III

Consider, now, a second claim, this time from Barthes's well-known essay, "From Work to Text," which is as close to a canonical formulation of what Krauss originally wished to borrow as one could possibly find:

In opposition to the notion of the work of art or literature [says Barthes] there now arises a need for a new object, one obtained by the displacement or overturning of previous categories. This object is the Text. . .. The Text must not be thought of as a defined object. It would be useless to attempt a

material separation of works and texts. . .. A very ancient work can contain "some text," while many products of contemporary literature are not texts at all. The difference is as follows: the work is concrete, occupying a portion of book-space (in a library, for example); the text, on the other hand, is a methodological field.^[15]

Two distinctions need to be made: first, there is no doubt that Barthes never means to abandon a reliance on referential facilities, all the while he clearly intends to subvert conventional views about reading a text (for instance, views somewhat like Monroe Beardsley's New Critical view of interpretive reading); second, there is no doubt that Barthes does mean to constitute, by a certain sort of reading and serial rereading, that "object" that thereby becomes (what he calls) the Text. The notion of the Text, for Barthes, therefore, is not the notion of an antecedent referent to which interpretation is directed but rather the notion of what is productively yielded by interpretively addressing "something else" that, in the ongoing (serial) process of reading and rereading, is uniquely affected by that very process.

It is impossible to pursue the theme without citing Barthes's famous distinction between the "readerly" and the "writerly" (the lisible and scriptible ) offered at the very opening of S/Z, which is close in spirit (and even language) to the paper just mentioned:

Why is the writerly our value? [asks Barthes.] Because [he answers] the goal of literary work (of literature as work) is to make the reader no longer a consumer, but a producer of the text. Our literature is characterized by the pitiless divorce which the literary institution maintains between the producer of the text and its user, between its owner and its customer, between its author and its reader. This reader is thereby plunged into a kind of idleness—he is intransitive; he is, in short, serious: instead of functioning himself, instead of gaining access to the magic of the signifier, to the pleasure of writing, he is left with no more than the poor freedom either to accept or reject the text: reading is nothing more than a referendum. Opposite the writerly text, then, is its countervalue, its negative, reactive value: what can be read, but not written: the readerly. We call any readerly text a classic text.^[16]

Of course, Balzac's Sarrasine is the classic readerly text that Barthes ingeniously shows us how to read as a writerly Text. In doing that, Barthes confirms: (1) that writerly reading does not eliminate readerly reading or its eligibility and contribution; (2) that reading of either sort presupposes a referent—the point of mentioning the "signifier," necessary for both readerly and writerly reading; (3) that the readerly text, which may have been derived, at some earlier time, from a writerly

Text, is now a fixed or bounded text the unity of which (in the modernist sense) has become a function of its particular interpretive history; alternatively, its (Sarrasine's ) being reread now (as a readerly text) commits us to recovering what constitutes it as a canonically fixed text; hence (4) that even a writerly Text is constituted (in our second sense) by interpreting something else—the "signifier," in Barthes's Saussurean usage; (5) that, for Barthes, the "Text," taken as the internal accusative of reading, is not an actual referent for further writerly reading (though the signifier is) but is collapsed into such a fixed referent only for readerly reading; and (6) that reading in the writerly way is not in the least incompatible with admitting readerly texts; in fact, it may be practiced on such texts.

A great deal of nonsense has been spread abroad maligning Barthes's intelligence, when what is wanted is a careful understanding of the remarkable thesis Barthes has bequeathed us. As it happens, it affords the best clue we are likely to find regarding the second sort of theory of interpretation. (I shall say nothing against Barthes's literary playfulness. Barthes was clearly impatient with or amused by the rigidities of structuralism. I see no evidence that his verbal games adversely affected his argument; and I see no evidence that Barthes ever played fast and loose with the constraints of coherence.) In any case, in his terribly freewheeling way, Barthes shows us how to entertain the idea that a text (in our sense, not quite in his, though congruently enough with his own notion) need not be presumed to have a fixed nature throughout a responsible reading (that is, in what we—once again, not Barthes—are calling interpretation) in spite of the fact that, however that nature may change, it remains a changing or changeable nature assignable (by reference) to this or that text (as we are prepared to say) or "signifier" (as Barthes would say). What shall we make of that?

Barthes does speak of interpretation but only either to dismiss it or to allude to what he calls "the Nietzschean sense of the word."^[17] What he means is that conventional interpretation (interpretation somewhat in my first sense) addresses readerly texts, texts construed as "products," referents with fixed natures; whereas writerly "texts" (texts treated in a "writerly' way) invite interpretation in the "Nietzschean' sense (in something close to my second sense), a sense applied to "production without product, structuration without structure": "To interpret a text [in this sense, says Barthes] is not to give it a (more or less justified, more or less free) meaning, but on the contrary to appreciate

what plural constitutes it. . .. This text is a galaxy of signifiers, not a structure of signifieds; it has no beginning; it is reversible; we gain access to it by several entrances, none of which can be authoritatively declared to be the main one; the codes it mobilizes extend as far as the eye can reach; they are interminable . . . their number is never closed, based as it is on the infinity of language."^[18]

Barthes is most exact here, despite the extravagance of his prose: "as nothing exists outside the text [he says: his remark follows Derrida's into print by about three years^[19] ], there is never a whole of the text [; that is,] for the plural text, there cannot be a [fixed] narrative structure, a grammar, or a logic."^[20] His meaning is plainly designed to preserve adequate resources for reference ("a galaxy of signifiers") but also to disallow a complete fixity of predicable nature ("a structure of signifieds"). He rejects the idea of exhausting the interpretive undertaking practiced in a cultural world that acquires new resources without limit, new "codes" of meaning drawn from its ongoing experience.

Interpretation in the "Nietzschean" sense subtends a responsive reading all right, but it is a reading that employs (as Barthes's own reading of S/Z shows) a selection from the "codes" of reading that gather and increase (somehow) in the life of our society—that do not and cannot lead to closure, to hierarchical preference, to mere correctness by way of reference to an antecedently closed textual nature. There is no longer an explication de texte, except by fiat: there is only a reading of the signifiers that thereby constitutes, reconstitutes, leaves indefinitely or "plurally" open to endless further reconstitution, the signifiers that acquire that interpretive history (that yield a text). This is the meaning of that otherwise impenetrable remark (playfully Rousseauesque): "narrative is both merchandise and the relation of the contract of which it is the object."^[21]

What Barthes means but will not say outright is that the interpretation of a "text" ("a galaxy of signifiers") is inseparable from a society's interpretation of its own open history. The meaning of whatever is singled out as a readable referent is, by that very act, judged apt for bearing whatever further signification its evolving "codes" dare ascribe to the evolving history of such ascriptions. That history is collected, referentially, as this "text" or that. The practice presupposes that texts may be individuated and reidentified as remembered histories (or careers), as referents that lack fixed natures. Barthes does not discuss the rigor of such a practice, it's true: he is more interested in its emancipatory power. But he never violates the limits of coherence, and we have (in S/Z ) the paradigm of the practice he recommends.

So seen, Sarrasine is not a standard story or a two-part story that contains a story within a story inviting explanation; it is a story "of a contract [says Barthes] of a force (the narrative) and the action of this force on the very contract controlling it": we are invited (in effect, by the writerly contract) to invent, by applying to certain signifiers the codes of reading of our world, whatever functional equivalences of structure may be imaginatively produced in an exchange of readings applied to the admitted structures of Sarrasine (the apparently "nested narratives").^[22]

It is helpful to notice that Barthes's way of reading—"writerly" reading—is entirely capable of its own characteristic discipline. It is indeed opposed to the exclusionary rights of "readerly" reading, though Barthes does not exclude such reading. The pertinent discipline depends on our sharing in some generous sense the consensual cultural resources of our own society. But that, as will gradually become clear, is also necessary for the preferred discipline of the "readerly' strategy. In short, the two ways of reading cannot be entirely disjoint. It is true, however, that Barthes is not interested in featuring any particular constraints—historical, traditional, intentional—on the interpretation of texts, in the manner favored, one way or another, by those who work within the hermeneutic tradition. There is that difference between the modernists and the poststructuralists.

It would be wrong to say that Barthes's sort of playful interpretation abandons reference or predication or a disciplined reading: it merely abandons the full fixity of texts favored in readerly readings in standard modernist accounts, and it shifts the focus of reference from finished text to enabling signifier. What Barthes alludes to is the neglect, among modernist theorists and literary commentators, of the changing context and history of reading—hence, of the contextually and historically changing "codes" of reading accessible to a living society. There is the clue to the infinite "plural" that is a Text, on Barthes's account. The very practice of reading—an entire society's practice, its Lebensform, we may say—ensures it. (I hasten to add that we must read "Lebensformen" or historical "traditions" in a much less conservative sense than Wittgenstein seems to have favored.) The "infinitude" of interpretations is simply the openendedness of a text's interpretability within a historical society's practice of reading. (This is also what Krauss attempts to assimilate.)

There are two pressure points in Barthes's theory of Texts pertinent to my second sense of interpretation. First, there is literally nothing to be interpreted (in the first sense of "interpret") until after a "text" is

"constituted" (Barthes's own term) by interpretive work (in the second sense of "interpret"); secondly, constituting a "Text" in Barthes's sense does not yield a "product," an ordinary text to be further interpreted (in the first sense of "interpret"). The point is regularly neglected by the would-be anarchists and irrationalists of interpretation. Remember: Barthes never disallows the disciplined option of readerly interpretation practiced on a textual product in accord with the first sense of "interpret."

Barthes's thesis holds only that the two sorts of reading arise together within the same societal practices (what he playfully identifies as its "codes" of reading) and may even be regarded as sequentially ordered phases of reading (or interpretation) within an increasingly normalized use of particular texts (or "galaxies of signifiers"). Barthes's emphasis is on the jouissance of (preferring) the writerly over the readerly, not the ineligibility of the latter.^[23] (The sexual joke is Barthes's, of course.) Liberty with texts or signifiers does not escape the normal constraints of discourse—only the presumptions of jejune literary theory. In a logical sense, reference has its rigor but is unavoidably informal; in a confirmatory sense, predication may be consensually apt or inapt but its possibilities evolve and are culturally openended.

Consider the following remark:

Reading a text cited by Stendhal (but not written by him) I find [says Barthes] Proust in the minute detail. The Bishop of Lescars refers to the niece of his vicar-general in a series of affected apostrophes (My little niece, my little friend, my lovely brunette, ah, delicious little morsel!) which remind me of the way the two post girls at the Grand Hotel at Balbec, Marie Geneste and Celeste Albaret, address the narrator (Oh, the little black-haired devil, oh, tricky little devil! Ah, youth! Ah, lovely skin!). Elsewhere, but in the same way, in Flaubert, it is the blossoming apple trees of Normandy which I read according to Proust . . . this does not mean that I am in any way a Proust "specialist": Proust is what comes to me, not what I summon up; not an "authority," simply a circular memory [that is, a memory that "circles" or stalks a text]. Which is what the inter-text is: the impossibility of living outside the infinite text—whether this text be Proust or the daily newspaper or the television screen: the book creates the meaning, the meaning creates life.^[24]

Reading in the writerly manner is a form of living, not a form of research; it involves know-how (savoir aller, not savoir, at least not in the sense of assuming the predicative fixity of objective texts). But it is disciplined. It involves a practice interesting to others (for instance, in S/Z ) only if the reader is really civilized, witty, inventive. We, then, re-

trace the play of S /Z in order to become similarly motivated and (perhaps as) expert. But reading in that way resists the (readerly) "bifurcation" of the reader/read text—in order to allow that same distinction to be made again in a freer (writerly) way. There is no "explaining" the Text (Barthes's "Text"), and there is no "knowledge" of the meaning of that Text: because, of course, there is (then) no definitive "text" and no one way of motivating readings "which would be definitive" of any meaning.^[25] Nevertheless, there are "galaxies of signifiers," socially habituated practices, disciplined options of reading, and above all the customary meanings of sentences and sedimented readerly texts. One sees at once Krauss's mistake—as well as the mistake of such postmodernists as Jean-François Lyotard.^[26] For, savoir faire and savoir lire do presuppose savoir—at least referentially. Writerly reading presupposes readerly reading—again, at least referentially.

Barthes effectively acknowledges the point: it is the only possible condition on which a complete chaos of reading (or of cultural life in general) can be avoided. It is in part at least what Wittgenstein means by "forms of life," what Bourdieu means by "habitus," what Marx means by "praxis" and "modes of production," what Hegel means by "Sitten," what Gadamer means by "wirkungsgeschichtliches Bewusstsein," what Husserl means by the plural of "Lebenswelt," what Foucault means by "epistemes." It is no more than the acknowledgment of the preformative historical practices by which culturally apt individuals first become apt. Their world is already culturally preformed for them: that is the reason they may be said to learn their native language and their native culture; that is the reason they can specify the "signifiers," the culturally (already) prepared materials, that, by interpreting (in the productive sense), they first constitute texts or artworks as such—what thereupon prove to be usable as referents apt for interpretation (in the adequational sense). In a word, they learn to share the consensual memory of an encultured society. If you press the point, you grasp in a single step the impossibility of modeling such interpretation computationally, for there is, and can be, no algorithmic accommodation of a genuinely emergent, historicized culture. If interpretation is open in anything like Barthes's sense, then, unless some very strong nativist account of concepts holds (Fodor's, for instance), which is most unlikely for the concepts favored in the interpretations of the arts, computationality must fail.^[27]

Barthes's emphasis, of course, is on the initial process of writerly reading. On the evidence (on his view), the process has been forgotten

or ignored or misconstrued. My own emphasis, for the moment at least, is focused rather on the option of continuing a critical discourse about whatever (a "text") is thereby so constituted—without in the least reneging on Barthes's fine lesson. On the argument, we preserve both themes merely by distinguishing with care the logical requirements of unicity or individuation from the prejudice of certain substantive (modernist) presumptions about unity or fixity of nature.

The constraints of reference and predication are not violated by Barthes, only displaced from produced or finished texts to interpretable (openendedly interpretable) signifiers—from "natures" to "careers," as I should say. Barthes himself does not tarry long enough to give us a theory of the social habituation of the practices of reading that support the distinctive discipline of readerly and writerly reading. He presupposes such a theory—or such theories—but he moves on to offer examples of what be recommends. For my part, we could easily pause to construct a theory of social practice—from Hegel or Marx, or Nietzsche or Foucault, or Weber or Lukács, or Husserl or Heidegger, or Adorno or Benjamin, or Lévi-Strauss or Althusser, or Wittgenstein or Bourdieu, or Gadamer or Kuhn.

The point remains quite constant, however: the waiving of texts in the sense suited to (a modernist view of) the first sort of interpretation does not eliminate referential discourse elsewhere (for readerly texts, say), does not preclude referential and predicative discipline within the writerly reading recommended (as in the identification of relevant signifiers, the identification of other readerly and writerly read texts, a certain civilized familiarity with the details or "codes" of one's culture); and it does not even preclude a rapprochement between readerly and writerly reading before and after the play of a particularly agile exercise of the latter sort (the charm of S/Z, say). Certainly, in yielding in an openended way at least minimally hospitable to speculations of Gadamer's and Foucault's sort, Barthes deliberately opposes any structuralist conception of interpretation (the point of the playful use of the term "codes"). He is, in effect, exploiting his own idiom of an earlier phase of work; and he is also accommodating (within an enlarged vision) all alternative practices of interpretation, so long as they do not entrench false presumptions of fixity or totalizing. (I shall have occasion to reconsider the matter when I examine Michael Riffaterre's late-structuralist model.)

In short, Barthes's preference of the writerly is not even a denial of the ontology of texts—or of the likely dawning of gradually normal-

ized texts for which such an ontology could be retrospectively constructed (if we wished); and it does not itself supply an adequate analysis of what a signifier is, or a practice of reading, or even a human being capable of reading in either the readerly or writerly way. It is one thing to grasp the fresh discovery Barthes bequeaths us; it is quite another to make a shambles of every effort to understand interpretation. After all, the "bifurcation" of the signifier and the would-be reader remains, after the provisional "bifurcation" of the readerly text and the reader (and author) is first disallowed—and then (of course) civilly permitted to be recovered again in Barthes's educated sense.

Barthes offers an instance (in S /Z )—after the fact of a readerly deposit of Sarrasine in the canon of conventional texts—of what it would be like before such a reading, to have read Sarrasine in the writerly way. The "galaxy of signifiers" lacks fixed meaning; but, as the competent readers we are, we possess the know-how for grasping what may be taken to be their meaning. Barthes suppresses this hermeneutic or habituative dimension of reading—but it is surely there. The semantic and semiotic potentialities of signifiers are already built into the minima of any socialized habit of reading and using language. Nevertheless, Saussure, whom Barthes had taken his original departure from (but now supersedes), had never successfully explained the "original" relationship between writing and speech or writing and thought that he insisted on; and without that "originary" source—or the effective replacement of it more perspicuously advanced by Wittgenstein and Gadamer, say—there remains a critical lacuna in Barthes's own account. Certainly, the familiarity of Wittgenstein's and Gadamer's views considerably domesticates the question of a "writerly" discipline. (Saussure's "failing," of course, is just what Derrida had so mercilessly exposed in Of Grarnmatology.^[28] )

But the deeper theme, missing also in Derrida, is this: that the deconstructive or poststructuralist or antimodernist rejection of the bifurcation of reader and text itself entails a competent practice or activity on the part of readers vis-à-vis something (signifiers, say) within a preformed or habituated cultural space in which (and by using the processes of which) what Barthes calls the "plural" or "infinite" Text is first constituted (and what we are now calling a "text" may be constituted for "infinite" or, better, indefinitely many interpretations). In a sense, "the deconstruction of hermeneutics" is therefore reversed and outflanked by being shown to require and presuppose a "hermeneutics of deconstruction."^[29] It is not, however, thereby disallowed

or repudiated. What the argument shows is that the rejection of a cultural world bifurcated between inquiring subjects and subjects inquired into—or between such subjects and what they do or produce (texts, in the idiom I have proposed)—is itself the work of subjects (competent selves) active in such a bifurcated world and affected by their own work in it.

We theorize in a critical moment about a preformative condition we cannot originally fathom (that Saussure thought he could fathom, that Husserl also thought he could fathom) within which the bifurcation of world and word (or text and reader) first arises. Barthes's splendid game of writerly reading tenderly texts serves a double purpose: for one thing, it affords a miniature exemplar of the impossibility of radically disjoining the double function of subjects as observers and observed (in much the same sense in which one cannot beat oneself at chess); and, for another, it subverts the fixities of privilege, of readerly reading, of the "metaphysics of presence," of all the bugaboos of failing to remember that the steady structures of our now-bifurcated world depend impenetrably on whatever we critically postulate as the preformed world within which our own salient "objective" world arises.

So seen, Barthes's invention is an attractive toy—no more than a toy, no more than a toy for Barthes himself: for we could easily (and would need to) interpose a conception of numbered, reidentifiable texts that could support interpretation in the first and second senses and that would, at the same time, subvert a "metaphysics of privilege" (the notion of fixed and bounded texts) just because—for reading purposes at least—texts do and must remain referentially accessible. Barthes's conceit of the infinite Text (that is not itself a referent) is, then, merely the deliberately posed extravagance of a disappearing limit for the more modestly interposed texts I am now recommending. Barthes nearly says as much:

The Text (if only because of its frequent "unreadability") decants the work from its consumption and gathers it up as play, task, production, and activity. This means that the Text requires an attempt to abolish (or at least to lessen) the distance between writing and reading, not by intensifying the reader's projection into the work, but by linking the two together in a single signifying process.^[30]

The point is, a theory of texts adequate for interpretation at the present time would favor Barthes's double lesson—but would do so in an ampler and more systematic way than Barthes actually does. We must:

(1) detach the full theory of the nature of texts (literary, visual, musical) from the mere referential and predicative constraints of discourse about them, so that all notions of fixity, essence, analogy with physical particulars are attenuated as far as possible or challenged as much as necessary; and (2) we must develop a positive theory of texts, of how texts (or culturally emergent phenomena and entities in general^[31] ) are actually constituted—first, from precultural physical materials and, second, from culturally prepared materials. Item (1) trades on the lesson drawn from Krauss and Barthes: namely, that unity and unicity are distinct though not altogether separable notions; item (2) requires an entirely fresh start and cannot fail to center on the peculiarities of Intentional properties and their incarnated relation to material properties.^[32]

All this may seem unnecessarily heavyhanded, because Barthes was so ponderously intent on being lightfingered. But there is no philosophical carelessness in what he says. On the contrary, what Barthes conveys by his practice is the logical viability of preserving reference and numerical identity and yielding on fixed natures. (He mistrusts theory, but he never violates the constraints of coherence.) I accept Barthes's innovation: I reconcile the two notions by treating texts as individuatable histories. (You will have noticed, by the way, two important terminological adjustments: first, that "interpretation" need not be confined to linguistic acts only—in the most literal-minded sense; second, that "texts" include any and all culturally constituted entities—intrinsically apt for interpretation. Thus, the ballet is interpretive, and persons are texts.)