Abstracts of the Network Perspectives in the Archaeology of the Ancient Near East and Eastern Mediterranean workshop

The abstracts of the Network Perspectives in the Archaeology of the Ancient Near East and Eastern Mediterranean workshop 13th and 14th of December 2021.

Pottery studies have been part and parcel of archaeology since its origins as a scientific discipline. This long research history has resulted in standardized approaches to (macroscopic) pottery analysis focused on sets of observable parameters, such as for example in the Peacock classification system of modes of production (Peacock, 1977, Pottery and Early Commerce). Traditional material studies often focus on the establishment of two main axes of structuration: 1) typology; and 2) chronology, often pursued in combination as typo-chronological frameworks. For this purpose, endless heaps of material are amassed, described and compiled in voluminous catalogues. Especially in Mediterranean and Classical archaeology, where scientific methods such as 14C dating continue to wander largely on the fringes, traditional studies of material artefacts often remain the alpha and omega of the disciplinary practice. The immense merits of these traditional approaches for the development of reliable frameworks of interpretation as the foundation of archaeological reasoning cannot be overstated. Yet, it is becoming increasingly clear that the field is suffering from decreasing returns on investment as the net gain of every new publication gradually dwindles towards zero.

 The recent upsurge in computational approaches in archaeology has produced a fresh breeze that is invigorating the stale state of affairs in material studies. In this paper, I will present two novel approaches to material culture based on network science. Each of these case studies addresses one of the two main goals of material studies. The first covers the effects of time averaging on pottery chronologies using a dataset of Hellenistic and Roman tablewares from the Eastern Mediterranean. The second discusses the usage of morphological similarity networks for creating typological classifications, using a dataset from Late Bronze Age to Hellenistic ceramics from the area of Sagalassos (southwest Turkey).

The aim of this paper is not to advocate for the replacement of traditional material studies with computational approaches. Instead, through these two case studies, I will show the potential of network approaches for archaeology in general, and Mediterranean archaeology in particular, in tandem with traditional artefact studies and material specialists. The end goal of this approach is to develop a disciplinary framework where archaeological and computational approaches mutually reinforce each other through their respective strengths of rich interpretability and analytical tractability. Computational archaeology can only truly descend into the core of our discipline if it makes a serious effort to do justice to the works of material specialists. This paper aims to provide an outline for a potential path forward in this vein.

In this paper, I explore the potential for the development of a theory of state formation informed by recent advances in network science. Specifically, I argue that settlements situated along the overlapping boundaries of multiple communities of practice experienced greater inter-site connectivity compared to core settlements belonging to communities with more limited diversity. That is, settlements located at the boundaries of multiple overlapping communities of practice were afforded greater opportunity for cultural interaction and exchange, and this contributed to their subsequent development into the core or ‘capital’ cities of newly founded polities.

This network approach to state formation helps explain, in part, the phased and pulsating nature of urbanism in the Bronze and Iron Age Levant by specifying a mechanism through which urban centers and political capitals moved across the landscape in a cycle of peripheries turning into cores, and cores into peripheries. As preliminary test of this approach, I examine the case of overlapping ceramic communities in the Northern Levant during the second half of the third millennium bce, and reflect upon the rise of Ebla as a hegemonic state around this time.

I will discuss two projects to demonstrate the unique utility of network analysis. These projects, which I developed over a decade apart, were initiated out of opposite circumstances. First, my earliest foray into network research was a project I started by accident in 2006. In this case, my quantitative data led me to seek mathematical collaboration and pursue digital methods. Specifically, as an art historian, I had collected measurements of ancient Levantine ivory sculptures with the goal of determining possible canons of proportions and revealing prototypes of ideal feminine beauty. To do so, I would need to calculate, compare, and analyze thousands of ratios—a task that turned out to be impossible to manage by hand. I therefore sought help from mathematical experts, who recommended we use data mining to discover clusters of proportional prototypes. While we did find some interesting information relating to proportions, what I did not expect to find were that some of the clusters overlapped with interpretations of the much-debated regional workshops where the ivories may have been produced. Other outcomes added new possibilities to established interpretations. I, thus, surprised myself with output that pointed to social and production networks. Through this serendipity I embraced the value of network analysis, not only for its known potentials, but also for its potential (much like field archaeology) to reveal the past in unexpected ways.

More recently, and in a research culture where network analysis is much more common and encouraged, I have intentionally initiated a network analysis project. In this case, I was intrigued by the output of a Helsinki-based social network analysis (SNA) project to analyze networks among Neo-Assyrian deities (Alstola et al., JCS, 2019). The results of this project suggested that the god Assur lay at the center of the network, and result that reinforced the importance of the god Assur to the Neo-Assyrian king. Myself being interested in the religious networks of queens, and having previously had a very productive interdisciplinary collaboration on my ivory project, I proposed that we join forces (which entails joining data, research questions, differing expertise, and labor) to examine the royal-divine networks of kings of queens. Instead of starting from scratch, we are adapting the methodology and tools developed for the initial analysis of divine networks to support the analysis of royal and divine data as well as new questions. The SNA results so far have reinforced and nuanced our established knowledge, and revealed new perspectives and exciting foundations for continued research.

Through these two very different projects and experiences, I am convinced that network analysis has a unique utility, and that complex network analysis projects are best pursued collaboratively. I also assert, however, that we must acknowledge that the data going into network analysis, the management of the data, and the interpretation of the results must be contextualized within established scholarship (in order to measure, compare, and apply it), supported by qualitative research (in order verify and expand interpretations), and balanced with common-sense (in order to weed out faulty data and computational decisions). Finally, I encourage us to contextualize the results of computational network analysis in mainstream qualitative scholarship by presenting and publishing in conventional venues, as well as in necessary, specialized places of technical discussion such as this one.

 History of scholarship on the Late Neolithic Levant is full of gaps and disagreements. This is particularly the case for the later part of the Late Neolithic (5700 – 5300 cal. BC), characterized by the Wadi Rabah ‘culture’ in the south (normative and variant) and Néolithique moyen, Amuq C and Halaf ‘cultures’ in the north. While there are differences among these regional groups, there are also significant similarities, including the use of some pottery forms, decorations, iconography, and raw material sources. The similarities between the northern and southern Levant, and east and west of the Jordan Valley have been documented and discussed in some detail, but there has been no systematic, large-scale comparison of material culture from these regions. In this regard, network analysis is a useful tool for directly analyzing and visualizing the multivariate nature of material interactions among archaeological sites at multiple scales. The individual characteristics of each site can be examined in reference to their relationships with other sites, which allow for the recognition of larger-scale patterns of interaction.

 In this case study, network analysis is used to model material similarity among 43 Levantine Late Neolithic sites to explore community membership, and the nature and intensity of interaction among community groups. Material similarity was determined through the presence/absence comparison of ceramic typological data (24 form types, and 58 decorative elements) and iconographic motifs (such as on anthropomorphic figurines and seals). Presence/absence data collection was necessary due to the disparate level of detail in the excavation and publication records of Levantine Late Neolithic sites. Ceramics and iconographic elements were chosen for analysis because they represent the most common artifacts recovered during excavation and the most widely reported. Moreover, it has been shown that sharing in elements of ceramic production and consumption implies a strong degree of communication, an indication of shared socially acquired dispositions, and potentially even face-to-face learning.

The Jaccard coefficient was used to evaluate assemblage similarity, assigning a value between 0 (no similarity) to 1 (the same) for the comparison between each assemblage. This value was then used as the edge weight, linking each site by the strength of their similarity. Network measures including modularity (Louvain community detection) and eigencentrality were applied alongside exploratory techniques such as node and edge deletion to investigate community structure and resilience. Node and edge deletion are two related methods of deconstruction that identify a network’s ‘core’—the integral relationships that structure the network. Edge deletion systematically removes weaker ties based on edges weighted by the results of the Jaccard similarity coefficient. In removing the ‘weak links’, the strongest relationships in the network can be more easily identified. Node deletion includes two techniques: hub-removal and k-core deletion. Hub removal systematically deletes the most well-connected nodes, providing a test for network resilience and structural dependence on central places, while k-core deletion removes the less connected nodes, allowing the most important hubs to come to the foreground. These techniques are also combined with comparisons with random network models (degree preserving, Erdös-Rényi, Watts-Strogatz) to assess the robustness of the archaeological data.

Results show robust and enduring connections among the Levantine Late Neolithic sites, with at least three distinct ceramic community groups: coastal/northern, Hula Basin/Galilei, and Jezreel/Jordan Valley. The ceramic form and decoration networks show slightly different community clusters, with less clusters in the form network, indicating that the way food was prepared, consumed, and stored was very similar across the Levant, with decorative techniques slightly more localized. The networks were largely resistant to node and edge deletion processes, maintaining core structure and not immediately collapsing into multiple unconnected components, suggesting that sites maintained multiple relationships with surrounding settlements, and were not solely dependent on central places. This process also highlighted the importance of the Hula Basin as a bridge connecting the northern and southern Levant, and the Jezreel/Jordan valley as the core hub of the Levantine network.

Central to this paper is the idea that the landscape dimensions of a network are critical to the network’s formation and function. Starting with the CNRS ERC Desert Networks Project (ERC-2017-STG, Proposal number 759078) dataset, which records spatial and historical information about hundreds of archaeological sites and routes in the Eastern desert of Egypt, we explore the possibilities of incorporating high-resolution environmental data into our analyses of the ancient network. We also consider the applicability of Granovetter’s (1973) theory of “weak ties” as an analytical framework for approaching multi-period inscriptional galleries along the desert roads.

The Desert Networks project is engaged in constructing, among other things, a network model for the flow of people, goods, and information across and within the Eastern desert of Egypt from the New Kingdom to the end of the Roman period. Working from the location of archaeological sites and a calculation of the paths across the landscape that are most traversable by pack animals (in this case, camels) the project has previously identified a dense network of potential paths that connect places to one another. While large sites such as desert fortresses surely served as network hubs due to the opportunities they made for rest, news, water, and trade, the function of other sites in the network is more obscure. It is clear, though, that many sites with no apparent resources for travelers were nevertheless important nodes in the road network, as evidenced by dense palimpsests of landscape inscriptions. These places are often distant from major sites, but bear evidence of people stopping and contributing graffiti and other marks for millennia. Intensive Graeco-Roman interaction with these inscribed places raises questions about how locations of “off-site” activity should be identified and incorporated into models of archaeological networks. 

 Using a high-resolution environmental model in conjunction with the camel network mentioned above, we determine the profiles of three inscribed stopping places based on topography, temperature, sun shelter, wind speed, and proximity to water sources and major sites. We compare these profiles to the surrounding landscape and discuss the environmental factors that we find play a significant role in the location of these nodes. We then discuss the content and function of some of the inscriptions at these sites, and consider their enduring importance in the larger network and the role that they play as locations where important encounters among “weak ties” may occur. This situates the inscribed stopping places as vibrant spaces of exchange in the desert, where practices of graffiti and other landscape marking communicate information across social groups, language barriers, and time.

In the present contribution, the corpus of published compositional analyses carried out on archaeological copper-base objects from Anatolian and Northern Mesopotamian contexts dated to the fourth and third millennium BC will be re-assessed through the network approach of the modularity maximisation method in order to identifying supply networks between copper-using communities.

In archaeometallurgical studies, scholars have traditionally tried to reconstruct metal supply networks by applying cluster analysis to chemical compositions of ancient metals in order to assign them to certain sources and production groups based on their trace element patterns. The various compositional clusters thereby obtained could be visualised as spatial nodes in distribution maps, which would eventually allow for the identification of exchange networks (e.g. Junghans et al. 1954; Junghans et al. 1960, 1968, 1974; Coghlan and Case 1957; Britton 1961; Chernykh 1966; Ottaway 1994). However, although a network is by definition a set of nodes connected by links, the distribution maps that have been generally employed in archaeological studies show lots of nodes but no links between them.

In an attempt to overcome these limitations, in recent years, an increasing number of archaeological studies have borrowed and adapted theory and methods from network science to the study of past societies (Knappett 2011, 2013; Collar et al. 2015; Brughmans et al. 2016). Indeed, complex network analysis produces graphs representing the intricate relationships connecting elements in either natural or artificial systems. In this respect, network analysis can be applied to chemical datasets of ancient metal objects in order to investigate the interaction and cooperation patterns that are hidden behind the spatial distribution of the various alloy types.

Among the wide range of network methods, the modularity maximisation analysis – first applied in archaeology by Radivojević and Grujić (2018) to the Balkan archaeometallurgical data with encouraging results – has been adapted and applied to the legacy dataset of metal chemical analyses from LC and EBA Anatolia and Northern Mesopotamia with the aim of identifying community structure in networks of metal production and exchange. In these two areas, technological developments in metal production occurred more or less at the same time, despite the differences in the relevant natural resources availability.

 The remarkable abundance in Anatolia of poly-metallic ore deposits combined with the lack of mineral resources of Mesopotamia must have been important in the creation and maintenance of an extensive network of interaction and exchange, especially thanks to the easy communication channels on the Tigris and Euphrates rivers. Applying the network analysis method to the Anatolian and Northern Mesopotamia archaeometallurgical data could enable to reconstruct how and to what extent Anatolia and Northern Mesopotamia were interconnected and thus enhance our understanding of the socioeconomic structures and relations underlying the production, exchange and use of copper-based objects during the periods under examination.

Between the Late Aceramic Neolithic and the Ceramic Neolithic several large and very likely, densely populated settlement punctuated the landscape of Levant and Anatolia. These sites are frequently referred to as megasites and they share a series of common traits: their large size in terms of aerial dimensions, their clustered, agglomerated, modular arrangement and the apparent absence of top-down spatial planning and of any sign of hierarchical arrangement of power and authority.

 Using Çatalhöyük as a case study, this research investigates the way these large Neolithic aggregate communities formed and the way they were internally structured, it explores the social mechanisms that developed within these dense communities and the manner these mechanisms enabled them to flourish and endure for generations seemingly without any evident administrative communal institution, it additionally addresses the social geography of these large agglomerations and diachronic variations in their social and economic structures. While no hierarchical stratification has been observed at Çatalhöyük a vast array of horizontal differences between buildings, which form the main element of the social fabric of the site, have been largely noticed. These differences have been variously explained through time.

From a methodological point of view, the social fabric of Çatalhöyük and its changing social geography is studied via formal socio-material network methods incorporating a large and diverse archaeological dataset produced by the Çatalhöyük Research Project over 25 years, combining a vast array of material categories. For this study, network models are constructed using the individual building as the smallest unit of analysis; links between buildings are traced using similarities of material culture which are used as proxies of processes of affiliation, belonging or social co-operation.  Therefore, this study takes place on a local scale (meso-scale) by analyzing socio-material relationships at an intrasite level. In this study, scale, both in time and space, are cross-cut through the medium of “bridging” objects capable of linking different times and places. Scale is, therefore, seen as fluid and flexible and constantly redefined through interaction. Objects, the practices of using them discarding them and associating them with other objects, are what kept the community together in threads of continuity within the site and beyond it. This study acknowledges the emergent nature of communities and the pivotal role of objects in framing social relations. Furthermore, it relies on a practice-based perspective which provides the research with a tool for interpreting the character of human experience, action and “micro-processes of daily life” and the physicality of things within them.

  An array of 2-mode material networks have been created and have been projected into weighted 1-mode networks of buildings and weighted 1-mode networks of objects. When possible, analyses were conducted on weighted networks and these were binarized only to perform specific calculations (e.g. density). As a way of studying the obtained networks I integrated a selection of analytical tools chosen from the entire network science methodological toolbox. I used exploratory, modeling, and validating approaches together. Additionally, I investigated the networks at both a node level using different centrality indices and at a global network level focusing on cohesion and hierarchy. Methods of community recognition (Louvain algorithm) have been used to highlight areas of more intense connections between specific buildings. Network permutations and the assortativity index are used to validate the obtained community partitions. Furthermore, QAP correlation and regression is used to test the hypothesis of geographical proximity in producing patterns of material culture similarities.

Weighted networks of objects have been used to evaluate the intrasite relationships among artefacts and to examine patterns of co-occurrence through time. Associations and disassociations of material culture items in networks are evaluated using a probabilistic model that highlights the items that are co-present with other items more than expected by chance alone.

This study allowed me to highlight changes in the social fabric of the site both through time and through space. While I interpreted the observed chronological changes as a process of community formation through a “rhizomatic” process of aggregation from a multiplicity of different strands, the differences noticed through space illustrate a multilevel social arrangement characterized by nested sets of affiliations starting from groups of close-by buildings that were probably sharing portions of the landscape.

Network methods proved to be a productive approach towards the goal of disentangling some of the aspects of the complexities of Çatalhöyük datasets and past societal arrangements. Working with networks granted me the opportunity of appreciating the role of connectivity and connections within the social world of the site. Chiefly, socio-material networks have been useful in thinking at material culture as constitutive of the social fabric of the site and in integrating a large variety of material culture items into one environment. While community partition could probably be replaced by other statistical methods (e.g. hierarchical clustering), measures of cohesion and hierarchy are achievable though network statistics. The opportunity of observing the data zooming into the node level and out to the global structure of the entire network model, therefore, integrating different scales of analysis is another very valuable essential feature of network methods. At the same time, I would like to stress that the patterns traced by socio-material network methods should not be considered as direct reflections of social relations in the past; instead, these patterns should be always interpreted contextually through the nuanced integration of different datasets.

Across a fragmented landscape, the Neolithic communities of the Zagros Mountains (in modern Iraq and Iran) maintained complex networks of material exchange and knowledge transfer. From the early Holocene, small groups of people explored new ways of doing and being in the world, sharing innovative ideas with one another through tangible material media. Drawing on my research at sites in the Central Zagros, this paper considers case studies for differing approaches to the curation of networks amongst the inhabitants of highland and lowland landscapes from the Epipalaeolithic to the Chalcolithic. Through key material strands and shared networks of practice, we can identify catalysing factors behind the growth of communication networks in the Early Neolithic and consider the implications of intensified connections.

This research is conducted under the remit of the ERC-funded MENTICA project (the Middle East Neolithic Transition: Integrated Community Approaches), led by Professor Roger Matthews, which aims to examine the ways in which communities (on the local and trans-regional scale) engaged with ‘Neolithic’ ways of living, as well as working with modern communities to support heritage and cultural rights. The research addresses transects through time and landscapes through inter-disciplinary research at multi-period Hotu Cave on the Caspian shores, PPNA Sheikh-e Abad and Jani in the high Zagros of Iran (Matthews et al. 2013), and in the Zagros foothills at Epipalaeolithic Zarzi Cave and a new open air site, PPNB Bestansur and Shimshara in Iraqi Kurdistan (Matthews et al. 2020; Matthews et al. 2019). This paper will look at material case studies from these sites and consider how engagement with networks was selective and contingent for individual communities.

The application of materials analysis to clay, stone, and shell objects can shed light on resource use in local and regional landscapes and identify intersections between communities. The well-established practice of geochemical analysis of obsidians sheds light on one particular set of networks in operation, marine shells on another entirely, but through integrated approaches to the materials, complex relationships between shared materials, technologies and ideas can be explored in greater detail. Through these interactions, we can begin to conceive the breadth and depth of the complex networks that connected Neolithic people across Southwest Asia. The prism of social network analysis facilitates a vision of communities selectively engaging with multi-layered and richly textured networks, but also demonstrates the challenges of working with limited datasets in prehistoric communities. Examination of material networks in comparison with emerging evidence for genetic networks highlights the extent to which not just things, but also people were mobile over vast landscapes. This paper considers ways forward for examining the MENTICA sites and introduces new projects. 

 

Matthews R., Matthews W., Mohammadifar Y. (eds). 2013. The Earliest Neolithic of Iran: 2008 Excavations at Sheikh-e Abad and Jani. Central Zagros Archaeological Project Vol. 1. Oxbow Books and the British Institute of Persian Studies, Oxford.

Matthews R., Matthews W., Rasheed Raheem K., Richardson A. (eds). 2020. The Early Neolithic of the Eastern Fertile Crescent: Excavations at Bestansur and Shimshara, Iraqi Kurdistan. Central Zagros Archaeological Project Vol. 2. Oxbow Books, Oxford.

Matthews R. et al. 2019. The Early Neolithic of Iraqi Kurdistan: Current Research at Bestansur, Shahrizor Plain. Paleorient 45: 13-32.

Investigating past human networks invariably requires the predominant use of proxy data, in the form of material culture, texts, or other indications of social interaction, to determine connectivity. Very rarely is it possible to observe the networks themselves, indeed the very idea of doing so is counter to their perception as an abstract concept. The so-called “Black Desert” of eastern Jordan, however, presents an unusual case of being able to observe interaction networks directly in situ. This undulating landscape was formed by volcanic activity featuring large-scale lava flows that occurred between 30 million and 400,000 years ago, resulting in a surface of loess sediment covered in a dense layer of basalt blocks, locally called the harra. This means that with the exception of some mud flats and seasonal river valleys (wadis), traversing this landscape is extremely difficult, slow, and potentially hazardous. Therefore past and present populations in the region have cleared pathways to facilitate access, which due to the region’s extremely low taphonomic processes remain visible over millennia. Thus even a cursory observation of the region using remote sensing data, particularly satellite imagery, reveals a dense palimpsest of various cleared pathways constructed for differing purposes over a long period of time.

While this is doubtless a significant asset to the study of this region’s networks, it also presents its own range of challenges. These include ambiguity over the origins of visible tracks, since some are created by the feet of sheep kicking aside stones as they are herded across the terrain. Others are modern vehicle tracks cleared in the 1980s for oil prospection. Therefore a typology first needs to be created to distinguish pre-modern anthropogenic pathways from others. Furthermore there are difficulties in establishing associations between pathways and occupation periods or site types, since it is hard to determine whether certain pathways were used during the same time periods as the sites which they appear to connect, especially as subsequent re-use, over millennia and up to the present day, is likely for the majority. Finally, the visualisation of network vertices and edges through geographically accurate maps can be misleading in this region, since these do not clearly illustrate the significant variations in travel times and effort between moving along cleared paths, following mud flats or river valleys, and traversing the basalt surface.

To investigate this, the author conducted a detailed study of cleared pathways in a section of the “Western Harra Survey” area together with the project’s co-director Marie-Laure Chambrade of CNRS Archéorient, Lyon. Using satellite imagery, DEMs derived from drone imaging, and surface analysis, it revealed that “sheep tracks”, modern vehicle tracks, and anthropogenic pre-modern tracks can be distinguished based on their width, linearity, and topographical footprint. With the knowledge of which paths constituted the latter, a field experiment was conducted where a walk of 6 km was timed first along a pre-modern path and secondly over the basalt rocks right next to that path. This revealed that walking along the cleared track was 20% faster. Using this data, the pathway routes were first vectorised in ArcGIS, and the created shapefiles then modified to produce cartograms which clearly visualise travel times across this section of the desert and illustrate the true “distance” of sites from each other for practical purposes of pre-modern populations. This new method of presenting such data for the region will facilitate dissemination and allow for easy intra-regional comparisons.

The Archive of Zenon is the largest private archive to have survived from Ancient Egypt. It comprises c. 1845 ancient texts, covering a period of thirty-five years (263-229 BCE), and provides detailed information about the lives and affairs of individuals living under various conditions in Egypt and beyond under king Ptolemaios II Philadelphos. The combination of its documentary nature, grand size and short time span is one reason for the archive’s importance; another is the powerful positions held by the archive owner, Zenon son of Agreophon, and his superior, Apollonios the dioketes. As the financial minister of Ptolemaios II, Apollonios filled one of the most important functions in the bureaucracy of early Ptolemaic Egypt. Moreover, he had private businesses all over Egypt and her provinces.

To administer his official and private affairs, Apollonios – like the king – relied on a complicated network of allies. One of the most important persons in Apollonios’ inner circle was our Zenon. As such, the archive in question offers unique insights into the functioning and particularities of this complex and dynamic machinery.

The texts reveal that Zenon’s responsibilities changed numerous times, but also that he remained a thrusted agent of Apollonios’ until Ptolemaios III Euergetes came to power and Apollonios disappeared from history (c. 246 BCE). Even after Zenon stopped working for Apollonios, he kept adding documents to his collection of texts, now as a private person living near Philadelphia but with businesses spread all over the Arsinoite nome.

Although Zenon seems to have grown old in Egypt, he only settled there late in his life. He was born in Kaunos in Caria and when he first appears in history, he was traveling as Apollonios’ private agent in the Levant. Among the earliest documents are thus contracts and reports about businesses, sales and gifts sent between places, letters about outstanding debts and requests for travel arrangements, accounts and lists of payments with information about planned trips and the people involved in them, etc. Since several attested individuals and locations reappear in more than one document, the sources and information they reveal can be meaningfully conceptualised and modeled as k-partite networks.

The current paper discusses how formal methods of Social Network Analysis (SNA) can be fruitfully applied to the study of ancient texts. By means of a case study of the c. 40 earliest manuscripts associated with this archive, the author demonstrates her approach to studying socio-economic relationships and connectivity in the Zenon papyri. In doing so, she maps, visualises and analyses concrete examples of animals, things and people moving between and within regions in the Levant c. 261-258 BCE.

The sixth-century text, On Buildings (de Aedificiis, Peri Ktismatôn), by Procopius of Caesarea is an invaluable guide to the building works carried out during the reign of Emperor Justinian (r. 527-565) across the Roman Empire, and in particular in the capital city, Constantinople. The city forms the subject of the first book of the work, and includes magnificent descriptions of monuments that are still extant today, like the Hagia Sophia, as well as lost works, like the Church of the Holy Apostles. Procopius’ text, however is a work of literary production, hovering somewhere between panegyric and ekphrasis in terms of its genre, and is therefore often (though inconsistently) dismissed as unreliable by archaeologists and art historians seeking to use it to reconstruct lost monuments and historical realities.

Despite its literary slant, however, the information contained in the text can be broken down and used for data-driven analysis. Buildings can be grouped according to type, whether secular or religious, and their geography, whether intra- or extramural. In the case of religious buildings they can be further grouped either according to their dedications to particular saints, or according to the typological form of their plan (centralized vs basilical). Links can also be made to specific rhetorical devices, such as locus amoenus descriptions of milieu, to see in what contexts Procopius takes greatest artistic liberties.

The data from Procopius can then be supplemented and contrasted with data obtained from other historical and archaeological sources, regarding chronology and social networks of construction by including other actors, such founders and commissioners of works whose role Procopius glosses over in his drive to glorify Justinian, be they previous emperors or contemporary civil officials.

This paper will show the results of Social Network Analysis carried out on the patrons of specific building projects in 4th-6th-century Constantinople as known from other sources and contrasted with the information derived from Procopius’ text. The two models generated make it possible to systematically compare Procopius’ representation with facts from other sources. The resulting visualizations will provide a consistent overview of how Procopius' literary strategy diverges from archaeological reality, and provide a better idea of how to use the text for understanding lost monuments.