Seven years ago, in a small room at the top of a winding stair in Palo Alto, a strange experiment got underway. A group of literary critics began to build a laboratory. Labs themselves are not unusual on university campuses, but in 2010 it was far from clear what a group of critics would do in one. In fact, it wasn’t clear that critics could work productively in groups at all. Literary criticism is usually written alone. However, this volume of essays from the Stanford Literary Lab, Canon/Archive, has 14 authors. With scores of graphs and tables, the book looks like a collection of scientific papers. And yet, the suspenseful narratives that link the graphs don’t sound at all like (modern) science. Where did this hybrid creature come from?
If you read the newspapers, you’ll suspect the answer is “digital humanities.” But newspapers focus on recent symptoms. The deeper explanation of this puzzling hybrid is that the humanities only divided from the social sciences about a century ago, and the boundary between the two projects never fully solidified. There is actually a long tradition of quantitative reasoning in the humanities, stretching from the economic history of the Annales school to the survey-based interpretations of literary culture in a book like Janice Radway’s Reading the Romance (1984). Radway’s surveys might even be described as “experiments” on readers.1
But for most of the 20th century, numbers only seemed to illuminate a few marginal questions about literature. Sociologists of literature could use numbers to describe the responses of readers, and book historians could measure sales, but numbers couldn’t reach inside the volumes to describe the life within. There was, admittedly, a different group of scholars using numbers to describe style—in order to identify, say, syntactic patterns especially typical of Jonathan Swift.2 But that, too, looked like a specialized project, a rather dry linguistic one. In short, quantitative arguments only seemed to work inside a few bounded domains. Truly significant conclusions about culture required creative interpreters who were free to link tiny descriptive details to vast social forces, unencumbered by the rules of math.
As long as quantitative methods were confined to special enclaves, most literary scholars didn’t feel compelled to engage them. But over the last 30 years, the enclaves have joined to produce a practice of quantitative interpretation that is no longer purely sociological, or purely linguistic, but able to range freely across the spectrum from single words to social trends. Computers are certainly useful in this mode of interpretation, but they aren’t the new element. (Literary scholars have been using computers for at least the past 50 years, after all.3)
Will this book have any influence on the way other literature professors teach and write about literature?
The spark that gave quantitative arguments deeper significance was a new human connection between scholars. Canon/Archive illustrates this well, because the research gathered here arguably got its start when Franco Moretti, who had experience connecting literature to social history, began in the first decade of this century to collaborate with Matthew Jockers, who had experience connecting literature to linguistics. Stanford is by no means the only place where linguistic and social approaches to literature have fused. In the 1990s, the ARTFL project at the University of Chicago pioneered this approach; today, similar projects are underway at universities around the world.4 But since this is a new kind of research, stretched across an unmapped space between disciplines, each project understands itself differently.
In particular, the scholarly cultures at different universities have developed very different ways of grafting humanistic interpretation onto scientific experiment. The Stanford Literary Lab has approached this problem in a particularly distinctive, dramatic way. Instead of taking the experimental method as an object with a fixed meaning that could be imported from the sciences, these essays seem determined to invent it from scratch, inside the humanities, as a narrative form.
Experiment is presented here not just as a test of reliable knowledge but as a style of intellectual growth: “By frustrating our expectations, failed experiments ‘estrange’ our natural habits of thought, offering a chance to transcend them.” At moments, the point of experiment seems to become entirely aesthetic. In the book’s introduction, Moretti admits that he set out to write “a scientific essay, composed like a Mahler symphony: discordant registers that barely manage to coexist; a forward movement endlessly diverted; the easiest of melodies, followed by leaps into the unknown.”
This account of the book’s aesthetic achievement is candid, immodest, and accurate. The essays within are unified by a deliberately wandering structure, which keeps its distance both from scientists’ predictable sequences (methods → results → conclusions), and from the thesis-driven template that prevails in the humanities (counter-intuitive claim → evidence → I was right after all). Instead, these essays become stories of progressive disorientation, written in the first-person plural, and arriving at theses that were only dimly foreshadowed.
This narrative form has given the Literary Lab a coherent authorial persona, which may lead readers to assume that the experiments gathered here are also unified by shared methods and theories, more or less identified with Moretti. That would be a mistake. The Literary Lab is genuinely a collective project, and these essays have been shaped by many different approaches to the literary past. To fully appreciate this volume, a reader needs to resist the illusion of unity.
The slash in the title, Canon/Archive, is the first fracture we ought to notice. Back in the year 2000, Moretti presented the project of “distant reading” as a turn away from canonical authors and toward a mass of volumes forgotten in the “slaughterhouse of literature.”5 But the opposition between a “canon” of great works and an “archive” of obscure ones, while still central to this book’s title, is no longer central to its argument. Many of the essays gathered here discuss prominent writers along with obscure ones, and when the ninth chapter finally contrasts the two groups of writers, it finds them moving in the same direction: “The canon [i.e., the literary practice of prominent writers] precedes the archive by about fifteen to twenty years; but the historical trajectory is the same.”
Nor is this book really unified by an argument about the relative importance of history and form. The “quantitative formalism” mentioned in the book’s subtitle is sometimes foregrounded, as when the book discusses the structural function of paragraphs in narrative. But other chapters are mostly historical. The stylistic analysis in chapter 6 allows Ryan Heuser and Long Le-Khac to advance an argument about the transformation of the 19th-century novel by urbanization. Some arguments look beyond literature to other institutions. Chapter 7, “Bankspeak,” shows how the reports of the World Bank gradually transmuted “social forces into abstractions” as the Bank was bureaucratized. (The essay that became this chapter had dramatic consequences at the Bank itself, arguably leading to the fall of a chief economist.)
The most important fracture in the book involves the nature of the connection between numbers and interpretation. Several of the essays make this connection with statistical models. In chapter 1, for instance, Michael Witmore reveals that a model of textual similarity based purely on word frequencies can group Shakespeare’s plays into comedies, histories, and tragedies. But when Moretti describes the book’s methods in the conclusion, he downplays the interpretive connection provided by models, in order to tell a story that leaps from observation of opaque “patterns” to the “discovery of a causal mechanism.” This account again reveals Moretti’s commitment to framing the work of the Lab as a humanistic narrative. Scientists don’t usually understand their methods as a process of pure induction that produces meaning at the last moment; instead, they tend to begin with a hypothesis, and find a way to test it (often using a model).
Experiment is presented in these essays not just as a test of reliable knowledge but as a style of intellectual growth.
Of course, for many readers in English departments, debates about different ways of using numbers matter very little, because many literary scholars see numbers as wholly antithetical to literature. Although quantitative reasoning has always preserved a foothold in literary study, it has never been central to scholars’ conception of their mission, which is shaped instead by a view of literature as a necessary counterweight to modes of thinking that prevail in the sciences and the market.6
This brings us to the real question about Canon/Archive. Will it have any influence on the way other literature professors teach and write about literature? If new ideas always produced social effects, the answer would be a clear yes. This book offers new, consequential arguments about topics that have long been central to literary study: the nature of genre, the history of the novel, the stylistic strategies that make some writers more prominent than others.
But if you want to be imitated, it helps to do something other people can easily imitate. If I want to imitate New Historicism, for instance, I can look for surprising connections between works of literature and historical events, and start my essays with anecdotes. Since literature majors largely know how to do those things, New Historicism was able to spread rapidly.
If I want to write like the Literary Lab, on the other hand, I need to learn methods that are not taught in literature departments (experimental design, programming, statistics). Or I need to find a group of collaborators who already have those skills—which isn’t much easier. While many campuses are now discussing “digital humanities,” it isn’t yet clear that diffuse interest in computers will produce the curricular changes needed for experimental inquiry. Many literary scholars still understand their discipline as a counterweight to modes of thinking that require numbers. Quantitative methods are doing very well in a few places (McGill, Northeastern, Carnegie Mellon), but on most campuses a student would need to look outside literature departments in order to learn how to imitate Canon/Archive.
In short, there is no danger that a quantitative approach to literature will spread like wildfire, displacing other modes of interpretation. The project is difficult, and curricular winds blow in a different direction. On the other hand, there will certainly be more books like this one. Some are already in press. It has been clear for several decades that quantitative methods could produce real discoveries about literary history; this volume also proves that those discoveries can be communicated in a lively, suspenseful way—perhaps even with an echo of Mahler.
- Janice Radway, Reading the Romance: Women, Patriarchy, and Popular Literature (University of North Carolina Press, 1984). ↩
- Louis T. Milic, A Quantitative Approach to the Style of Jonathan Swift (Mouton, 1967). ↩
- The journal Computers and the Humanities was founded in 1966. ↩
- For the ARTFL project, see Mark Olsen, “Signs, Symbols, and Discourses: A New Direction for Computer-Aided Literature Studies,” Computers and the Humanities, vol. 27, no. 5/6 (1993/1994). ↩
- Franco Moretti, “The Slaughterhouse of Literature,” Modern Language Quarterly, vol. 61, no. 1 (2000). ↩
- For the early history of this assumption, see Raymond Williams, “Mill on Bentham and Coleridge,” in Culture and Society: 1780–1950 (1958; Columbia University Press, 1983). ↩