Distant Horizons: Digital Evidence and Literary Change by Ted Underwood (review)

The cover of Ted Underwood's Distant Horizons: Digital Evidence and Literary Change, a self-help-style image of remote mountains at dusk, alludes to the vistas of literary history Underwood proposes to reveal through his computational method and perspectival modeling. We see the world as flat from where we stand; only at distant horizons is the curve of the globe perceptible. The cover evokes the book's calming as well as inspiring purpose: nature-loving humanists should not fear machine-learning algorithms; textual analytics still care about aesthetics, as Underwood states. The title invokes Franco Moretti's distant reading, in spite of Underwood's divergence from Moretti, in particular from the latter's claim that genres rise and fall within one generation. Traits of science fiction, for example, are more persistent across centuries than criticism on that genre has asserted. Although no one would mistake this book for a contribution to contemporary Victorian studies, it is of broad interest as a kind of antithesis of C. P. Snow's The Two Cultures and the Scientific Revolution (1959): beyond "technophilia and -phobia," it emphasizes the "grounded character of supervised algorithms—which explicitly acknowledge the circularity of historical interpretation, describing human artifacts only in relation to other artifacts characterized by humans" (158–59). The book interprets several deft experiments in ways to classify changing characteristics of novels (with one foray into a poetry corpus) over several centuries, guiding readers to the data and documentation in appendices and online.

I would recommend this book to Victorianists interested in the novel and textual studies, as well as to students of periodicals or reception history, specialists in information science, or anyone professionally tasked with assigning genre in library or publishers' catalogs. Underwood's work, in earlier iterations and this book itself, is already having an impact in digital humanities and cultural analytics. Certainly, an initiated reader could learn how to replicate Underwood's models and dispute or approve how he deploys them; there are such readers already doing so. Underwood genially invites us to look at the recent findings of quantitative methods first applied to curated samples and then tested on large corpora (more than 90,000 novels). These collaborative experiments produce, with high rates of accurate prediction, measures for some perhaps unexpected trends in characterization of male or female characters or in traits associated with literary prestige, with clearly explained visualizations. Underwood is a reader-friendly writer who does not mimic Moretti's hypnotic pronouncements: "[d]istant reading is simply a new scale of description. It doesn't conflict with close reading" (xvii). As he measures the qualities of the works of poetry or fiction that were reviewed in a selection of respected journals, including The Westminster Review and Fortnightly Review as well as The New Yorker and The New Republic, he comes to the view that "aesthetic judgments … change very slowly," though they are "historically contingent." Numbers, he says, "allow us to be patient" with the shifting terrain; "they can characterize literary value as durable without affirming it as universal" (88). Surely the new formalists and various schools of criticism have reached similar views without the aid of enumeration on this scale.

While Underwood is wise about the challenges of large-scale textual analysis, the horizons of his study are both soft-focus and not distant enough. A horizon that Underwood does not claim to espy are the varied ranges, peaks, and foothills of digital humanities worldwide that deploy computers and digitized texts (not to mention models of buildings, artifacts, and geographical or social data) without engaging in machine learning and modeling of words. Many projects in this spectrum of digital humanities, not known as distant reading, are quantitative and span centuries but are well adapted to the recognition of historical contingency and even urgent advocacy within the big picture. To take a patient view of the longue durée seems like an evasion of many...