Source: Diacritics, Vol. 21, No. 4 (Winter, 1991), pp. 21-31 Published by: The Johns Hopkins University Press Stable URL: http://www.jstor.org/stable/465374 Accessed: 06/03/2010 01:06 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=jhup. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. The Johns Hopkins University Press is collaborating with JSTOR to digitize, preserve and extend access to Diacritics. http://www.jstor.org L'ECRITURE ET LE NON-DIT ANN BANFIELD Some twenty years ago, a linguistics formal and unyielding in confronting the evidence of a language that its analysis could distance from the individual speaker was a continuous point of reference for any discipline whose object could not be so rigorously defined. As such, it functioned above all for a literary theory that had itself already drawn many of its categories from an older grammatical tradition. Consequently, the central tenets of a contempo- rary critical theory, one whose foremost representatives have been French- Blanchot, Barthes, Foucault, and Deleuze-were grounded on a modem linguistic notion of language as a body of rules and principles held together by something other than a speaker's intentions. The radically new conception of language that emerged from a literary theory thus acknowledging modem linguistics in one of its three forms- comparative grammar, structuralism, and generative grammar1 -was then one in which the speaking subject ceased to occupy a place at the center of language and to provide the key to its structure. What has characterized modem linguistics in any of its three forms is the irrelevance of the speaking subject to the formulation of the rules predicting this structure, whether they are the laws of sound change, phonological derivations, or syntactic rules. None of the features of these rule systems can be deduced from any conception of the speaking subject, or of his function or intentions, nor are any of the structural properties of grammar plausibly attributable to properties of the speaker. This is not to say that the notions "speaker" and "subject" have no place in grammatical theory, but that it is they who are defined by the properties of language and not vice versa. So the notions of "the death of the author" and of a linguistic performance with a mode of existence independent of its producer (ecriture) arose naturally in a literary theory recognizing the language defined by linguistics. For it is modem linguistics that-as Jean-Claude Milner has put it, with an allusion to Freud's statement about the blow to man's narcissism dealt by psychoanalysis-has dislodged man from the center of his own language by demonstrating that no speaker creates his own language but instead confronts one that was already there before he came to speak it, setting limits on what 1. See Michel Foucault, The Order of Things, in particular, the section entitled "Labour, Life and Language," and Jean-Claude Milner, For the Love of Language, for two accounts of modern linguistics that restore to comparative grammar its important role. See also "What DoLinguists Want?," my introduction to Milner [28]. For Foucault, it was already "at the beginning of the nineteenth century" that "language was burying itself within its own density as an object and allowing itself to be traversed, through and through, by knowledge" [300]. Cf. also Ducrot: "la linguistique moderne, terme qui recouvre a lafois le comparatisme, le structuralism et la grammaire generative" [171]. It isperhaps only in Britain and the United States that comparative philology has fallen from memory. diacritics 21.4: 21-31 21 diacritics / winter 1991 ]i O 0 -3=: r r-; could be said [For the Love of Language 137]. These limits do not simply set restrictions on a speaker's or a writer's freedom; they rather define a non-ego-centered freedom, one not conceivable as a breaking down of boundaries or a violation of rules, but as the creation, in accordance with certain laws,2 of an area of choice. All of this has been reiterated in many different ways. But the consequences of this conception of language and literature have not provided the premises for a continuing research. On the contrary, the radical challenge to certain orthodoxies a non-speaker- centered conception of language contained has been imperceptibly revised, and the old has returned under the guise of the new. As the fitting accompaniment to this revisionism, another linguistics is currently appealed to, one Milner has labeled an "antilinguistics," which renounces the difficult task of providing a formal account of the often recalcitrant data of language and of submitting this account to the standards of logical economy. This is a linguistics that, unwilling to or incapable of looking directly at the impersonal laws of language itself and analyzing its forms and systems, turns instead to the "human voice divine" within linguistic performance, substituting communication for language. Such a substitute linguistics goes by various names: discourse analysis, pragmatics, speech act theory, communications theory.3 It provides the justification for the return of a unified authorial voice, in the guise of the speaker, to literature. That return has been disguised: it poses as a challenge to the egocentric theory of language, which it submerges in the notion of the "unicite" or oneness of the subject, and yet reintroduces this egocentricity under the form of a multiplication of speakers in certain configurations. These are the various theories of "dual voice," of "dialogism" and "heteroglossia," of "polyphony," enjoying currency at this moment. Their appeal explains in part the discovery and republication of Bakhtin; it likewise explains the following Oswald Ducrot has in France. In the face of the declaration of the death of the author in writing, these theories perform the function of a kind of animism, repeopling the text with the sound of personal voices. Moreover, the "carnivalesque," "heteroglossia," "polyphony"-all are proffered as figures of an imaginary freedom whose model, whose utopia, is a crowd, overflowing boundaries, breaking rules and mingling many voices, and whose ultimate weapon against power is an irony conceived either as the interpenetration of voices or as a preordained intentionality which intervenes deus ex voce to provide an ultimate harmony and meaning that are personal. What is Ducrot's theory of polyphony and how does it reintroduce a speaker-centered linguistics at the very moment it denies it? To answer this question, I will counterpose Ducrot's "sketch" of this theory to what he rejects as that theory's contrary-the most recent defense of the "dogme intouchable" [172] that "chaque enonce possede un et un seul auteur" [171], namely, the theory of the language of narrative fiction as a system of unspeakable sentences, in order to argue that Ducrot's very conception of the actual and 2. The question ofthefreedom possible within a language conceived as a system of rules might be pursued with respect to Chomsky's insistence on linguistic creativity and thefact that he is often misunderstood to mean by this the ability ofthegrammar togenerate an infinite number ofsentences from a finite number of rules and the speaker's ability to produce and interpret sentences he has never encountered before. It is the latter which defines the parameters of this creativity but does not explain or predict it. 3. See For theLoveof Language[138]. The qualificationofthesedisciplines as "antilinguistics" should not be taken to mean that the questions they treat are illegitimate or without interest but only that they fall outside the range of a formal linguistics. Just as the limits of a formal linguistics are not an argument against the validity of its claims, so the informal nature of such disciplines is in direct relation to the issues they address. What renders them "substitutes" is the claim of certain of their practitioners that the inability of formal linguistics to account for this or that aspect of language demonstrates the superiority of an nonformalpragmatics. It is in this way that those who take this position deprive themselves of the discoveries of a formal linguistics. 22 possible alternatives shows how the issues have been obscured. He assumes that the relevant dichotomy exists between the hypothesis that every utterance has one and only one speaker, "ancre dans la tradition linguistique" [172], and a "polyphonic" theory of the utterance that aims to "mettre en doute le postulat selon lequel un enonce fait entendre une seule voix" [172]. Ducrot's implicit reasoning assumes that the only challenge to the traditional egocentric view of language is one that counterposes a multitude of voices, assuming various functions of the speaker, to the "unicite" of a single speaking voice. But the issues can be posed otherwise and indeed have been. Another alternative exists, one that renders the polyphonic theory as only a variation of the existing one, because the crucial issue does not turn on whether an utterance is restricted to one or more than one "voice" but on whether every utterance must have a "voice." This question is central to the theory of narrative I have presented in Unspeakable Sentences. Yet it has gone so universally unremarked that the theory of unspeakable sentences is grounded on the claimed existence of speakerless sentences that it is difficult not to suspect that it is here the attempt, whether deliberate or unconscious, appears to "ward off the menace" to the traditional egocentric conception of language, to echo Ducrot [172]. But the empirical consequences of this claim and the challenge it poses to traditional assumptions cannot be grasped until the notions "author," "voice," "speaker," and "subject" have been rigorously defined in terms of the data. This neither Ducrot nor other defenders of dual voice do, contenting themselves with what Ducrot calls a "sketch." It is in this way that the hard and inalterable givens of language and the laws they yield can be avoided and more reassuring notions smuggled in. For what if there existed utterances unmarked by the person of even a single speaker? Then such instances of language would have lost the authority and intentional coherence attributable to that person called by grammatical tradition the "first." What would then be the consequence for the "unicite" of the speaking subject, for the possibility of a plurality of voices? To answer these questions we must turn to a more detailed account of the evidence of narrative as well as to the theory of unspeakable sentences. This theory requires several notions. Some are already a part of generative linguistic theory and hence not introduced solely to account for the particular data relevant to narrative style, and a restricted number of others are justified by the sole evidence of narrative. Their meaning is thus specific to the theory, even when, like subject or SELF, they also have a meaning in ordinary language, and it is the theory-specific meaning which must be understood when interpreting the theory and weighing its claims against any alternative. The first notion to be singled out is that of an E, or Expression, which rewrites Chomsky's S" (S = Sentence), with the further stipulation that this "highest S" cannot be embedded, that is, may not appear subordinated to any other E. I have provided justification for this revision of Chomsky's S" [see "Conditions on Transformations" (1973)] in "Narrative Style and the Grammar of Direct and Indirect Speech" and Unspeakable Sentences. Once introduced, further consequences follow from it, permit- ting a unified account of what can be referred to as linguistic subjectivity [see Unspeak- able Sentences]. It is thus for purely syntactic reasons that the two principles I formulated in these two works, and whose function Ducrot sees as warding off the threat of a"plurality of subjects" [ 172]-namely those summarized as "1E/1 SELF" and"Priority of Speaker"4- 4. 1 Ell SELF is a revision of Ell I or Speaker to accountfor the data of represented thought. The formulation of Unspeakable Sentences is the following: a. 1 E/1 SELF. For every node E, there is at most one referent, called the "subject of consciousness," or SELF, to whom all expressive elements are attributed. That is, all realizations of SELF in an E are coreferential. b. Priority of SPEAKER. If there is an I, I is coreferential with the SELF. In the absence of an I, a third-person pronoun may be interpreted as SELF [93]. diacritics / winter 1991 23 are principles whose domain is the E or entire sentence. (Moreover, it should be recalled that the notion Sentence in transformational grammar is an initial idealization of the notion "utterance." Such a grammar, accounting for "competence" and not "perfor- mance," is a grammar not of utterances, which belong to performance, but of Sentences, or in the terminology used in "Narrative Style," of Es.)5 Every E may have at most one subject of consciousness or SELF, a second notion the theory defines. The SELF is the point of reference for a set of subjective elements and constructions-deictics, nouns and adjectives of quality [see Milner, De la syntaxe d l'interprdtation], exclamations, and so forth-which share features of distribution and syntactic behavior. The generalization expressed in 1 E/1 SELF thus relates these two notions-E and SELF-in a precise way. Note, for instance, that the generalization applies to the unique reference of a speaker in an E only by virtue of the obligatory coreference of speaker and SELF, that is, by virtue of Priority of Speaker. Moreover, no comparable claim as to the uniqueness of the referent of the SELF is made about any unit larger than an E; in particular, the claim is not extended to the notion TEXT, which, in the special meaning attributed to it in Unspeakable Sentences, is the unit constituted by a series of Es related by the rules of anaphora and of "concordance of person and tense." On the contrary, it is explicitly stated in Unspeakable Sentences that the evidence supports no such extension of 1 E/1 SELF to the TEXT, held together, then, only by the rules of anaphora and Concordance of Person and Tense. Concordance of Person requires that all instances of the first person remain coreferential throughout all Es of a TEXT, but it makes no such requirement for the referents of SELF. As a result, the subject or SELF may change from E to E, giving rise to the famous "shift in point of view" of the moder novel, what Blanchot refers to in "The Narrative Voice" as the multiplication of little egos [136]. In the theory of unspeakable sentences, the novelistic text may consist of a plurality of different third-person subjects, each one unable to penetrate, so to speak, the E or sentence occupied by another, what Sartre once called "la solitude de chaque unit6 phrastique" [117]. The result is a kind of monadology of the sentence whose logical consequence is a plurality of subjective worlds. Insofar as an intuitive response to the novelistic form has been the acknowledgement of this plurality of perspectives, the theory of unspeakable sentences provides one model of this atomism of points of view, one in which the perspective, that is, the SELF, is a center to which linguistically subjective elements and constructions are referred and in which only a single such center is permitted in an E. But it is not a model that does not account for multiple points of view. To point out the possible plurality of perspectives which is deducible from a theory in which 1E/1 SELF and Priority of Speaker (or I) apply to Es and not to TEXTS is not, however, to say that the theory of unspeakable sentences is a notational variant of a "polyphonic theory of the utterance." Both Ducrot's "polyphonic theory" and that of unspeakable sentences confront, in part, the same set of facts, which may be described- an initial description depending on a terminology that further analysis may reveal to be misleading-as texts containing a plurality of perspectives, in some sense yet to be defined. But the relegation of each single SELF (as one theory's concept of a perspective, counterposed to the other's notion of "voice") to the confines of one or more E and the Ducrot does not seem to be familiar either with the fuller presentation of these principles in Unspeakable Sentences or with the criticism of the dual voice theory in that context or in "The Formal Coherence ofRepresented Speech and Thought," but only with my 1973 article [ "Narrative Style"]. 5. Ducrot also insists on "une distinction rigoureuse entre l'enonce" et 'la phrase"' [174]. But if the distinction is appropriately invoked at the start, he unfortunately does not establish an equally rigorous system relating these notions. This undermines his attack on 1 Ell SELF, since hecanpresent claims about "I' enonce" without specifying inwhat way this constitutes an argument against a theory of Es, which have the theoretical status of sentences. 24 impossibility of more than one SELF appearing in a single E yield a quite different picture of the novelistic text than the one in the position Ducrot outlines. (It should be pointed out here that the precision required and made possible by linguistic methodology and the formally representable evidence of language allow one to go beyond variously interpret- able nonformal accounts, such as that of Bakhtin. It may not be entirely possible to decide whether Bakhtin's position is closer to that of Ducrot or that of Unspeakable Sentences- at times his phrasing seems to permit the possibility of two points of view in a single sentence, and at other times it does not-because Bakhtin never posed the question in these terms.)6 What is decisive is that the possibility of shifts in perspective arises in the theory of unspeakable sentences under quite specific circumstances. For Concordance of Person, the principle defining the larger unit of the TEXT, among other things, as a sequence of Es in which all referents of the first person remain coreferential, together with Priority of Speaker, a principle affecting Es, means that a TEXT may shift perspectives only when it has no first person. The crucial condition for the appearance of a style displaying a plurality of perspectives thus emerges as the possibility of sentences (technically, of Es) with no speaker, that is, sentences in which the first person is excluded, in which that central authority of the personal voice of the speaker, interpreting, evaluating, expressing and bestowing coherence and unity on a discourse,7 is absent. In the linguistic analysis of that narrative style known in French as style indirect libre and which I call "represented speech and thought" can be found the syntactic arguments that Es with a third-person SELF cannot also contain a first person, or otherwise they could not receive the interpretation in which they represent a third-person point of view. To account for such sentences, a subjective perspective must be isolated independent of the first person and thus not be assumed by the authoritative speaking voice of any narrator. In this way, a novel may "represent" many such subjectivities, each one equal from a linguistic point of view, without giving one point of view priority over the other (as would be the case if a speaker's voice were allowed to occupy the same E as a third-person perspective), and "unity" and "meaning" come to reside elsewhere than in the continuous, personal voice of a narrator commenting on and guiding the reading. Each single perspective remains uninterpreted, however, at any higher level in the language of the text. 6. See Reboulfor a discussion of Bakhtin with respect to Ducrot's polypony and the theory of unspeakable sentences. For another discussion ofBakhtin and Ducrot, see Moeschler. Ducrot' s characterization of the theory of unspeakable sentences and the logical relations holding among its various claims is, at any rate, inaccurate: arrivee... au moment ou une pluralit6 de sujets pourraient etre introduits dans 1'enonce, Banfield formule deux principes qui ecartent la menace. Elle pose d'abord qu'il ne peut y avoir, pour un enonce donne, qu'un seul sujet de conscience, repoussant d'emblee dans le domaine de l'anormal les exemples qui feraient apparaitre une pluralit6 de points de vue juxtaposes ou imbriques. Et ensuite, afin de traiter les cas ou le sujet de conscience n'est pas l'auteur empirique de l'enonce, elle pose qu'il n'y a pas, dans ces enonces, de locuteur. Certes, je ne reprocherai pas a Banfield-bien au contraire-de distinguer le locuteur, c'est a dire l'etre designe dans l'enonce comme son auteur (au moyen, par exemple, de marques de la premiere personne), et le producteur empirique, etre qui n'a pas a etre pris en compte par une description linguistique pr6occupee seulement des indications semantiques contenues dans l'enonce. Ce queje reprocherai a Banfield, c'est la motivation qui 1'amene a cette distinction, a savoir le souci de maintenir coute que cofte l'unicit6 du sujet parlant. [172] 7. Here I am using "discourse" in its ordinary sense. In Unspeakable Sentences the term is reserved for that kind of sequence of Es, that is, that kind of TEXT-a unit that is not necessarily written-which is marked linguistically by the presence of a speaker and an addresseelhearer. diacritics / winter 1991 25 It is in a quite different fashion that a multiplicity of points of view are accounted for in Ducrot. Indeed,"polyphony" for Ducrot does not refer to an unlimited number (in terms of any limits imposed by the grammar) of equal points of view-equal because each has as its domain an equivalent grammatical unit, the E, in which each plays an equivalent role. For Ducrot every sentence is divided between several different functions, in a way reminiscent of the division between speaker and self in the theory of unspeakable sentences [see "The Formal Coherence of Represented Speech and Thought"]. But the similarity only goes so far. For the division between speaker and SELF does not endow the speaker with the independent status of a full speaking subject grammatically incarnated in the first person and separate from the third-person SELF, whose perspective is subordinated to it. The speaker may appear only when it is coreferential with the unique SELF. As long as a third person subjectivity is represented, no speaking voice can be realized. In Ducrot's account, on the contrary, as well as in other versions of a "dual voice theory," the text in which a third-person subjectivity is represented retains a higher authority that places this represented subjectivity within a system of values-those "spoken" by the "narrating voice"-higher than itself. I will not enter into a detailed refutation of Ducrot's position, restricting myself to the more important task of correcting certain recurrent misconceptions about the differences between the theory of unspeakable sentences and what seems to me yet another version of the position I have dealt with elsewhere under the name of "dual voice theory." There I have answered the objections of proponents of this theory in detail.8 Moreover, for these objections to carry, the citing of a few purported counterexamples does not suffice, as any version of scientific methodology will point out. A countertheory is required, one accounting for all the data the first theory accounts for, as well as for the counterexamples. Until then, the burden of proof is on any counterproposal. Ducrot, however, more than anyone, cavalierly dismisses the detailed edifice of rules, arguments, and data-"je signalerai rapidement une recherche americaine" [172]-he wishes to replace with a hastily constructed "sketch." His methodology does not rely on logical argumentation or the marshaling of a rich body of evidence. Rather, each example is just another in a long list of examples of different, supposedly subjective aspects of the text, each different example calling for a new theoretical construct, invented ad hoc. Indeed, the distinctions Ducrot makes seem ultimately to be undone by a methodology sorely in need of that tool of logical economy, Occam's razor. A finite number of theoretical constructs and principles are not the means for Ducrot of accounting for a great variety of different donndes, which in turn justify these constructs and rules; instead, a number of isolated examples or cases-quite different notions than those of data and evidence-are the excuse for the proliferation of loosely defined notions. Where is the Ducrot of the elegantly argued "Peu et un peu"? The Ducrot of the "sketch of polyphony" is not really elaborating a systematic theory argued on the basis of linguistic evidence but is making a loosely classificatory inventory of examples supposedly inserted with difficulty in a theory of unspeakable sentences. The issue of method and argumentation is not unconnected to that of the unquestioned assumptions and misconceptions of Ducrot's attack on a theory of language centered on a single subject. For his whole countertheory of polyphony depends on a series of distinctions-between "l'nonciateur," "le locuteur," and "l'auteur empirique," for example-that, if he is to make good the claim to have replaced a speaker-centered linguistics, he must demonstrate are empirically distinct, that is, that they account for quite different and clearly defined linguistic phenomena in a consistent and systematic way, as 8. Indeed, Ducrot's critique of "Narrative Style" seems to rely on two secondary sources; it is not upon his own careful analysis of the theory of unspeakable sentences that he bases his objections, but upon these two essays [see Ducrot 173]. 26 well as theoretically distinct from the single, unitary subject. But the gestures he makes toward a theory in which the authority of a dominant speaking voice is challenged never transcend the idea of a hierarchy of speaking subjects, reduplicated at different levels. In insisting, for instance, on the distinction between "le locuteur, c'est-a-dire l'etre designe dans l'6nonce comme son auteur (au moyen, par exemple, de marques de la premiere personne), et le producteur empirique" [172]-an important distinction-he cannot conceive of the latter, the "empirical producer of the text," except on the model of another speaker. That the author's role in writing, in composing, need not, cannot, receive a linguistic representation unified around the notion of a voice does not, however, seem to have occurred to Ducrot, and he can conceive of him only as in some way "speaking" in the text, even if he never says "I," unless it is to represent a fictional persona: The author sets before us characters who in what I call,following Anne Reboul, a 'first speech" exert a linguistic and extralinguistic action, an action the author himself does not take responsibilityfor. But the author may, in a "second speech," address the public through the characters, either by assimilating himself to such and such a one who he seems to make his representative ... or by thefact that it seems significant that the characters speak and behave in such and such afashion. [205; ed. trans.] When discussing Benveniste's notion of histoire, Ducrot seems to allow an "enonce" which "n'exhibe aucun auteur de la parole." But the consequences of this possibility are never pursued, nor is it explicitly integrated into the larger "polyphonic" system. Thus, the author, at one level seemingly silenced and banished from the text by receiving no linguistic representation, nevertheless returns to address an audience via a "secondary speech." For this is ultimately a metaphoric way of describing certain phenomena for which it is the task of linguistics and a literary theory aided by linguistics to find a more accurate account. This includes an analysis of what it might mean to represent a subjectivity other than by expressing it as a speaker would to a listener. The notion SELF is meant to capture such an "unspoken" notion of subjectivity. What follows from this unspeakability is that subjectivity is unmediated by another interpreting voice. The inability to conceive of the subject as other than a speaking voice prevents Ducrot from grasping the notion of SELF, that is, of a subject which is not also a speaker, crucial to the statement of 1 E/1 SELF: It is this "1 Ell SELF" theory that permits the use of the expression "subject" while presupposing as self-evident that there is a being who is the unique author of the statement [enonce] and responsible for what is said in the statement. If one has no scruples or reticence in using this expression, then, it is because one never even dreams of calling into question the uniqueness of the statement's origins. [189; ed. trans.] But the subject is not "responsible for what is said" in the Expression, according to 1 E/ 1 SELF; he is the point of reference for a specific set of linguistically subjective elements in the E, which is thereby taken to represent his perspective. Yet nothing is "said" in such a sentence; an author who is not directly embodied in a first person, as a speaker is in his speech, may manipulate language in accordance with the possibilities inherent in it-and at the same time respecting the limits it imposes, in order to represent a fictional subjectivity-but he does not speak in it. He writes,rather, and in writing disappears. This inability of Ducrot's to get beyond the concept of a speaking voice is tied to the inability to see revealed in the evidence of language-in particular, the evidence of narrative-a vision of language other than the egocentric one, a vision in which a territory is opened diacritics / winter 1991 27 up within man no longer governed by man, in which forms and constructions appear emptied of all human presence. Contrary to Ducrot's explicit claim to have destroyed the oneness of the speaking subject, it is preserved at every level. What seems to be the unstated conceptualization behind this claim is one in which a certain unified subject- namely the author-is divided into a number of separate subjects, each unified and entire, but each nevertheless hierarchically subordinated to the author in some undefined semantic fashion. In this manner, phenomena such as irony are brought back to and contained within the familiar form of a voice, whereas no such "explanation" is available to the theory of unspeakable sentences. Indeed, in this latter theory irony receives no explanation. But far from being a deficiency of this theory, this lack of explanation reflects the fact that irony is not a linguistic phenomenon, in the strict sense of that which is formally representable within linguistic theory, and can be represented in it only in an imaginary and nonfalsifiable way.9 For this reason, a formal linguistic theory, unlike Ducrot's, concedes that there are aspects of linguistic performance that escape explana- tion within its framework and acknowledges the existence of another discipline with a long and rich history analyzing those aspects of literature about which linguistics has nothing to say-literary criticism. The alternatives, therefore, counterpose not a theory centered on a single unitary subject and a polyphonic theory but rather one in which a plurality of isolated and noncommunicating points of view or centers coexist in a narrative style in which there is no first-person, single omniscient voice, imposing a personal unity, and one in which polyphony consists in a hierarchy of voices, each conceived on the model of the other, yet one providing a single, overarching center. Ultimately, of course, the choice between the two competing models will depend on argumentation appealing to the evidence. But currently the issues have taken on an ideological coloring, and it suffices to name one position "unitary" and the other "polyphonic" for it to appear that the sides are chosen. I have tried to suggest, however, that either side might choose to call itself "polyphonic," so long as this term remains evocative rather than clearly defined. I have also tried to show that the position roughly sketched by Ducrot is far from being the one presenting the greatest challenge to the traditional egocentric view of language, far from being the counterpart in linguistics of a contemporary literary theory. Instead, it is just one of the current ways in which the return to the traditional formulation is negotiated. On the other hand, a theory of the language of narrative in which a configuration of sentences possessing no subjective center whatsoever (sentences of Narration per se) alternating with sentences possessing different third-person centers of subjectivity but in which there is no single center of subjectivity holding together the whole text is by no means a unique and isolated one. I have argued elsewhere [see "Ecriture,Narration and the Grammar of French"] that it is such a view of the novel taken as the paradigmatic case of literature that emerges in the fifties and sixties in France and that lays the empirical groundwork for the new theory of literature that came to mark those decades. Its central tenet, as we have seen, took various forms: the death of the author or the notion of an 6criture which was not the personal voice of a writer conceived of as a speaker. Its principles were to be sought in the language of the novel. Each in a strikingly similar fashion, Barthes, Blanchot, Butor, Foucault, and Deleuze saw a radical division between 9. In other words, the formula for irony is "saying one thing and meaning another." But, of course, that "meaning" is other than the semantic meaning assigned to the sentences by the grammar. It is for this reason that ironic intent is difficult to prove in libel suits and also for this reason that irony can be a tool ofpolitical opposition when the consequences of direct opposition seem too severe. For a different position on the possibility of aformal account of irony, see Sperber and Wilson, Relevance, esp. 237ff. WhileSperber and Wilson treat ironyformally, they nonetheless place it outside formal linguistics. 28 two possibilities of novelistic style that set the language of narrative off from that of the epic, that is, a narrative tied to the spoken language. Blanchot spoke of it as a division of the il into an il/it and an il/(s)he. The first becomes "the impersonal coherence of a story" [135], "the indifferent-difference that alters the personal voice" and "does not speak from a centre" [142]; the second "marks the intrusion of character: the novelist is the person who refuses to say 'I' but delegates that power to other people; the novel is filled with little 'egos'" [135-36]. Blanchot's distinction corresponds to the division of labor Barthes, in the chapter "Writing and the Novel" in Writing Degree Zero, sees performed by the narrative past (in French the pass6 simple) and what he calls the "third person of the novel," that is, represented thought. It is clear that for both Blanchot and Barthes these divisions are ultimately grammatical, based upon differences of person and tense. In the development of the novel out of the epic, what are essentially two new forms of narrative sentence are accompanied by the gradual silencing of the storyteller and the development of a form in which that "experience one does not recount but that is involved when one recounts" creates a distance that "decentres the work" [Blanchot 139]. The condition then of "polyphony" in this view-of a plurality of subjectively centered worlds, to use a terminology that avoids suggesting a speaker-is the silencing of the central person of an author as narrator, who, if he continues to exist, exists somewhere outside the language of the text, "in another dark," as Beckett puts it in Company. The result is the creation of linguistic units-nonembeddable sentences, or Es-that owe their structure to some otherprinciple, some other center than the speaker and the time and place of his speech act. In the case of sentences of Narration, those appearing in French in thepass6 simple,10 there is no subjective temporal and spatial center whatsoever. Such sentences recount the past by placing discrete entities that are past events in a linear order. But there is no privileged moment to which the others are referred; the order of time is that of history, outside of any experience. One recognizes in this account the conception of time that Bergson, in L'essai sur les donn6es immn diates de la conscience, defines by contrast to la dur6e; it is "le temps que l'astronome introduit dans ses formules, le temps que nos horloges divisent en parcelles egales" [80]. This explains why the French narrative past, as Benveniste has pointed out, tolerates no deictics. In the case of sentences of represented thought, the result is even stranger and more counterintuitive. Such sentences are organized, so to speak, around subjectivecenters that are not egos but third-person centers of subjectivity, each representing spatially and temporally a here and a now which does not designate any speech act. Stranger still, there is evidence that versions of such sentences exist that are not occupied by any third-person SELF, but nonetheless contain spatial and temporal centers. They present "unoccupied perspectives," linguistic representations of a theoretical construct belonging to Bertrand Russell's theory of knowledge, of "possible worlds," in a Russellian reading of Leibniz. Such sentences describe the sensibilia of the physical world as events grouped around an empty center." It might be said that this model "divides" or factors out the subject into various linguistically defined aspects of subjectivity, some of which may occur in isolation-the here/now of the unoccupied perspective-and others obligatorily in conjunction with this spatial-temporal center.12 There is no a priori reason for deciding that this account of linguistic subjectivity is less a "threat" to the oneness of the subject than another, since the unity of the subject is a notion that itself needs definition within a precise and empirically testable theory. 10. Such sentences correspond in crucial ways to Benveniste's notion ofhistoire. 11. The notion of an "unoccupied perspective" as a linguistic concept is the subject of my "Describing the Unobserved: Events Grouped around an Empty Center." 12. See, in particular, Unspeakable Sentences, chapter 5, for a discussion of reflective and nonreflective consciousness. diacritics / winter 1991 29 The world delineated in this theory is one created by a written language associated historically with realism, thatrealism which, carried to its logical conclusions, became the modernism of Joyce and Woolf, Proust and Beckett. It is a world-the one that sensation yields-that is "fragmentary, absurd, lawless, but not self-contradictory" [Russell 199].13 That world presents another figure of freedom than the Bakhtinian one. This is the freedom of a writing in which the writer, poised between two determinisms-a personal style, an individual and blind necessity, on the one hand, and a rule-governed language on the other-discovers an arena of choice where he can create, liberated from the orthodoxies of the past only at that point where he relinquishes personal expression and ceases to inhabit and use his language as a speaker. Then, at a certain remove from his style and his language-"that distance that distances even him, removing him from the centre, since it constantly decentres the work" [Blanchot 139]-he discovers in it and with a certain terror laws and truths whose mode of existence is impersonal and which it is no longer appropriate to qualify with the possessive pronoun. That discovery is of a possibility inherent in language, but it is the written language of narrative that reveals it. "A language and a style are blind forces; a mode of writing is an act of historical solidarity" [Barthesl4]. We recognize here Barthes's ecriture-that existential freedom Barthes claimed, in the face of Sartre's action and personal engagement, for an engaged impersonal writing rather than for an engaged writer. "Thus the choice of, and afterwards the responsibility for, a mode of writing point to the presence of Freedom" [16]. It is a freedom that must engage in an untiring struggle against that "doxa" which it itself creates. "Writing as Freedom is therefore a mere moment. But this moment is one of the most explicit in History, since History is always and above all a choice and the limits of this choice" [17]. Nor does this conception of freedom reject a priori the figure of the carnivalesque crowd of voices disrespectful of laws and limits. It rather recognizes that the nostalgia for such a freedom is a longing for something imaginary, for a freedom seized once and for all, a freedom in which no choices are required because no real consequences would follow from them. The freedom of writing is haunted by a past far from prelapsarian-that past of its own creation, its own history, which continually returns to remind it of the unintended consequences of its own former freedom, to deprive its present of the old possibilities. Such aconception of language and literature offers no imaginary consolations. In this period desperate for consolation, a period that identifies itself with no more decisive label than a "post-" which only rejects its own past, the hard vision of literary theory that was inaugurated in the years of the cold war and that lived through 1968 is pushed back into a past now pronounced irrelevant. But this does not change the fact that all the possibilities inherent in it have not yet been exhausted, despite their having fallen out of fashion. WORKS CITED Banfield, Ann. "Describing the Unobserved: Events Grouped around an Empty Center." The Linguistics of Writing. Ed. Nigel Fabb, Derek Attridge, Alan Durant, and Colin MacCabe. Manchester: Manchester UP, 1987. 265-85. ". "Ecriture, Narration and the Grammar of French." Narrative: From Malory to Motion Pictures. Stratford-Upon-Avon-Studies, 2nd series. Ed. Jeremy Hawthorn. London: Edward Arnold, 1985. 1-24. . "The Formal Coherence of Represented Speech and Thought." Poetics and Theory of Literature 3 (1978): 289-314. 13. The exact sense in which thisworld is "lawless" would require afuller account ofRussell's theory of "our knowledge of the external world." For that, see note 11. 30 . "Narrative Style and the Grammar of Direct and Indirect Speech." Foundations of Language 10 (1973): 1-39. . Unspeakable Sentences. London: Routledge, 1982. Barthes, Roland. Writing Degree Zero. New York: Hill & Wang, 1967. Beckett, Samuel. Company. New York: Grove, 1980. Bergson, Henri. L' essai sur les donn6es immediates de la conscience. Oeuvres. Paris: PUF, 1959. 1-156. Blanchot, Maurice. "The Narrative Voice (the 'he', the neuter)." The Gaze of Orpheus and Other Literary Essays. Barrytown, NY: Station Hill, 1981. 13343. Butor, Michel. "L'usage des pronoms personnels dans le roman." L'entretien infini. Paris: Gallimard, 1969. Chomsky, Noam. "Conditions on Transformations." A Festschriftfor Morris Halle. Ed. S. Anderson and P. Kiparsky. New York: Holt, Rinehart & Winston, 1973. 232- 86. Ducrot, Oswald. Le dire et le dit. Paris: Minuit, 1984. Foucault, Michel. The Order of Things: An Archaeology of the Human Sciences. New York: Pantheon, 1970. Milner, Jean-Claude. For the Love of Language. Trans. Ann Banfield. London: Macmillan, 1990. De la syntaxe d l'interpretation: Quantite', insultes, exclamations. Paris: Seuil, 1978. Moeschler, Jacques. "Dialogisme et dialogue: Pragmatique de l'enonce vs pragmatique du discours." Actes du colloque dialogisme etpolyphonie. Universite de Neuchatel- Suisse, Institut de Linguistique, 27-28 Sept. 1985. 7-3. Reboul, Anne. "Dialogisme, style indirect libre et fiction." Actes du colloque dialogisme etpolyphonie. Universite de Neuchatel-S uisse, Institut de Linguistique, 27-28 Sept. 1985. 45-81. Russell, Bertrand. The Analysis of Matter. New York: Dorer, 1954. Sartre, Jean-Paul. "Explication de L'Etranger." Situations I. Paris: Gallimard, 1947. 99-121. Sperber, Dan, and Deirdre Wilson. Relevance: Communication and Cognition. Cam- bridge: Harvard UP, 1986. diacritics / winter 1991 31