Computers, Writing, Rhetoric and Literature

Return to previous page

"Textualizing" Literature: Barthes' S/Z

6. This incomplete or interrupted textualizing of disciplinary knowledge is visibly at work in-- is indeed the substance of-- Roland Barthes' S/Z (1970). I choose to focus on Barthes here both because his work exerted a significant influence on the development of literary criticism from the 1960s on, and because American theorists working on computer-mediated textual forms (Bolter, 1991b, 1991c; Moulthrop, 1991a; Landow and Delany, 1991; Landow, 1992) treat Barthes as a sort of ancestor-figure whose ideas were all the more remarkable because he "did not know about computers" (Bolter, 1991b, p. 161). As S/Z reveals, however, Barthes did know about computers. He evidently didn't know very much, but his awareness made a difference to the way he thought about literature and, consequently, to literary studies more generally.

7. I want to highlight two things about S/Z. First, information theory provides a general framework for Barthes' understanding of literary discourse. Second, the computer-- or at least the idea of the computer-- plays an important part in his thinking about reading and writing.
   Information theory

   Formulated by Claude Shannon and Warren Weaver in the late 1940s, mathematical information theory is at the heart of modern communications and computer technology. Addressing a practical problem with important economic and social implications, information theory provided a way to calculate with great precision how human speech could move through a "channel" while keeping distortion within acceptable limits. A radical simplification dependent, like Boolean algebra, Saussurean linguistics, and their descendants, on fundamental binary pairs (signal/noise, 0/1), the Shannon/Weaver theory also provided thinkers in many other fields-- genetics, for example, or economics, or the emerging field of computer science-- with a way to reconceptualize their areas of concern as information processes.

   It did so for Barthes as well, though it is difficult to say how much he knew either about the theory itself or the fundamental ways in which modern digital computers depend on it. Uniting Shannon's information theory and the binary Turing machines developed by British mathematician Alan Turing just before World War II (see, e.g., Bolter, 1984; Penrose, 1989; Hofstadter, 1985), the digital computer is the most potent embodiment of information theory except for the human genome. In S/Z it affords Barthes a metaphor for thinking about both reading and writing.

   Ourselves writing

   S/Z offers itself as a "reading" of Balzac's "Sarrasine," the story of a not-very-bright but oh-so-Romantic and culturally illiterate French sculptor, Sarrasine, who makes the artist's obligatory pilgrimage to Rome. Unaware that there are no female singers on the Roman stage and that soprano roles are performed by castrati, Sarrasine falls head-over-heels in love with a soprano named La Zambinella and is thereby doomed.

   Barthes' reading rests on a number of binary oppositions, of which the most important are those between readerly texts and the writerly text; ideal communication and noise; denotation and connotation; and the formalized sciences and literature. The primary opposition is that between the writerly text (which exists nowhere) in its infinite plurality, and which has as its goal to "make the reader no longer a consumer, but a producer of the text" (p. 4); and the readerly text, which presents a bland, smooth surface of "Repleteness" and casts the reader as a more or less passive consumer. The distinction between the readerly and the writerly is also an historical distinction, between the classical and modern conceptions of text. Classical textuality is of course represented by Balzac; modern textuality is impersonated by Barthes' own writing in S/Z: for "the writerly text," as he says, "is ourselves writing" (p. 5; emphasis in the original, that is, in the translation).

   Manhandling the text

   In order to shatter the realistic text's blandly confident illusion of unity, Barthes proposes to disrupt the Balzacian text. "The work of the commentary," he says, " . . . consists precisely in manhandling the text, interrupting it" (p. 15). In order to "obtain" at least the illusion of a parsimoniously "plural text," Barthes "star[s]" the text of "Sarrasine" and "cut[s] it up" into "a series of brief, contiguous fragments," or "lexias." These are "arbitrary in the extreme," yet each is carefully chosen as a stage or "space in which we can observe meanings" (p. 13) moving about like actors in a play or specimens in a Petri dish.

   Computing Connotation

   Connotation becomes a central concern. Barthes writes in Digression IV ("Connotation: For, Even So") that both scientific discourse and ideology have taken aim at connotation, seeking to eliminate its ambiguity and to force all discourse into precise denotative habits. (See also Hayles, 1991.) "Connotation," Barthes writes, "must therefore be rescued from [this] double contestation and kept as the namable, computable trace of a certain plural of the text" (p. 8; emphasis in the "original").

   It seems almost perverse to call connotation computable, since the word connotation itself connotes a freewheeling multiplicity of meaning that seems the very opposite of computability, like certain irrational numbers that "cannot be generated by any Turing machine at all" and are therefore defined as non-computable (Penrose, 1989, p. 50). [1] But it is precisely with respect to connotation (as the extension of the double entendre that Barthes discusses in the passage below) that information theory enters into Barthes' calculations, providing a key distinction between communication and noise:

   if we grant that the double understanding far exceeds the limited case of the play on words or the equivocation and permeates . . . all classic writing . . . we see that literatures are in fact arts of "noise"; what the reader consumes is this defect in communication, this deficient message; what the whole structuration erects for him and offers him as the most precious nourishment is a countercommunication; the reader is an accomplice, not of this or that character, but of the discourse itself insofar as it plays on the division of reception, the impurity of communication.... (p. 145)

   As William Paulson explains, "Mathematical information theory... begins by quantifying information: the information of a message can be measured as the number of binary bits required to encode it" (Paulson, 1991, p. 39). [2] Thus it is because Barthes treats connotation as a function of information that he can regard it as "computable" (though from a technical standpoint it would probably be more accurate to say that this is why the question of computability arises).

   And it is through this notion of computable connotation that reading acquires a certain programmatic aspect. Barthes writes that one name for the connotations he seeks to catalogue is "index" (p. 8). It is as if he were constructing a relational database in which the units of "Sarrasine" were indexed to one or more of the five codes (SYMbolic, SEMantic, REFerential, HERmeneutic, proairetic (ACT)) whose "grid" of intersection constitutes the text of Balzac's narrative. This would be a database in which the elements of the text-- the lexias-- would function as indices to their own connotation. [3]

   Bits, Bytes, and Briques: The Text as Program

   But Barthes is not constructing a database, as will become clear later on, and it is the programmatic aspect of "Sarrasine" itself that he seems most interested in. On four separate occasions, in his commentaries on lexias 39, 122, 141, and 279, he uses the word byte in a way which suggests that he envisions the text both as a sort of program and, at the same time, as the product or output of such a program.

   The word byte has to do with the way data is encoded for the computer: a byte consists of 8 bits, or binary digits, and effectively represents the minimum amount of space in memory necessary to store a single piece of information such as the letter S or the numeral 107, which as it happens is the numerical value assigned to the capital S by the American Standard Code for Information Interchange, or ASCII. Barthes seems to be using the word byte in a slightly unusual way, however-- or perhaps it isn't Barthes but his American translator, Richard Miller. The word Miller translates as byte is brique. It does not appear in any French dictionary that I can locate as an equivalent for byte; the closest I have been able to get is a definition listed as archaic by Robert, in which a brique corresponds to a million old francs. This could work metonymically, so that a million old francs corresponds to a million bytes. But Claude Levy, a French colleague who attended the seminar from which S/Z emerged and who heard a much earlier version of this chapter in 1992, suggested that byte might well have been a mistake on Miller's part, that brique carries with it the sense of a building block that would correspond more closely to Barthes' references to subroutines and "sections of program" (Claude Levy, personal communication, 10 April 1992).

   A section of program fed into the machine

   Commenting on lexia 122, Barthes indicates that the byte/brique encodes an entire sequence of events in the narrative: "The short episode which begins here . . . is a byte (in computer terminology), a section of program fed into the machine, a sequence equivalent as a whole to only one signified . . ." (p. 78). Barthes is careful to specify "computer terminology" as authorizing his usage; the byte represents a unitary, unequivocal identification of "only one signified." Thus is connotation tamed, brought under control, reduced to unitary meaning. But perhaps more importantly, the byte here is not merely a piece of data upon which the program operates: it marks a code segment, a "section of program fed into the machine," suggesting that Barthes is imagining "Sarrasine" itself as a sort of program.

   The real is not operable

   Just a few pages later, Barthes makes it clear that designating sequences as bytes marks their separation from the real. The "discourse" of realist fiction, he writes, "has no responsibility vis-a-vis the real: in the most realistic novel, the referent has no 'reality': suffice it to imagine," he goes on, "the disorder the most orderly narrative would create were its descriptions taken at face value, converted into operative programs, and simply executed. . . . In short . . . what we call 'real' (in the theory of the realistic text) is never more than a code of representation (of signification): it is never a code of execution: the novelistic real is not operable. To identify-- as it would after all be 'realistic' enough to do-- the real with the operative would be to subvert the novel at the limit of its genre. . . ." (p. 80).

   Thus the novel is not only a program, but is simultaneously the output of a program as well: those who think the realist novel could generate the real (as Joyce hoped people of the future would be able to recreate the Dublin of June 16, 1904) are in for a rude surprise. The purpose of executing the program Barthes toys with imagining here is not to generate the real, as one might have imagined, but rather to generate the story, "Sarrasine." Anticipating what Artificial Intelligence researchers call Automatic Story-Generating programs (see Schank, 1984; Ryan, 1991; Bolter, 1991), Barthes-- or perhaps literary discourse-- has found a way to preserve the opposition fiction/reality after all.

   Print's ground

   S/Z is the fruit of a skirmish between a conception of text as defined by print at the height of its maturity in the 19th century (see Bolter, 1991, pp. 114-15), and an emergent conception of text as defined by information technology, in its perpetual immaturity, in the late 20th century. Or call it a flirtation between text-as-print and text-as-information. No wonder, then, that it's so difficult to tell just how far Barthes intends to go; no wonder that for all his masterly manhandling of the readerly "Sarrasine," Barthes can't quite manage to bring the "writerly"-- which is ourselves writing, ourselves no longer mere consumers but instead active producers of the text-- clearly into view. For the writerly is on the other side of the electronic frontier, in the domain of information. S/Z takes place on print's home ground, which is to say no more than the obvious, that S/Z is a book about a book, that it is deployed, enabled, and constrained by the conditions of print.

   Fearful symmetry

   Writing in a different but closely related context, that of scholarly editing, Jerome McGann (1995) remarks that

   When we use books to study books... the scale of the tools seriously limits the possible results.... This symmetry between the tool and its subject forces the scholar to invent analytic mechanisms that must be displayed and engaged at the primary reading level -- e.g., apparatus structures, descriptive bibliographies, calculi of variants, shorthand reference forms, and so forth. The critical edition's apparatus, for example, exists only because no single book or manageable set of books can incorporate for analysis all of the relevant documents. In standard critical editions, the primary materials come before the reader in abbreviated and coded forms.

   Following McGann's lead, we might say that the underlying "symmetry" between Barthes' heavily coded text and Balzac's-- the very symmetry Barthes wishes to repudiate-- engenders the monstrousness of S/Z. This, I think, is why the full text of "Sarrasine" appears as an appendix to S/Z. But this isn't quite the text of "Sarrasine," or rather, it is the text of Barthes' "Sarrasine": it includes superscripts that indicate where each lexia with its appended commentary begins. Thus Barthes' restructuring is now coded into the text of "Sarrasine," and that text is loud now, noisy.

   Go to next section.

   ---

   Page:"La Zambinella: 'Textualizing' Literature: Barthes' S/Z"
   Copyright (c) 1996