formation of tightly knit ceramic spheres that formed between the Khabur and the area of Diyarbakir. These newly emergent ceramic spheres were not simply a result of borrowing or mimicking certain motifs or vessel forms but seemed to have resulted from new social and structural overlap within these regions manifest in the spread of mass-produced Coba bowls.9
53The separated form and decoration networks for the Ubaid 4 show an intensification of the patterns seen in the Ubaid 3 networks (fig. 8), with more sharply defined cluster groups.
Fig. 8 – Ceramic form (left) and decoration (right) networks for Ubaid 4 period.
54The Ubaid 4 form network divides into three clusters, a northwest group centered on the Khabur (green), a northeast group centred on the upper Tigris and Zagros (blue), and a southern group (orange). These clusters are stronger and more sharply defined than in the proceeding Ubaid 3 form network. This means sites are interacting more intensely and exclusively with their direct neighbors, with less long-distance interaction between cluster groups. This pattern can be most dramatically seen in the partial separation of the southern group—suggesting the south was maintaining a more distinct repertoire of forms.
55Long-distance interaction between form groups is channeled more through bridge sites or central sites than during the Ubaid 3. This can be seen most apparently at the site of Tell Leilan, facilitating interaction between the northeastern and northwestern groups. The centrality of Tell Leilan could be influenced by its geographical positioning between the Tigris and Euphrates, but potentially also to the Anatolian highlands, enabling the movement of Anatolian forms into the Khabur.
56In contrast, the Ubaid 4 decoration network is much more interconnected, with clusters not as sharply defined (more connections between clusters), suggesting more sharing of decorative motifs between community groups. Interestingly, the south is not its own cluster in the decoration network (like we see in the forms) but represents a combination of the decoration clusters we see in the north. This suggests that the south is maintaining a distinct repertoire of forms, but actively engaging in the diversity of visual communication in the north. The lack of interconnection in decoration between southern sites also defies the pattern of geographic proximity influencing ceramic similarity and seems a deliberate act of difference.
57In terms of centrality, Tell Qalinj Agha in the northeast is the most central as it maintains numerous connections to sites to the west and south. The centrality and highly connected nature of this northeastern decoration group could be influenced by relationships to the Zagros, facilitating the movement of decoration patterns from the highlands into the west.
58The separation in ceramic trends is more visible during the LC1 period. This represents a continuity within the overall trend of regions drifting further apart. The difference here is that this separation is visible in both decoration and form (fig. 9a-b). Here we see the emergence of four regions rather than the three observed during the Ubaid 4. Southern Mesopotamia has disconnected entirely from the northern regions, especially in the ceramic forms. Further, the first two PCs account for most of the variation observed in both the decoration (74.2% of the variation) and form analysis (69.3% of the variation; tab. 2).
Fig. 9 – Graphs of LC1 PCA Results (a) decoration and (b) form.
- 10 . – The sites of Tell Brak, Tell Feres, and Tell Leilan in addition to Telul eth Thalathat, south o (...)
59Some very general form and decoration patterns persist from the Ubaid 4 during the LC1 across the study area (fig. 1C, E, N, Q, Aj, 2ba, 5b, 1j, 13p). In terms of forms, the most visible distinctions take place between southern Mesopotamia, the Tigris basin, and the northwest (western and lower Khabur, Balikh and Euphrates basins). The emergence of new forms that appear to be restricted to southern Mesopotamia (fig. 1: 3u, 4u, 5u, 6u, 7u, 8u, 9u) strongly impact the PC loadings (negatively in both PC1 and PC2). This has a negative correlation along both PCs with the new forms that were prominent in the Tigris basin (fig. 1: spectacle idols, grey/red holemouth pots [HMPs], and wide flower pots [WFPs], 6p, 7j, 9b, 4b, 9p,). Prominent northwestern types (fig. 1: dark flaring rim cooking pots, internally incised bowls, Coba bowls, beaded Coba bowls, red-slipped jars, 6j) correlate positively to the Tigris types along PC1 and negatively along PC2. These correlations create the pattern we see in the graph (fig. 9b) with clear separation taking place between the three regions. The types that distinguish the northwest were also prominent during the Ubaid 4 (with the exception of 6j). This implies strong continuity within this area after a period of innovation that took place during the Ubaid 4. The eastern Khabur and the Shahrizor share many elements with the Tigris basin. Sites in the eastern Khabur during this period10 share elements from both the Tigris basin and the northwest. While these sites were strongly linked to the northwest during Ubaid 4 and carried on producing many of the elements that characterized the northwest in the Ubaid 4 and LC1 (internally incised bowls, Coba bowls, red-slipped jars, 6j) they also incorporated elements that were characteristic of the LC1 of the Tigris basin (spectacle idols, grey/red HMPs, WFPs, 6p, 7j, 9b, 4b, 9p). Finally, in both the northwest and the Shahrizor plain we see continuity with many elements that emerged during the Ubaid 4 and remained important to these regions without spreading into southern Mesopotamia or the Tigris Basin. This stands in contrast to the dramatic changes seen in southern Mesopotamia and the Tigris basin.
60What distinguishes the Ubaid 4 from the LC1 is the divergence in decoration attested during this period. Divergence in decoration patterns during the LC1 mirrored the divergence attested in form. Some decoration persists throughout: painted black on buff decoration persists in the northwest, Shahrizor and southern Mesopotamia (fig. 1E, J, Q), and incised decoration persists in the Shahrizor (fig. 1: incised wares, Aa, Ac). However, we also see southern Mesopotamia and the Tigris basin clearly distinct from one another. In the south, three unique patterns absent from other regions emerged. These were rope applique, elevated ridges, and Ad (fig. 1).
- 11 . – Results of their work presented in the workshop on which this volume is based, 12th ICAANE 2021 (...)
61The Tigridian region was perhaps the most innovative in decoration during this period. In the Tigris basin sprig wares and black on red wares represent new ware groups that emerged during LC1 (fig. 1Ab, Bb, Cb, Eb, Jb, sprig wares). Additionally, pattern (I) also represents a new pattern that is common at several sites in the Tigris basin. Some of the new decoration patterns found in the Tigris basin were also found at sites in the eastern Khabur (Tell Brak, Tell Feres, and Tell Leilan) without crossing further west. This reinforces the notion that this part of the Khabur was a boundary zone that potting practices did not easily cross. Analysis of sprig and black- on-red wares (Rothman and Blackman 2003; Abu Jayyab 2019) point to a Tigridian origin for these vessels. More recently, Sconzo and Amicone11 have shown that similar patterns executed on buff fabrics were also part of the same Tigridian industry. This is supported by the presence of similar decoration patterns in both black on red and black on buff in the Tigris basin.
62The Tigris network during this period extends towards the Taurus highlands with sites along routes to the highlands, such as Hafuz Mevkii, and Türbe Höyük, showing close affiliation with sites further downstream. This divergence from the Ubaid 4 and the dissolution of the Ubaid network of interaction could have been caused by increased interaction with the highlands, a theme that follows from the previous period for the northwest and Iraqi Kurdistan.
63The eastern Khabur becomes a very important region during the LC1. Sites such as Tell Feres, Tell Brak, Tell Leilan, and to an extent, Telul eth Thalathat became intermediary between the Tigris region and regions further west. Here we see the eastern Khabur sharing in both forms and decoration patterns from both the northeast such as sprig wares, black on red wares, and WFPs, and the northwest such as Coba bowls, internally incised wares, and red slipped wares. This implies that this region acted as a boundary zone within the northern Jazira that perhaps controlled the flow of information and innovation between the east and the west (see next section).
64Separated forms and decoration networks for the LC1 again show general intensification of the patterns seen in the Ubaid 4, with more sharply defined clusters (fig. 10). However, during this period we see more alignment between the form and decoration clusters than we saw in earlier periods. This could suggest that during the LC1 form and decoration are transmitted or learned through entwined socio-economic mechanisms, whereas in earlier periods they may have been transmitted through separate mechanisms.
Fig. 10 – Ceramic form (left) and decoration (right) networks for Late Chalcolithic 1 period.
65In the LC1 form network, we see the emergence of four strongly regionalized form groups. The south (orange) has now completely separated from the north, and the north has become even more strongly clustered, forming two ceramic groups divided along the headwaters of the Khabur. The northwest group (green) includes the upper Euphrates and western Khabur sites, while the northeast group (blue) includes the eastern Khabur, upper Tigris and Anatolian highlands. Interaction between these clusters is facilitated by ‘bridge’ or central sites, all located within the Khabur headwaters, with Tell Beydar III emerging as the most central (eigencentrality score = 1). While Tell Beydar III is the most central, other sites located in the Khabur headwaters, including Tell Ziyadeh, Tell Mashnaqa, and Tell Feres, also have very high centrality scores (≥ 0.85), indicating this region as a whole was essential for maintaining connections between east and west Mesopotamia. The fourth form group includes sites located along the Zagros highlands in Iraqi Kurdistan (grey), with interaction to the west facilitated by the bridge site of Tell Qalinj Agha.
66Again, we see that the decoration network is more connected and less clustered than the form network, with only three decoration clusters, compared with the four clusters of the form network. The southern group (orange) is pulled back into the network through similarity in decoration to the northeast Iraqi Kurdistan region—which was its own cluster in the form network but combined with the south in the decoration network. Interestingly, the southern decoration group has a few outlier sites in the Khabur, Tell Beydar III and Tell Ziyadeh, both sites that were extremely central in the northwest form network. The northwest (green) and northeast (blue) decoration clusters closely match the northwest and northeast form clusters, though with less distinct division along the upper Khabur, suggesting more sharing of decoration repertoires than forms in this region. Decoration centrality during the LC1 shifts from the Khabur region in the form network toward the Mediterranean in the decoration network, at Tabara el Akrad – suggesting more influence from the northwest for decoration during this period.
- 12 For example, data from the stretch of the Euphrates from Jebel Bishri to Rawa is completely absent (...)
67Any future research on interaction networks during the periods studied must consider the absence of data from a few sub-regions, such as the Khabur during Ubaid 3 and the lower Syrian Euphrates throughout the three periods.12 Filling this gap in the dataset and further refining the periodization (see Volpi this volume) has the potential to expand on or change some of the observations presented here. These gaps in our knowledge during the Early stages of the formation of the Ubaid interaction sphere severely limit our understanding of how interaction networks emerged and the nature of connections between east/southeast and northwest.
68Another caveat worth mentioning here is that elements of culture did not necessarily circulate in unison, therefore the ceramic data discussed in this paper only represents one aspect of material culture in an otherwise much broader cultural phenomenon. While the sharing of ceramic elements is within itself varied and indicative of overlapping processes (Abu Jayyab 2019, in press), exploring other elements of material culture in conjunction with this data has potential to reorient these networks. However, ceramic interaction networks constructed through PCA and network analysis in this study allow us to address the Ubaid/LC1 phenomenon across three different scales:
- at the macro-scale through addressing the changing structure of the Chalcolithic network overtime;
- at the meso-scale through addressing the role of sites and sub-regions within the network;
- at the micro-scale through addressing ceramic assemblage formation at the site and subregional levels.
69One of our main aims from this study was to understand whether changes between the Ubaid and LC1 were gradual or abrupt. Initially, during Ubaid 3, we observed the emergence of two ceramic communities with close ties to one another. The ties were much stronger for decoration patterns then they were for forms. This tightly knit network gives way during the following Ubaid 4 period to three ceramic communities, followed by four (network analysis) to five (PCA) during LC1. The Ubaid interaction sphere, at least from the perspective of ceramics, was at its strongest during the Ubaid 3 and gradually dissolved throughout the Ubaid 4 and LC1 (fig. 11). In other words, LC1 was as much of a break from Ubaid 4 as Ubaid 4 was from Ubaid 3.
Fig. 11 – Maps showing community clusters with characteristic ceramics during the Ubaid 3 (a), Ubaid 4 (b), and LC1 (c).
70The wide-spread shared practices and gradual breakdown of an Ubaid interaction sphere fit the model proposed by Hole (2016) on the spread of Ubaid cultural practices in the lower Khabur. The tighter knit connections observed during Ubaid 3 and the increased modularity observed over time fit well with the notion of an initial southern Mesopotamian migration that diverged over time due to the adaptation to new circumstances (environmental and demographic) and amplified by the distance away from the homeland (Hole 2016: 392-394). Though this hypothesis could explain part of the picture that led to the emergence of a strong Ubaid 3 network, it is difficult to envision a mass migration from the south that led to a complete cultural change. Alongside migration several other factors were probably at play. If we consider that interactions between the southern Ubaid world and the surrounding regions began at an earlier period than Ubaid 3, as suggested by Baldi (2020a; also see Campbell and Fletcher 2010), then the data we see from Ubaid 3 represents the zenith or largest extent of the interaction sphere rather than its point of emergence. In this case, we may posit that the emergence of an Ubaid interaction sphere was formed through several processes that included migration, adaptation and adoption of a wide range of techniques and practices (see Stein 2010). This, however, is hard to address in full detail through ceramics alone (ibid.).
71Overall, the networks show increased modularity and centrality in potting through time. Meaning the emergence of more clusters of sites that interacted more with their surroundings than with the outside world. This phenomenon is coupled with the emergence of bridge sites that mediated interactions between these various emergent clusters. This observation is true for all three levels of the network analysis: the overall network, the form network, and the decoration network.
72Initially during the Ubaid 3 the overall network was characterized by extremely high clustering and short average path length. In network terms this represents a small world network, albeit not in its most efficient shape with plenty of redundant connections maintained (triangles between sets of sites). In this period networks were more ‘egalitarian’, that is most sites communicated directly with any other site in the network without being channeled through an intermediary site. This means that most sites are direct ‘neighbours’ of one another with more or less direct access.
73During the Ubaid 4 we see that form and decoration networks behave in a different way. The Ubaid 4 decoration network is much more interconnected, with clusters not as sharply defined, suggesting more sharing of decorative motifs between community groups. Interestingly, the south is not its own cluster in the decoration network (in contrast to forms) but represents a combination of the decoration clusters we see in the north. So, the south, while slightly more conservative in terms of forms from other regions, actively engaged in the diversity of visual communication in the north. This lack of interconnection in decoration between southern sites defies the pattern of geographic proximity influencing ceramic similarity and could have been a deliberate act of creating difference. In terms of centrality of the decoration network, we see that Tell Qalinj Agha in the northeast is the most central as it maintains numerous connections to sites to the west and south.
74In contrast, in the form network we see strongly defined clusters with more direct interactions with neighbors than across clusters. Any long-distance interaction between clusters is increasingly channeled through bridge or central sites. This is seen most apparently in the role that Tell Leilan played in facilitating interaction between the northeastern and northwestern groups. The centrality of Tell Leilan could be influenced by its geographical positioning between the Tigris and Euphrates, but potentially also its location relative to the Anatolian highlands, enabling the movement of Anatolian forms into the Khabur.
75Telul eth Thalathat plays a central role in the network starting from Ubaid 3 and continues to do so in the Ubaid 4. Telul eth Thalathat was a unique site; in the form network it clearly had stronger ties with the Northwest, while in terms of the decoration network it belonged in the Northeastern group. This bidirectional influence on the site could be an indication of the significance of the land route south of Sinjar that connected the Mosul/Makhmour area to the Khabur.
76The LC1 represented an intensification of the patterns seen during Ubaid 4. Here we see increased clustering on all levels. Very few interactions took place across northern regions that were not facilitated by a bridge site. Despite the presence of some connections between the south and other regions through some shared decoration patterns, the south was completely separated in terms of forms. The center of gravity during LC1 shifted completely from the south to the Khabur where the region became highly centralized connecting the east and the west. Despite Tell Mashnaqa having the highest centrality score in the LC1 full network, several sites within the Khabur had overlapping traditions between sites in the east and the west, such as Tell Beydar III and Tell Feres. An interesting pattern worth noting here is the significant role Tell Feres, Tell Brak, and Tell Leilan play in the network. In the PCA these sites seem to fall between the sites in the Tigris basin and those in the west while Tell Leilan and Tell Brak switch network affiliation in form and decoration networks between east and west (see fig. 10).
77The network topology during LC1 points to increased hierarchization through the emergence of clear bridge sites mediating and controlling flow of potting practices. It seems that the eastern Khabur acted as a mediation/boundary zone in terms of ceramic styles between east and west. This position in the network could be an indicator of the significance of the eastern Khabur for interregional interactions and a factor in the growth of sites such as Tell Brak, which becomes a significant site during LC2 and LC3 as a proto-urban and urban settlement (see Ur et al. 2007). This conclusion supports Baldi’s assertion that new regional centers may have contributed to the dissolution of the Ubaid system of territorial villages (see Baldi 2012b: 405).
78The structure of the networks presented here was based on the ceramic assemblages from the various sites. In a broad sense, assemblages in the study regions were composed of underlying Ubaid elements supplemented by highly localized and regionally-specific wares and types. The increased clustering witnessed in the networks over time is governed by an increase in local wares relative to underlying Ubaid ones. Regardless of how the network was conceived, several processes pertaining to assemblage formation were at play within the network. We can imagine that some of the shared Ubaid decoration patterns resulted from active signaling of affiliation (Wobst 1977), represented a cognitively undetected blueprint (see Lechtman 1977: 4), or resulted from mimicking. It remains difficult, however, to categorize the widespread patterns in our study, and beyond the scope of this paper. What we can explore are elements that seemed to have been highly localized. Some decoration patterns such as circles in reserve and horns were patterns executed on classical Ubaid forms but were only part of the ceramic repertoire in the Khabur and Diyarbakir sub-regions. Similarly, sprig wares and Dalma impressed wares had limited distribution in the network compared to other forms of decoration and only appear in the Tigris and Iraqi Kurdistan respectively.
79The appearance of Dalma incised and impressed wares in Iraqi Kurdistan, and circles in reserve and a particular type of horn decoration in the Khabur, points to engagement with adjacent highland regions alongside the participation in the supra-Ubaid network. This localization process begins to emerge in the Ubaid 4 in the Balikh, the Khabur, and Iraqi Kurdistan, and stabilizes in the LC1 (see Kennedy this volume). While these regions show more gradual developments from Ubaid to LC1, new innovative breaks in potting practices occur in the south and the upper Tigris during LC1. The upper Tigris during the LC1 shows a sharing in potting practices across its basin, following in the pattern of increased interaction with the highlands, seen during Ubaid 4.
80In terms of forms, we see that some forms characteristic of Ubaid assemblages first appeared in the north and found their way to the south. This shows that the Ubaid phenomenon was not as unidirectional as imagined (see Campbell and Fletcher 2010; Carter and Philip 2010).
81Starting from Ubaid 4 new structural and functional changes emerge based on the emergence of new pottery forms. This is clearly manifest in the proliferation of red slipped wares in the west, and the appearance of internally incised wares and grey flaring rim cooking pots across both the Khabur and western regions. The deepest structural divide, however, could be seen in the emergence of mass-produced bowls. The presence of mass-produced Coba bowls in the north (the Khabur, Diyarbakir, and western Syria) during Ubaid 4 implies the emergence of a clear structural separation, perhaps tied to new practices of feasting and organization of labor (see Baldi 2012b; Kennedy 2012, 2019). The emergence of WFPs in the Tigris Basin and Iraqi Kurdistan during the following LC1, while splitting the network in terms of vessel types, implies perhaps the emergence of shared practices across a broader region taking place in LC1.
82Our analysis shows that the Ubaid interaction sphere is at its height during Ubaid 3, gradually separates during Ubaid 4, and continues to separate during LC1. Many factors could be noted to explain the reasons behind this process, such as different sub-regions engaging in more localized interactions, sub-regions focusing on interactions with neighboring highland regions, increased settlement hierarchization within the sub-regions, and the emergence of new localized structural adaptations. When viewed from a southern Mesopotamian perspective, Ubaid interaction networks seem to be breaking down gradually. However, if we examine each region on its own terms, rather than from the lens of an Ubaid interaction sphere, we see this process as more of a continuous reconfiguration of interaction spheres cumulating in the more expansive LC2 and LC3 interaction spheres that follow.
- 13 See Campbell 2007: 105; Carter and Philip 2010: 13.
83Baldi (2016) aptly described the Ubaid to LC1 transition as a move from apparent homogeneity to a superficial fragmentation. What we show here is that a clear delineation of Ubaid and LC1, at least from a ceramic network perspective, seems to be elusive. Different sub-regions in the so-called Ubaid interaction sphere distanced themselves or were distanced from the overall Ubaid network at different junctures, and as such, a complete breakaway is near impossible to pinpoint. If the Ubaid period is not as ceramically homogeneous as perceived, and does not end as synchronously as expected, then we cannot justify the reification of the Ubaid and a post-Ubaid/LC1 period from a ceramic perspective, where such a break is somewhat arbitrary. In other words, a complete break between the two periods stems from the internal bias predisposition inherent in cultural historic terms such as Ubaid13 and the role it plays in prejudicing such comparative work (Baldi 2016: 118), especially when placed against terms such as post-Ubaid or indeed LC1. This terminology gives the impression that ‘Ubaid’ had an internal cohesion that stood in sharp contrast to preceding and succeeding periods. Though we still employ terms such as Ubaid and LC1 in this work, we built an analysis from the bottom up that aimed to detect some subtleties within the Chalcolithic of Greater Mesopotamia.