Algorithmic Criticism and Barthes: Two Perspectives on the Sentence

Literary criticism, magpie discipline, has long benefitted from borrowing techniques from other fields--philosophy, history, linguistics. In recent years, criticism adopted methods for dealing with aggregate data and text analysis originally developed to manage large amounts of data--more poetically, the “wine-dark sea” of texts. Michael Witmore, Director of the Folger Shakespeare Library used Homer’s epithet to foreground the size and mystery of looking at many texts at once in the title of his blog, (created almost exactly nine years ago on June 22, 2018) dedicated to the study of “literary and cultural history at the level of the sentence.” The sentence, as the largest unit of syntax, seems simultaneously to be the smallest unit to start investigating syntactic patterns, a trace of the literary critic’s commitment to organized units of meaning.

Many of the models most closely associated with algorithmic criticism, however, don’t work at the level of the sentence. Rather, many descriptive and predictive models look at feature sets--a range of lexical and grammatical features--instead. Even when I was involved in a project studying the sentence in the Stanford Literary Lab, we looked at relations among smaller grammatical units within each sentence in texts that had been tagged according to part-of-speech (by the Stanford Parser used by computational linguists). My book, Reductive Reading, is about the unit of the sentence: it looks at certain syntactic patterns, like shifts between past and present tenses or certain types of speech tags, the he said/she saids that attribute speeches to a character. Such patterns help identify sentences that are like cousins to each other, genetically similar and unpredictably different from one another. The specialized uses of language I look at (like the fictional narrative past tense or the idiosyncratic habits of a single author) appear in sharper relief against a backdrop of how language usually works, a perspective provided by computational linguistics.

Reductive Reading is a study of Victorian moralizing carried out by a person used to explicitly reducing the terms of inquiry to the kinds of claims that can be supported by the analysis of feature sets. What’s true, though, is that my book is equally indebted to Roland Barthes’ S/Z, in which Barthes breaks down, bit by bit, a story by Balzac that hinges on the gender of the character called “Sarrasine,” not “Sarrazine.” Barthes names his book after the one-letter difference that reveals how a linguistic convention for gendering names is no guarantor of gender. S/Z goes on to identify the codes at work in each of the story-bits to demonstrate that denotation, which we might think of as something like lexical and grammatical meaning, is in fact an “old deity, watchful, cunning, theatrical, foreordained to represent the collective innocence of language”--it obscures the many meanings each phrase encodes.  At times, the power of denotation seems to live in the sentence itself: “There is a force in the sentence (linguistic entity) that domesticates the artifice of the narrative, a meaning that denies the meaning.” My own book is a close study of the “force in the sentence” in relation to the narrative meanings it obscures--and enables.

In the Literary Lab pamphlet on “Style at the Scale of the Sentence,” the sentence was such a complex object of study that we decided we could best approach it through the relation among clauses, which served as a proxy for something more difficult to trace. In Barthes’s notion of the sentence, the sentence is a treacherously inadequate proxy for meaning. My “reductive” approach is deeply informed by the post-structuralist premise that the meaning of a sentence is encoded in, but not reducible to, its grammatical logic. To consider approaches to the sentence by Barthes and the Stanford Literary Lab together suggests that the value of frankly acknowledging the limits of an approach is not in its evocation of perfect clarity (denotative triumph), but rather in the way it sets the limits of one approach against the limitlessness of interpretive possibility.

Sarah Danielle Allison is an assistant professor of English at Loyola University New Orleans. She is the author of Reductive Reading: A Syntax of Victorian Moralizing. On June 29, 2018, Sarah Allison will present a paper on "Setting Limits" at DH2018.