Connecting Communities in Archaic Greece: Exploring Economic and Political Networks through Data Modelling

Loy’s book explores what he calls “big data” to reflect on the interconnectivity, both economic and social, of the Archaic Aegean. The term big data is intentionally vague, as each chapter ultimately investigates four very specific and carefully crafted datasets of marble, pottery, coins, and inscriptions. These data are primarily from nineteenth- and twentieth-century excavations—as the author explains, “we may all lament the state of old and bad data, but that does not remove the need to engage with it” (9). A quick skim through the many tables and graphs in this book will impress upon the reader the ambitious scale of this project, and the author confesses that “what is being proposed here almost looks like a backwards step towards processualism, an uncritical ‘archaeology by numbers’ that prizes manipulation of the data without full contextual discussion” (13). In many ways, this book is an encomium of Peer Polity Interaction and world-systems theory, and all the critiques of those ideas can equally be applied here. I favor those theories and find Loy’s conclusions persuasive, but I have reservations about the book’s engagement with digital humanities (DH).Unfortunately, the publisher has not served Loy well. Occasional spelling, grammar, and formatting errors appear throughout, and many of the graphs, visualizations, and maps, both in the book and at www.cambridge.org/loy, are poorly labeled and indecipherable. Most graphs are missing labels and legends outright, and those that do have labels are often difficult to read due to overlapping text. The author’s claims often depend on social network analysis (SNA) visualizations with nodes representing cities in geographic space, but none of these visualizations include labels (i.e., the names of the cities) or the contours of coastlines. The editors decided not to include basemaps for many of Loy’s geographic information system (GIS) visualizations, so the data on these maps appear to float in abstract space. For a project so dependent on charts, graphs, maps, and visualizations, the inaccessibility of these media makes the argumentation difficult to follow.The book contains seven chapters: an introduction; four chapters exploring the marble, ceramic, coin, and inscription datasets respectively; a chapter comparing the visualizations from each dataset; and a conclusion. The four datasets are hosted at www.cambridge.org/loy. As with any project of this scale, each dataset is subjective, and art historians, ceramicists, numismatists, and epigraphists will no doubt have reservations about the data in each collection. The book presents a series of conclusions based on its graphs and maps, but these media are simply visualizing the big data datasets. Although the author considers his big data to be “sufficiently full and diverse” (195), there are clearly significant gaps in many cases. In its discussion of the inscriptions, for example, the book claims that “there is little correlation between the type of writing produced and geography: all types of writing [i.e., small personal inscriptions, dedications, short public markers, longer texts like building records, and public laws] are found in all parts of the Aegean” (240). This claim, with which no specialist would agree, appears just centimeters below an empty map for public laws from 550–500 BCE that has only six points: Iolkos, Chalkis, Athens, Corinth, Olympia, and Gortyn (I am guessing the cities’ names because the map has no labels; fig. 5.9e). The online version of the inscription dataset includes Late Archaic laws from Axos, Eleutherna, and Eltynia, but these cities are not marked on the map—although the book includes references to modern treatments of these texts in its bibliography, it apparently did not consider these inscriptions to be public laws. This is just one example of how this book compiles, deploys, and then tethers itself to big data, and why the decision to avoid specialist concerns ultimately detracts from the persuasiveness of the thesis.Chapter 2 introduces the marble dataset and estimates the tonnage of marble that was shipped around the Aegean. This research is impressive and fascinating, and it persuasively demonstrates that the marble trade was both lucrative and gradually expanding over the course of the Archaic period. The chapter then deploys Proximal Point Analysis (PPA), which I will discuss further below, to reconstruct how this trade operated over space and time.Chapter 3 begins with a series of GIS maps showing where archaeologists have discovered “Attic,” “Corinthian,” “Ionian,” and “Local” ceramics and how these assemblages changed over time. This dataset compiles 50 publications on 32 sites (91), and the four ware types are intentionally vague to accommodate the various ways in which this material was published (94). Using R, the chapter presents a series of Principal Component Analysis (PCA) visualizations to measure the similarity of each site’s ceramic assemblage and its preference for certain ware types and shapes. The PCA visualizations have overlapping text and missing labels, so while the author’s conclusions are interesting and potentially groundbreaking, I am unable to confirm what he is seeing in the visualizations. This conversation ends by turning once again to PPA.The author is careful to insist that PPA visualizations create “corridors of probability” rather than realistic reconstructions (73). To make a PPA network, every node in an SNA visualization is tied to the three closest nodes. By closest, I mean as the crow flies—this system does not account for mountains or seas. And it is never stagnant: every new node creates three new lines. It is vitally important, therefore, to state exactly what a node is: Are they cities with marble, cities with pottery, all Archaic ports, just modern ports? As far as I can tell, the nodes are never clearly defined in either Chapters 2 or 3. Chapter 2 references 1,550 harbors and anchorages, but a footnote clarifies that these are likely not all Archaic (68). Then it states that “anchorages are used for the base unit of analysis here rather than formal harbours, as our knowledge of ancient harbours is undoubtedly incomplete” (72). What qualifies as an anchorage and the source of this data are never explained. As with the construction of the four core Archaic datasets, the absence of clarity in the construction of the DH visualizations makes it difficult to evaluate the book’s results.Despite these caveats, the corridors of probability for the marble and ceramic trades are extremely helpful. These visualizations reiterate what scholars have been arguing about the Archaic period for decades: conspicuous consumption and economic commodification drove complicated large-scale networks of exchange that encompassed the whole Aegean. In his defense of PPA, the author observes that his visualizations cross the isthmus between the Saronic and Corinthian Gulfs, and he then launches into a critical analysis of our evidence for the Corinthian diolkos, a paved and well-worn path that connected the two gulfs (71–72). These observations fit perfectly into the book’s broader arguments about the influence of economic networks on human interconnectivity and mobility in the Aegean, but they are presented as a defense of the SNA visualization rather than as an observation developed out of the application of a DH tool. As soon as we uncover another set of anchorages, however we define them, new lines will have to be drawn and the corridors of probability will shift. Unfortunately, the book consistently prioritizes the product of a DH tool that will soon be eclipsed by new archaeological findings, or even a software update, over intriguing questions that arose out of the application of DH.Chapter 4 tracks how each Archaic coin weight standard changed over space and time. These visualizations are quite impressive and could be immensely worthwhile to many numismatists, but they would benefit from basemaps and labels to improve their legibility. The patterns they capture could raise interesting questions about economic practices in the Aegean and the role of circulation to influence exchange.Chapter 5 similarly presents many interesting and inaccessible SNA visualizations, but its conclusion highlights one of the primary issues with this book’s approach to DH. To summarize the chapter in unfairly blunt terms, it creates an epigraphic dataset using Kirchhoff’s and Jeffery’s works, digitized legacy data from Inscriptiones Graecae, and Local Scripts of Archaic Greece (202–4); visualizes that data; and then points to the visualizations to argue that Kirchhoff’s and Jeffery’s identification of three dialectic regions, which they labeled as red, blue, and green, was correct. In other words, it asks a machine to regurgitate its training data and then presents this recitation as evidence that the training data were valid. I doubt anyone comparing the author’s wonderfully colorful visualizations of linguistic diversity, like fig. 5.7c, would agree that they resemble Kirchhoff’s three dialectic types, which are mapped in fig. 5.3. The abstract visualization of fig. 5.5c, for example, actually appears to contradict Kirchhoff. There are three linguistically distinct clusters in this graph, but one includes Corinth, Athens, and Sparta; one includes Miletus, Larisa, and Olympia; and the third includes most of Crete, Ithaka, and Thermon. This is quite the sociocultural and spatial mix up—geography and cultural identity appear to have little to do with dialect. The author also makes this observation (210, 215, 220) but then shifts suddenly into an endorsement of Kirchhoff at the end of the chapter (227–29). Chapter 4 stated the author’s formula more explicitly: after presenting a series of SNA visualizations, a final section begins with a sentence reflecting that “it remains to decide what this means” (175). By ascribing meaning to a visualization, these chapters inevitably regurgitate the theses of the authors who compiled the big data in the first place.Chapters 6 and 7 are excellent because the author finally turns to probing his DH visualizations. Chapter 6 considers the Kalaureian amphiktyony and Ionian League and predictably finds that the visualizations do not align with the texts (260). The visualizations in Chapters 2 through 5 found no evidence of the Kalaureian amphiktyony and only nominal evidence of loose linguistic similarities within the Ionian League at a time when these communities are said in our textual evidence to have had deep political ties. Exploring this inconsistency further, the author concludes that big data, “which by their very nature have featured more sanctuaries than any other type of site” (284), are ill-equipped to represent political or economic networks across the whole Aegean; they do not, ultimately, present a complete picture. Perhaps it is no clearer than at this moment that this book has deployed DH tools in the wrong order: rather than exploring the peculiarities in the data, the conclusion dismisses them because they complicate the established narrative. It attempts to use DH to prove a theory rather than query our assumptions.One of the selling points of the book is no doubt its use of “experimental data analysis techniques from the digital humanities,” as the back cover claims. This is a puzzling way to describe SNA, statistical modeling, and GIS, which archaeologists have been using for more than a decade. The author clearly has a deep understanding of these tools and deploys them to great effect, but many who do not work in DH might miss what has been achieved through this process. The book opts not “to rehearse here yet another explanation” of the pitfalls and value of DH tools (11), and it is stated later in the introduction that nonspecialists are its intended audience (28). Unfortunately, this approach leaves many important methodological questions unaddressed. To be reductive, the book ultimately uses visualizations of old curated data to reiterate old theses. Outside the final chapters, the book does not use DH tools to ask new questions but treats them instead as a black box to provide original insight. This is a dangerous way to use a tool that is so dependent on inputted data. DH tools can only reflect their training data and should not be used, therefore, to generate new conclusions. Rather, they excel at creating new questions and exploring potential inconsistencies or complexities in our understanding.Loy has a solid grasp of the evidence and makes intelligent observations about this complex period of history. The book’s missteps with DH tools are ultimately most unfortunate for us, his readers, because they misdirect so much of Loy’s attention. Despite these issues, the volume successfully visualizes four old and complex datasets and proves that there is still much to be said about big data. When Loy starts asking questions about oddities he discovered in these DH visualizations, in Chapters 6 and 7, he makes interesting observations about the complexities of Rhodian Archaic identities (262–63), the importance of island sanctuaries like Delos as socioeconomic conduits (263–65), and the ways in which competing Ionian powers collaborated in certain economic and cultural arenas to dominate certain commodity trades (266–67). These sections demonstrate how DH tools can highlight surprising patterns in the evidence that might shake up, qualify, or even revitalize old ideas. Ultimately, this monograph is a welcome addition to any DH-literate scholar’s library because it successfully visualizes how the great scholars of the twentieth century came up with ideas like Peer Polity Interaction and world-systems theory. Chapter 7 will likely make it into my next classical archaeology syllabus—but only following a discussion about the pitfalls, limitations, and value of DH tools.