NOTE: This is the working draft of an essay that uses social-network analysis to investigate the rise of literary criticism in English during the Restoration (1660-1700). The data still needs to be cleaned up and is not yet available. Comments welcome!
Allow me to begin with a simple fact: the term “criticism” was nearly absent from the English language until it experienced a sharp spike in the latter half of the seventeenth century. (See Figure 1.) To be sure, critical writing had long existed. Renaissance poets debated the merits and ethics of their craft, and they sometimes referred to their adversaries as “critics.” But throughout the century, few people saw criticism as a coherent practice unto itself, and they almost never described it as a kind of writing—until the 1690s, that is. What changed? Most written criticism appeared in paratexts like prologues, epilogues, prefaces, or dedications, where poets offered their work to readers and audiences, or where they defended themselves against the accusations of rivals, or where they heaped praise on potentially generous patrons. Rather than an authoritative discourse dictating to a mass reading public—an image often inferred from the examples of later canonical figures like Joseph Addison or Samuel Johnson—most early literary criticism was devoted to the reception and distribution of individual plays and books. Rather than an author/public dyad, the social structure delineated by paratextual criticism bound poets among a web of others: not only other writers, but also politicians, stationers, and players, as well as writers from earlier eras and foreign places. Criticism was the metadiscourse of the literary marketplace. For this reason, changes in criticism can be tracked by examining changes in the relationships that criticism was employed to enact.
In this essay, I experiment with a method that transforms publication metadata and the results of text analyses into network graphs like this one. (See Figure 2.) This sociogram displays the structural relations that prevailed among authors, stationers, patrons, and performers who were directly related to the dramatist, William Congreve. Working from a corpus of texts drawn from documents recently released by the Early English Books Online Text Creation Partnership (EEBO-TCP), this chart maps relationships encoded in books. At the center of the graph sit booksellers like Jacob Tonson, Richard Wellington, and Samuel Briscoe, who published works by Congreve and many other poets of the time. Alongside them sit actors like Thomas Betterton and Anne Bracegirdle, who spoke the prologues and epilogues to several of Congreve’s plays, as well as those of fellow playwrights like Thomas Southerne and John Dryden. On the outer edges of the graph are people involved in fewer co-publications. The groups of nodes near the top and the bottom of the graph (green and light blue, respectively) reflect Congreve’s inclusion in miscellaneous publications, Letters Upon Several Occasions (1696), a collection of correspondence edited by John Dennis, and Poems on Affairs of State (1697), a book that gathered new writers like Charles Gildon alongside poets from previous generations like Aphra Behn, John Milton, and the duke of Buckingham.
Figure 2 does not display Congreve’s social networks in the colloquial sense. Only some of these people were Congreve’s actual associates. Instead, it maps the persons involved in paratexts that involve Congreve. As has already been seen, such involvement could take a number of forms. These nodes include authors who contributed to books alongside Congreve, printers and booksellers listed in his imprints, performers of his prologues and epilogues, and patrons to whom his dedications were addressed. The links (called edges) that connect these nodes represent the documents in which these various people were involved.
My argument will proceed as follows. I’ll begin by providing a more detailed description of the model. I’ll discuss the process by which its data was gathered and analyzed, and I’ll say a few words about the goals and methods of social network analysis, particularly as they apply to what I’ll call historical text networks. Then my argument will delve into the model itself, using it as a tool to investigate three motivating questions:
- How did paratextual criticism change over time, from 1660 to 1700?
- What roles did stationers and patrons play in these changes?
- How did women’s contributions differ from men’s?
Throughout my survey of the network, I’ll also describe the results of a lexical analysis of the documents. Because each link in the network represents a document in the collection, it becomes possible to correlate changes in the lexical makeup of critical discourse with changes in the social structures that supported that discourse. The lexical analysis was performed using Latent Dirichlet Allocation (LDA) topic modeling. LDA modeling is an imperfect and controversial tool, but it provides a useful survey of the collection as a whole, and it can help delineate in outline some of criticism’s various threads. Despite the inevitable imprecisions caused by its randomized sampling of text data, LDA modeling provides a needed corrective to some misunderstandings of early criticism perpetuated by studies based on close readings of individual authors or others organized around clumsy categories like “literature” or “politics.”
Rightly or wrongly, early criticism tends to figure in history as the precursor to the literary studies of the twentieth and now twenty-first centuries. In my conclusion, I’ll posit that the Restoration period witnessed a transformation in the material condition of literary knowledge. Criticism became “modern” not when “literature as such” consolidated into a coherent and discrete category, nor when the discourse was institutionalized in review magazines or university departments, but rather when criticism was first recognized as a general social pattern—call it a “practice,” if you like—that transcended the disagreements of any individual poets and could be understood instead as the textualized condition of literary knowledge in general. I conclude by briefly discussing some possible implications this shift in perspective might have for debates about the so-called “crisis in the humanities” today, as well as for debates surrounding digital humanities and the increasing popularity of quantitative literary research.
The full text of the draft can be read in pdf form here: Historical Text Networks May 3 Draft.