Earlier this week, I posted an image of EEBO-TCP as a Giant Hairball. I call these historical text networks. The basic idea is to think of books as points of connection in a network: when printers and booksellers collaborate to distribute the work of an author, this action can be understood as a link that binds them together. Taken as a whole, these myriad ties form EEBO into a giant social field, or “hairball.”

hairball

Visualizations like these don’t do much except communicate the scope of the underlying data. However, if you look closely at the hairball, you can see that EEBO includes a number of fairly well defined clusters that appear throughout the network. Although the field as a whole is deeply interconnected, groups of stationers and authors who often published together appear closer together in the network. What are these groups?

When I use this technique on smaller subsets of the data – for example, taking just a decade or two at a time – the communities that emerge correspond to different segments of the print marketplace, separating, for example, poets and playwrights from religious radicals. I’ve used SNA to tease out some of these differences and to trace book circulation across disparate but synchronously proximate groups.

When the entire corpus is modeled, however, it streches diachronically over several hundred years. From this bird’s eye perspective, the clusters correspond to historical moments. The stationers who published together also lived together, and although some authors get reprinted posthumously (a point I’ll return to at the end), most don’t. It seemed to me, then, that finding communities over the whole network could be a way to think about how the archive organizes itself into historically defined clusters. I’m used to thinking about the “structural holes” that divide a social network into parts, but by extending the network far beyond the scale of a human lifetime, the gaps in the network become like holes in time.

After running a simple “walktrap” community detection algorithm, I briefly glanced through the results, which came out more or less as I had expected. The most-connected nodes in each community tended to be people who lived around the same time. So I re-did the “hairball” graph, highlighting in color the five most populous clusters. Here’s the graph again:

network

On the far left, the community in green is made up primarily of printers and writers from the sixteenth century. The yellow group includes Shakespeare, Jonson, and many of their Renaissance peers. The largest community is shown in purple: most prominent here are monarchs (both Charleses, as well as William and Mary) and their official printers. On the right the blue group includes John Dryden and Aphra Behn, as well as printers and publishers I know well from my earlier work on the Restoration literary marketplace.

Because the links in the network all correspond to EEBO-TCP documents, it’s possible to quantify how these different communities are distributed through time. Each link (or, “edge”) in the network has a corresponding publication date, so we can look at every edge for each node in these communities, then stack them onto each other in a timeline.

boxplot

The boxes in a boxplot show where most of the data-points are: outliers are arrayed as points to the left or right. The thickness of each boxplot is indexed to the size of the community: the largest group, with 1739 members, is the most central, featuring kings Charles I and Charles II. As you can see, the five largest communities are spaced in time, more or less according to conventional period divisions. Erasmus and John Fox cluster with printers from the sixteenth century. The late Elizabethan and early Stuart era appears quite distinctly, as does the late Restoration in blue at the top right. The large red and purple communities in the center correspond closely to the Civil War, Interregnum, and early Restoration. The break that separates the red from the purple in this chart sits uncannily close to 1649, the year of Charles I’s execution.

Yet, just as the network graph shows how deeply intertwined these social groups were in the marketplace, this timeline displays a lot of overlap, even disregarding the many smaller communities that populate the full graph. The largest communities form networks that stretch over centuries. Rather than clearly demarcated period boundaries, network analysis reveals the many layers of temporality that unfold through print’s social space.

Book History, Conceptual History, and the Problem of Time

Discontent with period boundaries is perennial. Ted Underwood’s recent book, Why Literary Periods Mattered, describes the consolidation of period divisions in nineteenth-century historiography, particularly, in literature, around the advent of the period-survey course. However, if periods have long been used to organize literary studies, their embrace has always been ambivalent.

In the field of book history, Leslie Howsam has warned that “studies of book culture must be attentive to their own complex problems of periodization” (Old Books and New Histories, 56). It is by no means obvious that time categories designed for literary or intellectual history are the best way to organize other lines of inquiry.1

Even within the (themselves varied) fields of intellectual history, though, periodization poses obstacles for analysis. For many, this obstacle takes the form of a seductive but misleading contrast between the modern and the ancient, or pre-modern. The category of the “early modern” reifies this distinction even while situating historical development on a continuum without radical breaks. Medievalists often find the distinction off-putting and reductive. Consider Kathleen Davis’s Periodization and Sovereignty or the responses to Stephen Greenblatt’s The Swerve.

Reinhart Koselleck’s history of concepts divided an ancient, feudal past from modern, Neuzeit, marked by capitalism and Enlightenment, but as Helge Jordheim has recently emphasized, Koselleck struggled against the very periodization his early work reinforced. Koselleck identified at least two kinds of historical analysis, synchronic (measuring the historical events and concepts that prevailed at any one moment) and diachronic (tracing conceptual change over broad swaths of time). Throughout his work, Koselleck stressed that these two modes of historical analysis complicated each other: “the diachronic organizes itself in terms of multiple overlapping layers, which run contrary to the conventional periods and opens up for different combinations.”2

Important research in book history by William St. Clair and Meredith McGill has pointed towards some of the ways these multiple temporalities were brought into being. Much of the discourse that circulates at any one time is permeated by texts from the past. Relying on a distinction between “synchronic” and “diachronic” analysis risks lapsing into what St. Clair calls the “parliament” and “parade” models of history: either the past is an empty space of contemporaries or history itself is presented as a mere sequence of thematically related events. By calling attention to the culture of reprinting, St. Clair and McGill show how the material conditions of textual distribution challenge the notion of a single coherent textual present.

Taken together, these graphs capture the reasons why periodization is tempting as a shorthand for historical difference, but they also stage the inadequacy of that shorthand. Each period appears clearly through the metadata’s noise as bursts of publishing activity. Yet, these bursts bear little resemblance to a neatly demarcated series of befores and afters. Literary periods are overlapping and thoroughly interpenetrating – not blocks of time neatly stacked next to each other from left to right like books on a shelf, but networks of human activity threaded into a complex fabric of temporality.

Graphs like these are also available for a wide range of studies. I haven’t yet correlated these communities to textual analysis, but they might well provide a map for tracing both the continuities and discontinuities of discourse. They’re also available for statistical analysis of structure. Who was the book trade’s most important cultural broker? (Spoiler: It’s Shakespeare. By a mile.) What subcommunities can be found within each group, and what differences, either in social structure or textual content, correspond to them?

Ultimately, what’s at stake with historical social-network analysis isn’t a consilience between the humanities and the social sciences, as many suppose. Instead, it bridges the even more stubborn disciplinary divide that separates conceptual from social history.


  1. See also Michael J. Suarez, “Historiographical Problems and Possibilities in Book History and National Histories of the Book.” He writes, “Practitioners of book history have often adhered to traditional divisons operative in either literary or political history because we have inherited these temporal categories as a consequence of our professional training. Yet, it is by no means clear how the reign of Queen Victoria—or, as some prefer, the interval from the passage of the Reform Act (1832) to the accession of Edward VIII (1901)—is especially meaningful for book history as an investigative tool for framing our inquiries, however well it may serve as an heuristic expedient.”
  2. Cited in Helge Jordheim, “Against Periodization: Koselleck’s Theory of Multiple Temporalities.”

2 thoughts on “Using Networks to Re-think Periodization

  1. K. Bender says:

    Your approach would be equally valid for art historical data, with painters, engravers and printers and publishers who disseminated the artworks on a large scale. Just one example. RAFFAELLO and his many collaborators, their famous engraver Marcantonio RAIMONDI and pupils, the house-printer and publisher Baviero de’ Carocci, and others at a later date.

    Reply

Leave a Reply

Your email address will not be published.