DH2017 Abstract
#New Methods for Studying the Critical Reception of Texts
Text reuse detection technology and approximate text matching have made possible the large-scale computational identification of intertextuality. These technologies have often been used in plagiarism detection and in studies of journalistic text reuse. Fewer studies, however, have applied these methods to literary research. We present a method for tracing the critical reception history of a source text by analyzing the density and chronology of its citations.
Our work builds on a recent set of digital humanities projects that use text reuse detection in order to study the afterlife of texts through textual quotation. The Viral Texts project and Digital Breadcrumbs of Brothers Grimm take algorithmic approaches to studying text reuse and circulation in their respective fields of research: nineteenth century popular press and folklore. In both projects, the focus is on using text reuse detection to uncover hidden networks of reuse––the reprinting of short news items and the re-use of motifs or minimal narrative units in folklore––in corpora without standardized conventions of citation. Our project seeks to take this work a step further, by applying a similar method in a different area of cultural production where more standardized citation conventions already exist: academic citations. We draw on the work of Dennis Tenen––in using extracted citations to critically visualize the "knowledge domain" of Comparative Literature––and a recent project by JSTOR Labs that uses the text of Shakespeare's plays and the U.S. Constitution to visualize the scholarship surrounding passages in those sets of texts.[^1] Like these projects, we hope to leverage the explicit and institutionalized nature of academic citation in studying text-reuse patterns, using text reuse to ask questions of critical attention in canon formation. By focusing on a single text in order to ask its critical reception history, we hope to provide new methods for studying not only text-reuse patterns, but the sociology of citation practices––studying changing in when, how, and what critics cite from a given literary text.
In applying these methods to literary scholarship, we've chosen to start on a relatively small scale. George Eliot's novel Middlemarch is an ideal test case, due to its length, copyright status, stable editorial history, and canonicity. Perhaps most appropriately for this study, Middlemarch is known for its narrator’s wise generalizations--highly quotable fragments that have been featured in more than one Victorian volume of George Eliot’s sayings. We use the Project Gutenberg critical edition of the novel, and programmatically compare it with a corpus of 483 critical articles retrieved from JSTOR.
Figure 1 shows a dispersion heat map of George Eliot’s Middlemarch. Here, the novel is broken into 50 segments along the horizontal axis, and each segment is colored according to the number of times its text has appeared in the critical literature--dark blue indicates that a segment has been quoted from not at all, and red indicates that a segment has been frequently quoted. A number of overall trends are noticeable here. The very beginning and the end of the novel show the most numbers of quotations, followed by the first quarter of the novel. Overall, the second half of the novel, except for the ending, is significantly less quoted from than the first half.
Viewed chronologically, we see that critical interest in certain sections--as expressed in numbers of quotations--appears unstable. In the 1950s, most of the critical citation was of the end of the novel, at its climactic third-to-last section, but this interest fades almost completely by the present day. At the same time, critical interest in the beginning of the novel was at an relative nadir in the 1950s, but quickly became a highly quoted segment by the 1970s. This shift may represent a change in critical attention to the parts of the novel, shifting from the end of the novel at midcentury to the beginning of the novel, where it rests currently.
Figure 2 shows an excerpt from our heatmap-annotated edition of Middlemarch. Here again, the color coding represents the number of quotations of each segment, ranging from unquoted black passages, to infrequently quoted blue passages and more frequently quoted green ones. In this case, the short, punchy “Her mind was theoretic,” and the even more powerful “yearned by its nature after some lofty conception of the world” have proved much more quotable than the more prosaic, although not uninteresting preceding passage, “She could not reconcile the anxieties of a spiritual life involving eternal consequences, with a keen interest in gimp and artificial protrusions of drapery.” This annotated edition might provide students of literature with a way to read the novel for passages that have been most discussed in secondary literature, and for passages that have been critically neglected.
The potential applications of this methodology are numerous and wide-ranging. Firstly, this methodology can be used in any discipline to investigate the discipline’s theoretical history. As with the 1980 study of Wundt's influence on the field of psychology[^2], our methodology could rapidly and easily produce similar investigations for the influence of Saussure in linguistics, Bourdieu in sociology, Mead in anthropology, Beauvoir in feminist theory, and so on. Moreover, this analysis would be much more fine-grained, registering not only the frequency of works cited but also specific sections, passages and even key phrases within them.
In addition, we see particular relevance of our methodology to disciplines substantially engaged in questions of quotation and commentary. In theology, commentaries on sacred texts could be analysed in this way to give insights into the texts' interpretive history. In philosophy, the citation of key texts in later periods could be used to assess the shifting priorities of subsequent generations of philosophers. Legal studies too, with the exceptional value placed on precedent in law, could use this methodology to better understand both general dynamics of legal citation and/or patterns in specific cases.
To return to literary studies, while our own initial project is focusing on academic literary criticism of the post-war period, the methodology could equally be applied to earlier aspects of literary reception. We are interested to examine whether patterns of citation change measurably since English literature is established as a discipline in the late nineteenth century. Going even further back, our methodology could equally be applied to quotations of literary works in non-academic formats, whether non-fiction (newspapers, journals, essays) or other literary works (citations of Shakespeare in Romantic poetry, say).
Our own next steps in this project will be to expand our corpus of quoted texts. First, we are working with JSTOR Labs to expand our Middlemarch critical corpus to the full collection of several thousand articles. Next, we hope to answer the questions: do similar patterns of citation apply to Eliot's other novels? Do they apply to novels by Dickens, or by Austen? These analyses will allow us to begin to formulate general hypotheses about patterns of quotation within literary scholarship of recent decades, while also offering unprecedented insights into the works of each of these authors.
[^1]: Dennis Tenen, “Digital Displacement,” in Futures of Comparative Literature, ed. Ursula K. Heise, Dudley Andrew, Alexander Beecroft, Jessica Berman, David Damrosch, Guillermina De Ferrari, César Domínguez, Barbara Harlow, and Eric Hayot (London: Routledge, forthcoming in 2017).
[^2] Brožek, Josef. “The Echoes of Wundt’s Work in the United States, 1887–1977: A Quantitative Citation Analysis.” Psychological Research, vol. 42, no. 1–2, 1980, pp. 103–107.