You Can Read (and So Can This Computer)

REMEMBER THE MOVIE Smart People from 2008? Of course you don’t. It wasn’t a very good movie and, had it not been both filmed and set in the English Department at Carnegie Mellon University, where I did my graduate studies, I wouldn’t remember it either. Its cast includes Sarah Jessica Parker, Thomas Haden Church, Ellen Page, and, most perplexingly of all, Dennis Quaid playing a curmudgeonly English professor. Quaid’s character is a direct descendent of Michael Douglas’s Grady Tripp from the 2000 film adaptation of Michael Chabon’s Wonder Boys (1995), a movie that had been shot at the exact same location only eight years before: he’s ill-tempered and disheveled, prone to appearing in public in his bathrobe, dismissive toward his students, and engineered, it would appear, for the sole purpose of turning the movie-going public against the very idea of English studies and English professors.

Unlike the marginally lovable Grady Tripp in Wonder Boys, though, Quaid’s Lawrence is a scholar, not a creative writer. His struggle to find a suitable name for his forthcoming scholarly monograph serves as a point of conflict in the film. At last, his daughter offers a suggestion that his publisher, sensing controversy, adores: You Can’t Read. The book’s title is meant as an illustration of how the public views academics of this kind, but what passes for a joke in the film actually conceals a level of sincere commentary. “Who are these jerks who make their careers out of labeling others as inept and stupid?” is the question that floats just beneath the surface of this otherwise extremely frivolous movie.

It’s a question that has been similarly dangling over the field that’s known as the digital humanities (DH) for some years now. This is because digital humanists have, quite often, fixated on pointing out the flaws in the their peers’ methodologies. Every DH-inspired monograph published since 2000 might as well be titled You Can’t Read, for that has been, as many see it, the recurrent punch line in these works. Franco Moretti, for instance, promoted a version of this argument in his “Conjectures on World Literature,” which appeared in the New Left Review in 2000 and spurred the practice (and the book) known as “distant reading.” Thirteen years later, Matthew L. Jockers appeared beating the same drum in his Macronalysis (University of Illinois Press, 2013), complaining that traditional literary analysis — also known as close reading — amounts to little more than “unsupportable speculation” and is thus “impractical as a means of evidence gathering.” This whole vein of critique is fueled by a spirit of admonishment: you can’t read the Western canon in an age of globalization; you can’t read individual works in an age of digital archives and libraries; you can’t read for subjective response when there’s objective data to be studied — on and on it goes, dispensing the word can’t with miserly glee.

On the surface, Andrew Piper’s recent work Enumerations: Data and Literary Study might look like another contribution to the broader field of can’t studies, as the digital humanities may alternatively be known. Piper, though, insists on the term “computational analysis” instead and steers entirely clear of all mentions of the “digital humanities,” perhaps in observance of the way that DH has become associated with castigation and judgment. Rather than telling others what they can and can’t do, Piper focuses on showing what computational analysis itself can do, stating that “we need to let the arguments of our research make the case for why this work is important.” This isn’t to say that Piper departs completely from the tradition of Moretti and Jockers, but that he aspires to a level of methodological brokering; he wants to temper his field’s reliance on the rhetoric of can’t. Conversations occurring among those interested in both computational analysis and literature, he observes, “have to date been far too hermetic,” and Enumerations seeks to widen both the scope of the conversation and its potential audience.

To that end, Piper’s study, which uses computational analysis to test vast archives of texts against a couple of broad “hypotheses,” is marked by a surprising amount of feeling and warmth. Feeling impinges upon the cold realms of data and computation, as when Piper interrupts a discussion about linguistic constraint and character development in chapter five with the comment that, “And yet when I read characters, I also feel difference.” These admissions align with some of Piper’s earlier work, including his elegant Book Was There (University of Chicago Press, 2012), and show that he has not lost sight of the affective possibilities of reading, even while his work hinges on outsourcing much of the work of reading to computer models. A similar moment occurs earlier in chapter one when Piper uses that much-maligned technique — reading — as subjective confirmation of a “suggestion” produced by a word-embedding model. The model, which has been produced in order to show evidence of “semantic diversity” within a specific set of poems, has wound up showing, instead, rates of standard deviation within individual poems. To Piper, this “suggests that these poems oscillate slightly more strongly between sentences that are more similar and less similar, a tactic that might help to foreground the sense of contrast that I felt when I read them.” This is characteristic of the way that Piper, in this work, uses computational analysis as a tool for not just for the assessment of literary language but for self-reflection as well. Sometimes the computer shows him what he wants to see, but often it doesn’t, and in between either of these outcomes rest profound admissions about the human investments and stakes that structure these supposedly objective, superhuman forms of information processing.

While it differs in its tone from much recent work occurring at the intersection of literature and computational analysis, though, Enumerations also resembles that work in certain, important ways. Chief among these is its focus on method. Piper announces his intention to sideline “debates” about methodology early on, observing that “‘debates’ has emerged as a kind of default genre within the field.” Instead of being about method, he declares, this is really a book about literature and about reading. These declarations emerge as part of an admirable attempt to tie his work to more traditional forms of humanities scholarship and to pave over the methodological divisions that have been the cause of so much “debate.” But try as he might, Piper can’t avoid the subject, and many of his “demonstrations,” which take the place of readings in his individual chapters, circle back to discussions of method. His chapters are furthermore organized around the prospect of trying and testing different computational methods, including using extraction of regular expressions to show the “economy of literary punctuation,” vector space modeling to show “emplotment,” topic modeling, machine learning, and so on. After each test — and sometimes before it, as well — comes an assessment of how well the computational strategy works as a tool for assessing the corpus, or range of texts, that has been submitted to it. In this way, chapters appear bracketed by the very kinds of debates that Piper has sought to avoid participating in, although I would argue that this bracketing is in fact quite advantageous.

For instance, chapter one, which works from a corpus of 75,000 poems to document rates and frequencies of punctuation across historical periods, begins by defending that practice with reference to the French philosopher Georges Bataille and his theories of excess. As Piper puts it, “computation enables us to inhabit, and more importantly to see, […] those spaces of aesthetic expenditure and luxury that offered keys to understanding human beings for Bataille.” In other words, quantitative methods may permit us to grasp the magnitude of “excess” described by Bataille, releasing it from the realms of the theoretical and the abstract. Piper assumes this defensive position once again toward the end of the chapter, arguing that quantitative studies (such as the one he has just reenacted and described) can guide us toward new ways of “testing and inquiring that can help us generalize about the group as a whole […] Number is a lens.” The word a is especially important here: Piper is not insisting that scholars use this particular lens, merely giving them permission to experiment with doing so.

One of the advantages of studies of this kind — that is, of studies that focus on method almost in spite of themselves — is that they rewrite the narrative of literary history by downplaying talk of “periods.” Others within the field, most notably Ted Underwood, have observed as much previously. Underwood, for instance, in his Why Literary Periods Mattered (Stanford University Press, 2013), notes that “by the end of the twentieth century” — that is, around the time that Moretti and his ilk arrived on the scene — “this model of literary history had lost some of its authority.” Piper’s Enumerations serves as a compelling testament to the idea that periodization matters less these days and not at all when it’s machines, not humans, that are doing much of the reading. In his demonstrations, he toggles between discussions of the classics, to 19th-century literature in German, to 20th-century poetry in English, showcasing a broad range of commonalities and trends in his quest for generalizable understandings of how (Western) literature works when it works. This is a quest that leads him from Aristotle to Amiri Baraka and thus mounts a claim for another, secondary advantage associated with this kind of reading: it dispenses with debates surrounding canonicity, if not necessarily with the idea of canonicity itself. Indeed, in place of the standard Western canon, Enumerations features a highly idiosyncratic, “curated” selection of texts — a private canon, if you will — in place of the kinds that previously formed by way of critical consensus, however problematically. The phrase “curated by my lab” is the one that Piper repeatedly uses to describe this new kind of canon, though he does not pause to mull the links between human curation and subjectivity, which computational analysis would, of course, have us avoid.

Another benefit of computational analysis, as Piper sees it, stems from the aforementioned merits of visualization. Enumerations is richly laden with tables, word clouds, and two-dimensional models that help demonstrate this. Yet even the most compelling visuals — and some of them really are spectacular — require elucidation and support from close-readings of the text, as when Piper turns to the subject of characterization in chapter five. A series of models in this section culminates in a “collocate network” that surveys keywords used by female authors to express the idea of interiority, or that particular, winning combination of “thinking-feeling” that appears on display in 19th-century novels featuring strong, intelligent heroines. The visualization usefully distinguishes between different forms of interiority, separating “perception” and its attendant keywords (like “anxiety” and “unconscious”) from affect (made manifest through words like “tender”). But this evidence becomes clear only with the help of detailed readings of individual scenes: Piper analyzes windows as “spaces of retreat” for the female protagonists who appear in these novels, pivoting from discussions of quantity and data visualizations to a reading of windows as technologies of visualization. This is a move that corroborates, albeit in a roundabout way, his claim that modeling “encodes a togetherness to reading.”

Piper’s visualizations help to keep the reader on track, even as they are continually bombarded with data. And Piper is, thankfully, able to sense and to sympathize with those feelings of bombardment. In his discussions of period frequency in chapter one, for instance, he grasps the link between excess and feelings of exhaustion or meaninglessness, commenting that “the period, in its excessive poetic state, does not appear to have a particular kind of meaning associated with it.” This observation leads him to a conclusion that is as damning as it is profound: “More is the knowledge of its own failure,” he says, describing how the very prospect of excess erodes meaning. Too many, he reasons, is a situation that can only be solved via the logic of more. One can’t help but feel that Enumerations is guided by a very similar kind of logic. To the problem of “too many” it offers more — more methods, more strategies, more texts, more options. In place of the rhetoric of can’t, one discovers an almost boundless and thus daunting array of possibilities.

Piper formally concludes with the point that “data can help us see […] imbalances.” The whole of Enumerations encourages us to see how this might be the case but, to paraphrase Karl Marx, the point was never merely to see them, but to fix them. How does the study of excess help us to do that? How do we go about forming arguments about the importance of subjective textual construction and consumption — about the future of writing and reading, in other words in the midst of all that excess? Granted, this is a burdensome question to ask of any book, but it’s one that most literary scholars and practitioners now confront on a daily basis. If we insist as a discipline on doling out can’t and calling it scholarship, then we are bound to meet with protest and one-dimensional mockery coming from audiences outside the academy.

Which brings me back to Smart People. In the movie, Ellen Page’s character, who is called Vanessa, is an overachieving high school student who, we are led to believe, has the potential to turn out just like her curmudgeonly father. She sees this and admits it to her Uncle Chuck, complaining, “Everyone hates me.” Chuck’s response is a comical no-brainer: “If you tell people they’re stupid, they’ll hate you.” Piper’s Enumerations undermines much of the discourse of hate and can’t that surrounds the digital humanities, and that is no small feat — Piper himself admits to being a victim of such “antagonisms” in a particularly heart-wrenching section of the book’s acknowledgments. But insular discussions of method can only do so much when it comes to broadening expectations about what the humanities are for, or assuaging the public’s suspicion of them. For that, you still need a good story.

¤

Sheila Liming is an assistant professor in the English Department at the University of North Dakota. She is the author of What a Library Means to a Woman, forthcoming from University of Minnesota Press, and Office, forthcoming as part of the Object Lessons series from Bloomsbury.