Computational approaches to literary studies are too often conjured as abstract shibboleths hailed as either messianic saviors or apocalyptic portents. Hoyt Long's new work, The Values in Numbers: Reading Japanese Literature in a Global Information Age, is perhaps the first book in the field of modern Japanese literary studies to actually utilize such tools. It masterfully cuts through this reified binary logic of “techno-utopian” endeavor (p. 277) versus “patently heretical venture” (p. 278) to demonstrate the potentialities and limitations of digital analytical techniques with sharp lucidity.
Those following Long's run of articles on network analysis and topic modeling will know how his scholarship moves beyond proof-of-concept toward full-fledged execution of modeling programs to demonstrate “how these new frames for interpretation alter the stories we are accustomed to telling” about literary texts (p. 173). Along with Andrew Piper and Richard So at McGill, Long, who co-directs the Textual Optics Lab at Chicago, is a leader in the melding of quantitative analytics and established methods of close reading to produce new understandings of corpus and canon. We are now treated to his long-awaited book-length study, which lives up to its promise through Long's “conviction that quantitative methods can productively refigure the relations of things, including texts, by projecting them into equivalence spaces that expand the possibilities of comparison via a deliberate reduction of dimensionality” and thus “generate new theoretical fictions” that “contend with existing forms of knowledge” (p. 269).
Long is clear that “numbers are a vehicle for the realities projected onto them by users” (p. 20), and his “modest proposal” is to take practices of “computer science, linguistics, and sociology and put them in dialogue with practices of knowledge making in literary study” (p. 271). The intellectual honesty of this endeavor encourages the reader to turn a critical eye toward the historicity of “disciplinary habits of mind that decide what to bracket and how” (p. 270), “exposing and reflecting the normative limits” (p. 206) at work in received methods of selection and analysis and clarifying the biases at play when we choose what and how to read, and how to assess claims made through those readings.
Each chapter combines a pair of conceptual terms with a particular corpus of texts, a computational methodology, and a theoretical problem. In each, Long weighs the “potential of computational models to serve as interlocutors” (p. 170), and while he does not spell out the programming in a how-to sense, the book can serve as a guide for how to responsibly conceptualize, design, build, assess, revise, and execute a computational model and analyze its results. Yet, The Values in Numbers is by no means a mere primer for computational analysis, but an extensively researched, richly erudite work of Japanese literary history in its own right. Every chapter is dense enough to challenge another quantitative limit: the word count of this review. All the more reason to read the book in its entirety.
Chapter 1, “Facts and Difference,” will feel the most familiar to those in the field. The chapter begins with Natsume Sōseki's well-known attempts to devise formulas for principles of literature in Bungakuron, then traces a fascinating, largely unexcavated genealogy of numerical approaches to literary study in Japan, including the linguistic thought of Hatano Kanji, Nakamura Akira's sampling methods, and Yasumoto Biten's statistical approaches, as well as Komori Yōichi's more recent work on stylistics. This trajectory allows the reader to locate contemporary computational methods within a longer history of attempts to create “equivalence spaces wherein new facts of literature could be generated” (p. 21).
Chapter 2, “Archive and Sample,” takes up the open access Aozora Bunko archive to ask larger questions about information management, access, and the “disciplinary practices and epistemological assumptions” and “representativeness” (p. 75) at stake both in the selection of texts and the theoretical claims made through those selections. The chapter considers “competing theories of how to generalize from the particular” through discussion of Aozora and the mid-twentieth-century zenshū compendia central to the crafting of canon heretofore. Long postulates that “the digital archive is a chance to rethink . . . the various meanings of the individual text and what textual interpretation stands to gain (and lose) from projecting this multiplicity” (p. 127) at scale.
Chapter 3, “Genre and Repetition,” explores the “slippery ontological status” (p. 128) of the I-novel (shishōsetsu) by building a model to synchronically and diachronically compare stylistic techniques and assess “the belief that I-novels represent a coherent genre” (p. 131). Examining repetition as a fictional mode, Long crafts a model that draws on word frequency and diversity through type-token ratio. Regardless of one's interest in the I-novel, the chapter's brilliance lies in how Long again puts the conceptualization of the model and its results into conversation with both the literary historical debates of the moment and contemporary critical and theoretical problematics.
Chapter 4, “Influence and Judgment,” considers the circulation and transformation of stream-of-consciousness narrative technique to examine the “nature of literary influence across linguistic and cultural borders” (p. 170) and how “judgments of equivalence or nonequivalence” and “algorithmic competence” can reframe the “(im)possibility of transnational and translingual influence” (p. 171). Long devises a syntactical and lexical model of “linguistic and stylistic features with distinct distributional tendencies” (p. 174) and describes the discursive figuration of stream-of-consciousness by authors such as Kawabata Yasunari and Itō Sei, identifying a moment in which experiments pushed “the boundaries of what it looked like to narrate the internal workings of the mind,” while making a general argument for how “latent formal innovations . . . prepare the ground for foreign forms” (p. 205).
Chapter 5, “Discourse and Character,” teases out a methodological reading of Edward Said's Orientalism about how “constricted vocabulary” and “categorizing logic” become “affixed to racialized Others under empire” (p. 213) through the description of characters and examines the voices of racialized colonial subjects in imperial-era discourse. Long uses sampling to ask “how numbers might intercede . . . in th[e] long running tension between general context and individual utterance in the study of racial discourse” (p. 210) by analyzing how the association of “discrete lexical items” with racial identifiers, in tandem with close readings of stories by Kikuchi Kan, Kunieda Shirō, and Wang Chaoxing.
Long has given us a truly pathbreaking book, one that at last brings the field of modern Japanese literary studies into ongoing methodological conversations that are rippling through the humanities. As someone who barely passed introductory computer science in high school, will I personally begin incorporating computational modeling into my scholarship? Unlikely. Will I eagerly look forward to future work by Long and the scholars sure to follow in his footsteps in opening new horizons for the field? Absolutely.
