Ambiguity is not the polysemy most words display as dictionary entries but results from the context's blocking of the reader's choice among competing meanings, as when, to use an example from Derrida, a French context hinders the reader from deciding whether plus de means "lack" or "excess" .1 In this case, the undecidability is due entirely to the fact that the reader is playing a score, the syntax, that will not let him choose. This must be because the score is (...) badly written; yet it is precisely this sort of willful neglect that critics have labeled poetic license, thereby underlining its literary nature. Undecidability has become a central feature in Derrida's analyses of literariness, and it is also the main underpinning of his creative writing.2 Better still, his own critical discourse has put undecidability to use, not a rare case of metalanguage imitating the very devices of the language it purports to analyze. My example are therefore drawn from Derrida on the assumption that his conscious practice of écriture, backed up by a sophisticated theory, will be particularly illuminating. For my own analysis of these phenomena, I shall be using a special word that Derrida has adopted and adapted from the terminology of ancient rhetoric. He proposes it in his commentary on this sentence of Mallarmé's: "La scène n'illustre que l'idée, pas une action effective, dans une hymen . . . entre le désir et l'accomplissement, la perpétration et son souvenir."3 Our critic points out that the grammar prevents the reader from choosing between hymen as "marriage," a symbolic union or fusion, and as "vaginal membrane," the barrier is broken through if desire is to reach what it desires. Undecidability is the effective mechanism of pantomime as an art form since from mimicry alone, without words, the spectator cannot tell whether a dreamed, or a remembered, or a present act is being set forth. This, in turn, Derrida shows to be fundamental to Mallarmé's concept of poetry. It is simply a pun or, as Derrida prefers to call it, a "syllepsis,"4 the trope that consists in understanding the same word in two different ways at the same time, one meaning being literal or primary, the other figurative.5 The second meaning is not just different from and incompatible with the first: it is tied to the first as its polar opposite or the way the reverse of a coin is bound to its obverse—the hymen as unbroken membrane is also metaphorical in both its meanings is irrelevant to its undecidability. What makes it undecidable is not that it is an image but that it embodies a structure, that is, the syllepsis. · 1. See Jacques Derrida's La Dissémination , p. 307.· 2. Because Derrida is a philosopher by trade, one would expect his undecidability to reflect the very precise logical and mathematical concepts of that discipline - which is to say, the limitations inherent in the axiomatic method. Kristeva, for example, tries to do this in her Le Texts du roman: Approche sèmiologique d'une structure discursive transformationnelle , pp. 76-78. So far as I can make out, however , Derrida's critical theory and reading practice do not pack more into the word "undecidable" than does the definition I offer in this paper.· 3. Stephane Mallarmé, Mimique, Oeuvres complètes , p. 310: "the scene [a drama, or rather a pantomime] bodies forth only an idea, not an action: it is like a hymen between desire and its realization, or between an act committed and the memory of it"; here and elsewhere, my translation unless otherwise cited. Derrida's commentary, "La Double Séance," has been rpt. in La Dissémination, pp. 199-317; see esp. pp. 240 ff.· 4. See La Dissémination, p. 249.· 5. This definition has prevailed ever since Dumarsais' treatise, Des Tropes . That syllepsis must be distinguished from the so-called grammatical syllepsis or the zeugma is apparent in Heinrich Lausberg's Handbuch der literarischen Rhetorik , pars. 702-7; on the acceptation chosen by Derrida, see pars. 7-8. Michael Riffaterre, Blanche W. Knopf Professor of French Literature and chairman of the department of French and Romance literatures at Columbia University, is the editor of Romantic Review. He is the author of Semiotics of Poetry, La Production du Texte, Typology of Intertextuality and A Grammar of Descriptive Poetry. "Syllepsis" developed out of seminars he led at the Irvine School of Criticism and Theory and at Johns Hopkins University. (shrink)

If we try to arrive at the simplest and most universally valid definition of the representation of reality in literature, we may dispense with grammatical features such as verisimilitude or with genres such as realism, since these are not universal categories. Their applicability depends on historical circumstances or authorial intent. The most economic and general definition, however, must at least include the following two features. First, any representation presupposes the existence of its object outside of the text and preexistent to (...) it. Readers feel, and critics pronounce, that the text’s significance depends on this objective exteriority, even though this significance may entail destroying the commonplace acceptance of the object; indeed, negating something still presupposes that something. Second, the reader’s response to the mimesis consists in a rationalization tending to verify and complete the mimesis and to expand on it sensory terms . The metalanguage of criticism accordingly prolongs and continues the text’s mimetic discourse, and critics evaluate representation in terms of its precision and suggestive power. Both processes—presupposition and rationalization alike—assume that referentiality is the basic semantic mechanism of the literary mimesis.There are, however, literary representations almost devoid of descriptive content, or so vague and so skimpy that their object cannot be analyzed or rationalized in sensory terms. Criticism is hard put to explain why readers feel compelled to evaluate them. And yet these texts not only lend themselves to interpretation but they are especially apt to trigger and control the reader’s hermeneutic behavior. In short, the represented object eschews referentiality yet refuses to vanish altogether, becoming instead the verbal vehicle of an interpretive activity that ends up by making the object subservient to the subject.1 1. See Roland Barthes et al., Littérature et réalité , esp. my paper, pp. 81-118, on the referential fallacy. Michael Riffaterre, University Professor at Columbia University, is the editor of Romantic Review. He is presently working on a book about Anthony Trollope . His previous contribution to Critical Inquiry, “Syllepsis,” appeared in the Summer 1980 issue. (shrink)

Literature is open to psychoanalysis as is any other form of expression—this much is obvious. Less so is the relevancy of analysis to the specificity of literary texts, to what differentiates them from other linguistic utterances; in short, the literariness of literature.The analyst cannot avoid this problem of focus. If he did, he would treat verbal art as a document for purposes other than an understanding of its defining difference. He would simply be seeking one more set of clues to (...) the workings of the human mind, as the sociologist or historian exploits literature to explore periods or societies through their reflection in its mirror.The only approach to the proper focus must be consistent both with the analyst’s method and with the natural reader’s practice. The analyst requires free association on the part of the analysand, and he matches this free flow of information with an attention equally open to all that is said. It is only after a passive stage of “evenly-hovering attention,” or, as the French nicely call it, écoute flottante, that he seizes upon clues to build a model of interpretation. These clues are revealed to him by anomalies such as parapraxes and repetitions or deviant representations, as well as formal coincidences between what he hears and the corpus of observations on linguistic behavior accumulated since Freud. The reader, on the other hand, is faced with a text that is strongly organized, overdetermined by aesthetic, generic, and teleological constraints, and in which whatever survives of free association is marshaled toward certain effects. The reader himself is far from passive, since he starts reacting to the text as soon as his own way of thinking, and of conceiving representation, is either confirmed or challenged. The text tends therefore not to be interpreted for what it is, but for what is selected from it by the reader’s individual reactions. A segmentation of the text into units of significance thus occurs, and it is the task of the critic to verify the validity of this process. In pursuing this goal he must restrict himself to a segmentation that can be proven as being dictated by textual features rather than by the reader’s idiosyncrasies, by those elements the perceptions of which does not depend on the latter and that resist erasure when they are in conflict with such individual quirks. The analyst’s advantage in identifying such features is that he is trained to recognize the above-mentioned anomalies and to explain them by repression and displacement, that is, by the traces left in the surface of the text by the conflict between its descriptive and narrative structures and the lexicon and grammar that we call the unconscious. Michael Riffaterre, University Professor at Columbia University and a senior fellow of the School of Criticism and Theory, is the editor of Romantic Review. His previous contributions to Critical Inquiry are “Syllepsis” and “Intertextual Representation: On Mimesis as Interpretive Discourse”. (shrink)