Pleiades: Beyond the Barrington Atlas (London 2006)

This is a partial text of a paper delivered by Tom Elliott to the London Ancient History Seminar (cf: ICU Events Calendar), 27 October 2006. The text breaks off at the point the demonstration portion began; that part was done extemporaneously.

Note that some of the text of this paper overlaps that in a shorter paper, delivered at the Annual Meeting of the American Philological Association in January 2007. Text that is substantially the same in both papers has been marked here in blue.

Visually impaired users should set their screen reading or brailling software to render the content of html "title" attributes; the "title" attribute on relevant paragraphs will contain the string: "This paragraph is repeated in the APA paper."


Alot has changed since the Barrington Atlas was published in 2000. Even as the 12-year-long Classical Atlas project was coming to a close, it was clear that a staggering series of changes in both mapping and geographic study was well underway. Where practical – and conducive to its ambitious schedule – the project had exploited new developments in Geographic Information Systems – or GIS – as well as other technologies.

At the same time, no one affiliated with the project believed that its success would be the final word in Greco-Roman geographic reference. Even as Richard Talbert and I Fedex’d the last batch of Map-by-Map Directory materials to the publisher in the early light of a Chapel Hill dawn, we knew that the rising technological wave would combine with the quickening pace of historical and archaeological research to produce new opportunities for exploiting and expanding the data embodied in the Atlas.

Of course you’re looking at one of these revolutionary developments – Google Earth. In 2000, GIS was already an established reality, and we could easily name specific Atlas-related tasks for which we wanted to use it. But GIS was still largely the province of highly trained specialists, equipped with the most up-to-date computer hardware and large budgets for software and data. We were still thinking in terms of using GIS to create new digital and print materials that would be provided, in more or less static form, to other projects and individuals. Few people would have been willing to predict the effectiveness with which Google, NASA and other players have since placed high-quality geographic visualization tools and imagery in the hands of computer owners in the developed world. Now anyone with a reasonably up-to-date computer, and a broadband internet connection, can fly with me from the shrinking rural countryside near Huntsville, Alabama to the erudite confines of Senate House, here in London.

But we are historians, classicists and archaeologists. We study the ancient past. Shall we break out of this presentist orbit and go somewhere interesting? What if we could harness these new technologies – though developed with modern geography and modern economics as their focus – to serve our research and pedagogical purposes?

It’s a relatively straightforward matter for any user to examine places of interest in Google Earth. If you know the modern name of a geographic feature – and if that feature is significant enough in modern terms to appear in the digital data Google purchases for their services – a simple search is sufficient to take you there. But if you only have a historical name, or if your feature is obscure, or if it is remote from the modern place with which it is commonly associated, this most basic of approaches will fail.

You could take matters into your own hands. Assuming a reasonable degree of both map- and computer-literacy, you could derive coordinates from the Barrington Atlas by hand and type them into Google Earth. In this way, you’d create so-called ‘placemarks’ – points on the Google Earth globe to guide your explorations. In this manner, you might ultimately find your way to this site: ancient Choma – known today as Hacimusular, in Turkey.

I think you’ll agree, however, that something more comprehensive and automatic would be preferable. You have students to teach and research questions to answer. Juggling a gigantic book, a recalcitrant mouse and a bunch of numbers is neither your heart’s desire nor your forte. Nor should it be. Google Earth is designed to retrieve data, as you need it, from other systems and services via the Internet.

The scholarly compilers of the Classical Atlas project documented over 50,000 physical and cultural features that are attested by the historical and archaeological record. All we need to do to make this data available in Google Earth is the will – and the resources to do it accurately and completely.

I am happy to tell you that the U.S. National Endowment for the Humanities has generously provided initial funding for a project that will start us down this road.

Specifically, the Ancient World Mapping Center’s Pleiades project is establishing a web-based environment for the perpetual update, perpetual diversification and perpetual dissemination of the data assembled by the Classical Atlas Project. This application will facilitate rapid editing of new geographic, bibliographic and analytical information relevant to the study of Greek and Roman geography. Moreover, the mapping center – together with its collaborators – will disseminate this data via in various print and digital formats.


Note: the APA paper includes, at this point, a discussion of the current and future planned phases for the project. This discussion does not appear in the present paper.


Reliable, comprehensive and up-to-date reference works are essential tools in all fields of humanities research and teaching. It was the realization of this basic truth that impelled the American Philological Association to launch the Classical Atlas Project during the 1980s. Although a number of important cartographic projects were already underway, this essential nexus of reliability, comprehensiveness and currency had been neglected – for the core areas and periods of Greek and Roman history – for over a century. And this tripartite truth may be underlined again now, some 6 years after the completion of that project. We need only to consider how completely we’ve come to depend on the book that the project produced.

Unfortunately, key reference works in the humanities – particularly geographic ones like the atlas – are increasingly prone to obsolescence. In a few minutes we will consider examples of the effects of age on the Barrington Atlas. For now, allow me to observe that – although there is no coherent body of scholarly literature on this subject – this problem is widely appreciated. For example, a search of the online version of the American Reference Books Annual, which provides brief but authoritative reviews of thousands of reference sources, returns over 100 reviews of works published between 1995 and 2005 in which obsolescence of content is already noted.

Beyond the implications for individual research, this fact of life should raise serious societal concern, both within and beyond the academy.  My daughter will live to see and use a global system of digital libraries containing – for better or worse – virtually all printed, graphic and audio works now preserved in one or more first-world, brick-and-mortar bookstores, libraries or museums. Moreover, she will join her contemporaries in increasing reliance on programmatically constructed summaries, extracts, collections and analyses of this vast data hoard. Accordingly, her economic, social and intellectual competencies and behaviors will be shaped by automated processes in ways impossible for me to imagine.

I am only confident in predicting that she will need, and deserve, automated help in dealing with it all: not only with the millions of old books that are about to be brought – by brute force – back into the public consciousness, but also with the rising tide of digital publications and data streams that will accompany growing populations, cheaper access to global audiences, and the popularization of traditional academic topics.

We cannot stop this train. It is powered by the twin engines of the market and the human thirst for knowledge. We would be wrong to try. Its troublesome aspects are more than outweighed by its potential for fostering emancipation, democracy, education and human achievement.

But scholars do have a part to play in conditioning the outcome. On what basis will the programmers’ of our children’s future construct the heuristics and algorithms intended to shape their information experiences? Reference works are already key components in both corporate and academic efforts to bring order to information chaos. Greg Crane and his collaborators at Perseus are not alone in demonstrating that software tools, designed to mine information from large, heterogeneous document collections, can be improved significantly through training with structured reference data. Biographical, geographical and historical reference works provide the information necessary to contextualize otherwise ambiguous references found in both primary and secondary texts. If scholars can grapple effectively with the tendency of their reference works to age, and if we can simultaneously restructure and redeploy these works to facilitate use – not only by individuals, but also by automata – we can improve the results.

We have deliberately positioned Pleiades in a liminal space. This initiative brings together both automated processes and committed human users to preserve, extend and reuse an existing, high-quality dataset, readying for the coming age.

I’d like to use the balance of my time today by returning to the details of that dataset and then demonstrating some of the tools we are building in order to accomplish these lofty goals.

The Barrington Atlas provides its viewers with 102 maps depicting the ancient landscape with modern cartographic conventions. Both physical and cultural features are highlighted, to the extent that modern scholarship can report their details. The maps therefore indicate elevation and salient terrain features, such as lakes, rivers and swamps. Where our sources provide us with names for these features, they are labeled. For example, in this excerpt from the Barrington map of Lycia (Map 65), we find a river labeled “Lysis.”

The general, spatial footprint of ancient regions and tribal territories are also indicated with distinctive labeling, as is the case here with the so-called “Praedium Plancianum” and the old stomping grounds of the Ormeleis.

Roads are traced, and our confidence in following their routes is indicated through the choice of solid and dashed linework. For example, here we find the Via Sebaste.

Similarly, hollow round symbols indicate the approximate locations of settlements, whereas other settlements may be placed more confidently and so earn solid round symbols. Compare, for example: Olbasa’s solid circle with Polyetta’s hollow symbol.

Compilers also signal occasional hesitancy in the assignment of a name, attested in literature, to a site securely located by archaeology. This is done appending a question mark to the label for a site, as here, with Kormasa?

A controlled set of distinctive icons are used to indicate sites of various types; for example, five-pointed stars mark the locations of extra-urban temples, sanctuaries, monuments, shrines and tombs, as is the case here with a shrine to Roma and Augustus.

The maps are supplemented by the Map-by-Map Directory. This 1,500-page work provides a compiler’s introduction for each map, together with a list of additional data for each labeled feature and a full bibliography. It is in the ancillary data and bibliography that – in our quest to keep the atlas up-to-date – the largest number of changes will be required.

In lockstep with the maps, directory listings provide historical names for features. Project compilers and editors exercised great care to admit only those names attested in ancient sources or – absent any such – those later names that might be reasonably inferred backward in time. Thus, latinate forms assumed for Greek sites by modern scholars, and used in print, have been rigorously excluded. Where a cultural feature is attested by archaeology but no ancient name can be associated with it, modern names have been introduced, using distinctive type. A series of editorial conventions for historical names, employed on the maps and in the directory, facilitates the marking of variants, reconstruction, interpolation and other editorial observations. The temporal aspect is treated as well. The 1,500-year span of the atlas is subdivided into 5 broad periods. The currency of names, and the relevance or habitation of built features, are communicated with letter codes for each period, postfixed with a question mark as necessary to indicate hesitancy on the part of the compiler. For all features, an attempt is made (via the “Modern Name / Location” column) to communicate the modern position of a feature via an associated modern placename or short description. Finally, a select number of bibliographic citations are provided in order to direct users of the atlas, as efficiently as possible, to a comprehensive or representative work that, in turn, facilitates discovery of all relevant scholarly publications. In the absence of any such modern work, primary source citations were substituted. Rarely, a compiler’s name is introduced to indicate an original conclusion or discovery whose separate publication was anticipated, but not yet scheduled. Let us turn to a specific example in order to demonstrate how we are restructuring this data for incorporation in Pleiades. With this brief demonstration complete, we will be better able to consider how Pleiades will facilitate updates to the data.

I give you Aphrodisias in Turkey, Barrington Atlas map 65, grid A2.

Because of its prominent ruins and on-going excavation, the Barrington compilers are able to locate the ancient city securely at modern Geyre in Turkey. Relying on previously published considerations, and eager to communicate one important change in the city’s history, the compilers assign the early name “Ninoe” to the classical period, and associate the now-better-known “Aphrodisias” with the city’s heyday in the Roman imperial and (especially) Late Antique periods. Pauly’s Realencylopaedeia is cited for both relevant articles. So too the Neue Pauly, whose earliest volumes were appearing as the atlas project finished its production phase and moved to publication.

I offer a less-compressed visual repetition of the same information, which more closely mirrors the manner in which we are restructuring it for storage and manipulation in Pleiades. Of particular interest is the conversion of graphical information on the map into textual and numeric data that can be manipulated programmatically. Specifically, I mean the feature type of “settlement” (indicated in the Barrington solely by means of a circular point symbol on the map). So too, the latitude and longitude coordinates corresponding to the site’s geographic location – available from the Barrington only by inspection and calculation from the map itself.

Over the past few years, I have had the opportunity to consult with the Inscriptions of Aphrodisias project team at King’s College. Professor Roueche assures me that both the archaeological and epigraphic record make it abundantly clear that the city was a going concern during the Hellenistic period, and that the name “Aphrodisias” was clearly in use at that time. It would be appropriate, therefore, for us to modify the Barrington directory to reflect this understanding in any future publication or reuse of data. Furthermore, the recent digital publication of a revised second edition of Professor Roueche’s work Aphrodisias in Late Antiquity, surely deserves citation. This addition will provide access to more recent considerations even than those cited in Neue Pauly. Given its electronic nature, we may also hope that it will provide a ready bridge to the full publication of the Aphrodisian epigraphic record, now in preparation.

Let’s visit the Pleiades website, and make these modifications to the data itself.