Factives and Perspectives: Making Sense of Presupposition as Exemplary Inference

Inference has long been the concern of assorted disciplines that vary in focus, rationale, apparatus, terminology, and achievement. There would appear an inverse proportion between the range and the orderliness of the knowledge accumulated by the various disciplines. Literary study tends to the first extreme, logic and its heritage in modern pragmatics to the second, either imbalance precluding a viable account of the field or even a comprehensive research program. To have the best of the two worlds, I argue, discourse inference (hence its study) at large must be reconceived as an integrative activity: out of the wide repertoire of sense-making resources available to humans, the inferrer opts for the mechanism(s) that will best integrate the given text in or with the context. Among inference types, presupposition figures here as exemplary because it is the most encoded and determinate (vis-à-vis, say, implicature) yet also themost controversial and themost resistant on inspection to the idea of Poetics Today : (Spring ) Copyright ©  by the Porter Institute for Poetics and Semiotics. 130 Poetics Today 22:1 formal systematizing, whether the logico-semantic or the so-called ‘‘pragmatic’’ way. After fifty years of formalist sound and fury, therefore, resystematizing presupposition by appeal to the inferrer’s quest for integration should also adjudicate between the respective analytic paradigms in general. Among presuppositional triggers, in turn, factivity (e.g., the verb ‘‘know’’) recommends itself for paradigmatic analysis by its centrality within this inference type, as well as within language, and by its intersection with various other disciplines, from epistemology to narrative theory. The argument starts by exposing the ‘‘dead end fallacy’’ that vitiates traditional approaches: they would rule out a priori, as ‘‘unacceptable/unreadable,’’ instances of factivity that apparently contradict themselves. Actually, such clashes abound throughout discourse, and multiple resources for integrating them emerge. Far from disabling or defeating inference, contradiction and lesser incongruities rather powerfully activate and channel it. Of the available integrative resources, again, the perspectival mechanism turns out definitionally attached to factivity as a branch of nondirect (e.g., ‘‘knew that . . .’’) quotation. Like all discourse about discourse, it necessarily joins together the viewpoints of quoter and quotee (e.g., attributor cum presupposer and subject of ‘‘knowledge’’) for us to unpack. So all disharmonies there may cohere via our distribution of the givens (epistemic attitudes included) between the partners within some quoting mold. In extreme, traditionally ‘‘dead end’’ cases, the inconsistent-looking factive presupposition just gets shifted to a more oblique, free indirect quotee—with appropriate changes in the quoter’s epistemic bond, narrative setup, and communicative (e.g., ironic) design. By way of ultimate test, the analysis then extends to negated factivity and to nonfactive triggers. Presupposition is accordingly redefined as an uncancelable yet shiftable inference type. By a further extension, it typifies the shuttle among language, world, and perspective whereby we make sense of discourse, always with an eye to the best fit. As literary discourse enacts this universal quest at its most artful, pragmatics would be wise to abandon the strange gods inherited from logic and align with poetics. 1. Presupposition among Inference Types: After Fifty Years Inference, as the process or product of reasoning from the manifest to the latent, deriving conclusions from givens, is essential to human thought, communication, inquiry, even existence.Nowondermanydisciplines, some ancient, some recent, have converged on it, under many names: exegesis, hermeneutics, syllogizing, allegoresis, psychoanalysis, interpretation, reading, decoding, deducing, disambiguating, processing, hypothesizing, entailment, implication, connotation, allusion, association, Gestalt, motivation, integration, sense-making, gap-filling, closure, etc., as well as inference proper.The babel is a telling sign and result of the ongoing disciplinary isolationism in the pursuit of common knowledge. Nor have the achievements, down to the goals envisaged and the measures of success, been anything like equal. On the record, there would apSternberg • Factives and Perspectives 131 pear an inverse proportion between the range and the orderliness of the knowledge accumulated: between particulars and generalities, wealth and explicitness, volume and systematicity. These extremes, more regrettably yet, would appear to correlate with the orientation toward and away from discourse as the arena of inference. Going by sheer quantity, onemight expect literary study to lead the field. It has by far wondered, quarreled, and learnedmost about the derivation of the unsaid from the said in a variety of genres, as about discourse at large. Beyond particulars, however, the derivers’ collected findings and insights have rarely grown collective, much less cumulative over the millennia. No textbook makes the fruits of studying myriad texts available to the literary student—or to adjacent disciplines so badly, if unknowingly, in need of them as correctives and correlatives, eye-openers and labor-savers and reminders of questions already broached. For a sorting out cum updating of all the currently known hermeneutic resources, we have to look as far back as the ancient Rabbis’ seven, then thirteen and thirty-two rules (middot) of (Scriptural) interpretation, quite possibly methodized with a glance at the Greek competition. In this regard, the lust for meaning displayed throughout the twentieth century hardly marks a turning point or a fresh start. It has established ‘‘interpretation’’ as the newest (to many, the central) branch of literary study with a vengeance: the emphasis does not usually fall on the ways, let alone the obliquities, of meaning across texts, but on the close reading of a single text as an end in itself. Today, the drives toward liberating the reader from (con)textual control would even valorize the distance between reading and reasoning. So libraries have been mainly filled with innumerable, invaluable, indiscriminate data that cry out for theorizing. As a believer in the twoway traffic between the practice and the theory, I would be the last to belittle either the value, the record, the joys, and the uses of literary interpretation or the sporadic attempts since theNewCriticism to understand,methodize, even channel it. (Both, traffic included, will figure below.) The paradoxical fact remains that the discipline that knows most also knows least compared with its enormous latent knowledge, because it has normally worked with rather than on inference. At the opposite extreme, all too neat and formalized, stands philosophy, especially the logical and linguistic or ‘‘ordinary language’’ branches. This alluring show of rigor helps to explain its following in other disciplines: not just in linguistics and semantics but also, much less happily, in the newer pragmatics, which traces back to it, and even among poeticians avid for scientistic legitimacy or equipment. Since theGreeks, formal logic has operated on an extremely narrow front 132 Poetics Today 22:1 of inference—that proper to its exclusive concern with ‘‘valid’’ as against ‘‘invalid’’ reasoning or deduction. For such a binary cut, only the sharpest measures of followability between given and inference will serve: the one must so necessarily imply (‘‘entail’’) the other that it would be nothing less than self-contradictory to assert the premise(s) while denying the conclusion. In the truth-conditional parlance now widely adopted outside logic, proposition X entails proposition Y if and only if in all worlds where X is true, Y is true.1 Hence the valid argument is consistent and true in reason, the invalid inconsequent and false. Arguments being diverse, however, entailment branches out into subforms or rules. Most familiar from those of the three-step syllogism, these rules also govern the deduction of ‘‘Julia owns three houses’’ from ‘‘Julia owns four houses,’’ say, or indeed the selfdeducibility of either. On the narrow front of entailment, as such examples show, the deduced inference is liable to combine the strength of the truth-conditional with the triviality of low informational value (and not just by poetic standards). By the same token, the inferencing is thoroughly rarefied, as context-free as it is indifferent to subject matter. The corresponding gain—unhappily, the envy and model of certain extralogical domains—is the amenability of this inference type to ‘‘formalization’’: into a coherent set of rules, each mechanically determining the validity of any possible argument that belongs to its class. These rules operate the way they (and the entailment relation, their umbrella) have been generalized, that is, within the predefined, abstracted, constructed, notated deductive system as an artificial language of symbols. Automatic application therefore counts here not as an ugly name but as the ideal and proof of formal systematicity. The last decades, though, have seen two major additions to entailment.2 Both of these inference patterns are likewise of philosophical origin, yet not or not standardly truth-conditional: as befits their originators, affiliated with the ‘‘ordinary language’’ school at Oxford. Both indeed mediate between logic and language use, which explains their enormous appeal to neighboring disciplines. In ‘‘Logic and Conversation,’’ H. P. Grice () famously launched the term ‘‘implicature’’ for the information that an utterance conveys (and we hearers derive) over and above what it ‘‘says,’’ beyond the propositional content. In turn, implicature itself subdivides, most notably by the rules . A yet stronger variant, on which more below, would add: and in all worlds where Y is false, X is false. . Apart from inductive reasoning, of course. But then, despite all claims and logic-driven elaborations to the contrary, these two (I would argue, as apropos the one reviewed here I do argue throughout) ultimately belong to induction, since they are probability-based. Sternberg • Factives and Perspectives 133 that govern it: according to whether they are ‘‘semantic’’ or ‘‘pragmatic’’ (as later exponents would tag the divide). ‘‘Conventional implicature’’ arises from the lexical meaning of the sentence’s words. (So ‘‘therefore’’ in ‘‘He is English, therefore brave’’ conveys, via its encoding, a tighter link than the mere conjunction yielded by its propositional equivalent ‘‘and’’.) However, the best-known subtype, ‘‘conversational implicature,’’ no more derives from linguistic convention than from the logical truth-conditions but from their interplay in context with the guidelines of rational talk exchange, namely, the ‘‘Cooperative Principle’’ and the set of ‘‘maxims’’: their observance and/or breach generate our inference, along lines grown too familiar to need rehearsal. (In what follows, unless otherwise noted, ‘‘implicature’’ will refer to this kind.) The other relative newcomer, and our immediate business, is ‘‘presupposition,’’ introduced fifty years ago by P. F. Strawson (, : –) in his attack on Bertrand Russell’s () then dominant theory of reference.3 The bone of contention, ‘‘The present King of France is bald,’’ has gained notoriety ever since. According to Strawson, an utterance in ‘‘The X is Y’’ form does not assert or entail (à la Russell) but presupposes the existence of X, so that the utterer ‘‘commits himself ’’ to it on pain of a distinct ‘‘logical absurdity.’’ X’s existence is ‘‘a necessary precondition not merely of the truth of what is being said but of its being either true or false.’’ If the referent proves nonexistent, like theKing of France today, the statement is therefore neither: the presupposition (prerequisite, self-committal) having failed, the issue of truth or untruth ‘‘fails to arise.’’ An ‘‘absurdity’’ or ‘‘outrage’’ certainly results, yet other than the plain self-contradictory kind, with its automatic falsity, to which assertions made and denied in the same breath are alone liable (ibid.: , : –). From Strawson’s neither-true-norfalse counteranalysis, there emerges the definitional feature and test of his presupposition, namely, an epistemic commitment (or, for the addressee, an inference) that remains constant when you negate the statement. ‘‘X is not Y,’’ along with ‘‘X is Y,’’ would then presuppose X’s existence, uniquely so among the uses and inferences of language. In the five decades since, presupposition has become an industry. As Levinson (: ) notes, more has been written about it ‘‘than on almost any other topic in pragmatics (excepting perhaps speech acts).’’ 4 The bib. For a parallel in a few earlier remarks by Frege ( []: –) on Voraussetzung, see Beardsley : , –; Atlas ; Levinson : –; Burton-Roberts : –; Beaver :  n. . . Levinson : –, drawing on themore technical Gazdar : –, remains the most accessible critical survey. However, much like the field surveyed, these overviews need to be approached with care, being heavily theory-laden and almost unaware of sharing tacit 134 Poetics Today 22:1 liography in the much-cited Oh and Dinneen  collection alone runs to over  items, and its counterparts in Beaver  or Krahmer  suggest the rate of expansion.The parties to the debate include logicians, language philosophers, linguists, semanticists, pragmaticists, discourse analysts, cognitivists, even semioticians (Eco and Violi ).The idea has also been occasionally imported into literary study (e.g., Prince , : – ; Chatman : –; Culler : –; Banfield : –; Toolan : –, : –; Doležel ), usually with unimpressive results (but see Beardsley : , –, –; Dry and Kucinkas ). In the process, the range of phenomena, or triggers, subsumed under this heading has widened apace, far beyond Strawson’s expertise. Here are some presuppositional triggers of evident interest to any student of language (drawn from the much fuller list in Levinson : –): — definite noun phrases, with their presuppositions of existence, whether straightforward, as in ‘‘The author of Emma is brilliant,’’ or troubled, like the King-of-France reference, if any; — factive predicates, such as know, realize, regret, mind, be sad/glad/odd/ amusing, whose occurrence in the main clause (e.g., ‘‘He knew/realized . . .’’) presupposes the truth of the embedded clause (‘‘. . . John had left’’); — change-of-state verbs, finish or stop, responsible for the insinuation carried by the proverbial query aboutwhether youhave stopped beating your wife; — temporal clauses, those headed by (say) before, after, since, while: ‘‘He laughed before she did’’ asserts his laughter but presupposes hers; — nonrestrictive relative clauses, whereby ‘‘John laughed’’ derives from ‘‘John, who laughed, must have got the point’’; — cleft constructions, as in the inference from ‘‘It wasn’t he that laughed’’ to ‘‘Someone laughed’’; — counterfactual conditionals and related counterfactives: ‘‘If he had arrived . . .’’ / ‘‘I wish he arrived’’ presuppose his nonarrival. These phenomena (a fortiori the entire list) are remarkable for both their variety and their unity.What with the basic elements subsumed, the variety underlines not only the centrality of presupposition within the linguistic code, but its ubiquity across language use, from daily talk to the masterpieces of verbal art. There is no avoiding it and much to be gained from fundamental assumptions (those that most concern us) with the very theories they criticize. The same holds for the updated overviews in Van der Sandt  and Beaver : the first admirably comprehensive, the second oriented to the newest wave, and both heavy with formalisms, notational as well as conceptual, that reflect the ideal of the domain under review. Sternberg • Factives and Perspectives 135 understanding the distinctive inferential operation that we so variously produce and perform all the time. Distinctive it remains amid the variety, because the odd feature originally diagnosed by Strawson—constancy under negation or questioning— runs throughout the triggers.Whether we negate the definite description’s predicate or the factive, or the change-of-state verb, and so forth down to the counterfactual, the presupposition yet survives: that the author ofEmma exists, that John had left, that you used to beat your wife, or that he had not arrived. This survival value makes an extraordinary feature, one startling and unique to the point of incredibility. Neither entailment nor implicature share it at all, any more than plain assertions do. Try negating ‘‘Julia owns four houses’’ or (in a recommendation for an academic appointment) ‘‘Adam has a good car,’’ and see how the inferences that Julia owns three houses or that Adam does not qualify for the job evaporate. Again, reapply the test to a statement headed by the verb ‘‘presuppose’’ or the harder ‘‘entail’’ or the umbrella ‘‘infer,’’ and you will discover that such verbs are nonpresuppositional, other than factive, as is ‘‘to be a fact’’ itself. Little wonder, since all these follow our intuitive grasp of negation that presupposing somehow eludes and defies. This odd feature has been questioned, though, and predictably so: less from its own exceptionality as from the rule that virtually nothing about presupposition has remained unquestioned over the years. This lack of minimum consensus—or worse, progressive fission and erosion rather than buildup—is in turn unusual by any standard, including that of other inference types. Having adverted to the unmatched extent of work on presupposition, Levinson (: ) judges ‘‘a great deal of it obsolete and sterile’’— and all of it controversial, he might have added, and in effect does. Nor has the state of the art changed meanwhile for the better—the familiar issues of the s recur in new guises and variants—so that some theorists have understandably despaired of the whole inquiry. What comes to the same thing, most inquirers would decompose the type into its ostensibly welldefined fellows, the logico-semantic and the pragmatic. As understandably, again, a survey like Beaver  nevertheless preaches optimism by appeal to recent developments. But the fact that it starts by waving aside the ongoing basic deficiencies—lack of an agreed definition and of an agreed range of phenomena (ibid.: –)—is enough to suggest the contrary. Another witness to the power of institutional inertia. Despair or overoptimism? A more appropriate response than either would be a radical fresh start, whereby to break the impasse. There is nothing wrong or unreasonable about the object of study—I hope to demonstrate—only with the premises underlying the otherwise embattled 136 Poetics Today 22:1 approaches to it via the traditional disciplines. In other words, these approaches share (at times, even outdogmatize) the wrong originary paradigm, one become too second nature for them to notice, let alone question, through the surface differences that loom so large: generators, as it were, of all the heat, the despair, and the revisions by inertia.This common ground amounts to ‘‘legislating inference’’ a priori, the univocal, formalized, yes-or-no way, on the deductive model or ideal. (So, for example, if ‘‘I didn’t know he had arrived’’ keeps the embedded inference while the stressed ‘‘know’’ may lose it, then the negation rule allegedly fails the test and compromises the inferential type.) Not that implicature study, on the face of it prosperous, has dispensed with this whole bundle of premises from the first or outgrown them—Iwould further argue—but that when it comes to the essentials of doing inference, presupposition affords the real test case. Of these two comparative newcomers, presupposition has the stronger claim to rethinking, on grounds apart from its superior intrinsic interest to my mind or even its fiftieth anniversary per se. For one thing, it is irreducible to other inference types—as numerous abortive bids for assimilation attest—hence also best qualified to differentiate and generalize inferential activity as a whole, to tell type-specific variables from constants.5 For another thing, it is both language-specific (unlike implicature, though proponents may deny the extendibility to all semiosis) and encoded in the language system (unlike conversational implicature on Grice’s own definition of it 6). Yet its actual inference must also fit with the enclosing coordinates: the presupposed reality, the presupposing speaker or subject. Accordingly, presupposition is a remarkable case of the shuttle among language, world, and viewpoint whereby we make sense of discourse.7 For yet another thing, implicature has flourished, while presupposition remains in straits despite all the extensive work on it. This antithesis links up with both the critical and the constructive side of my interdisciplinary theme. Implicature has maintained contact with discourse—starting with Grice’s own notes on irony or metaphor, as well as talk exchange, and an early application (Pratt ) to literature—while presupposition lost, indeed neutralized such contact, officially or otherwise, almost at birth. Even the Strawson of Introduction to Logical Theory () did not appear . Ironically, far from assimilating (elements of ) presupposition, conversational implicature itself reduces to wider mechanisms; nothing about it is distinctive. But my present argument does not hinge on specifying this undeterminateness, though glances at it will appear below. . As ‘‘nondetachable,’’ in Grice : –. . As much of my work has explored this universal threefold shuttle, various parallel cases and crosscuts will be found in References: some bear directly (e.g., via factivity, quotation, perspective) on the ensuing argument. Sternberg • Factives and Perspectives 137 logical enough to his early ‘‘semantic’’ followers and theirs since: against his insistence on presupposing as statement-, hence occasion-bound, these neo-Strawsonians associate presupposition with pure, abstract, entailmentgoverned sentences.Their nominal ‘‘pragmatic’’ opponents actually follow suit, in both the desire to get the presupposition of utterances codified and in the maneuvers executed to have this bare paradox come true.The study of presupposition could not, and cannot, possibly advance, I will argue, on the formalistic terms established by the philosophers and taken up with variations by linguists, pragmaticists, discourse analysts, even by poeticians who should know better than to apply the machinery, judging from their own meager results, if not from the sound and fury at source.That half a century of trial and error and patchwork has left the basis shaky compromises the entire formal paradigm, regardless of variants. Indeed, on the constructive side, gravitating toward the protean workings of discourse and reviewing the features of presupposition in their light shows the way to a viable theory: this alternative can handle the issues under a single inferential dynamics that captures and explains the type’s uniqueness while associating it with the universal resources of inference as part to whole. As with the rest of inference—only more distinctively so— the choice here lies not between order and anarchy, legislating and sheer improvising, but between two mutually exclusive idea(l)s of system: aprioristic, once-for-all legislation, whereby to (fore)tell right language from wrong, and goal-driven regulation (the righting of the ostensibly wrong included) by appeal to awell-defined set of possibilities that among themyield the best sense, the tightest fit available or wanted in context.The choice, in short, lies between an unearthly and an operative mirror to what happens in discourse—particularly to the element of uncertainty that typifies (i.e., empiricizes, relativizes, ambiguates, probabilizes) all its inferences, lingering even when they are most signaled (e.g., presupposed) and we happiest with the output. Here exactly literature’s endless context-switching and imaginative verbal artistry (not least in the mimesis of everyday talk) best qualify it for illustrating the protean ways of discourse inference at large while putting the regularities, unique or universal, to the severest test. Its superior qualifications have nothing to do with its reputation for specialness—‘‘parasitism’’ among the hardheaded, poetic license among the literati, fictionality among all—as parallel examples from history writing will establish, along with realistic utterances obviously modeled on life. Further, the very appeal to the record, literary or otherwise, opposes another counterproductive inheritance from logic and linguistics, whereby the analysts’ practice runs against the theory on its own terms. I mean the deplorable pragma138 Poetics Today 22:1 ticist habit of inventing unanchored (or worse, covertly or equivocally anchored) examples for analysis, even ‘‘proof.’’ 8The anchorage below should thus hammer home the operative implications and complications of the self-evident truth forgotten by such analysis, namely, that discourse (inference) is always in context, possibly in one out of our world and never all of our own devising.9 For balance between generality and other values—expository, heuristic, even theoretical—the argument will glance en route at assorted triggers but concentrate on factive presupposition, and not just for the sake of brevity. As some of the triggers have been disputed, uncontestable members will alone serve to establish the nature and, with it, eventually the range of the phenomenon. Among these, again, factivity comes only second to definite reference in its importance to language and outranks it by other criteria. The mind boggles at the thought of our doing without know, be aware, comprehend, grasp, find out, discover, learn, realize, regret, resent, deplore, ignore, forget, bother, mind, matter, amuse, make sense, suffice, alarm, fascinate, exhilarate, be odd, glad, sad, proud, crazy, happy, surprised, relevant, significant, tragic (mainly after Kiparsky andKiparsky ).Not to speak of further tokens co-optedwithin a particular verbal system, like hear in Biblical Hebrew: the tell-tale epistemic commitment to the following that-clause, absent even from the modern dialect, gets encoded there in the lexical item (Sternberg , : , –). Also, factivity involves a complex sentence (e.g., the asserted matrix clause followed by the presupposed embedding in ‘‘He knew that the bird was gone’’). And this syntax renders it a case par excellence of a (if not the) key issue in the field, the ‘‘projection problem’’: how to generalize, by way of ‘‘formal’’ rule, which presuppositions of component parts (e.g., the thatembedding here) are ‘‘inherited’’ by the composite whole. The case grows yet more exemplary in that it both freely subsumes nonfactive presuppositions (e.g., the existence of ‘‘the bird’’) and gathers further components, like modals or negatives: as the elements diversely compound, so will the testing of the alleged ‘‘projective’’ rules and their very rationale. (These all fail, I hope to demonstrate, beyond salvage.) Most significant, as I will argue, built into factivity is the mechanism— roughly, the dynamics of perspective—that resolves other presuppositional conundrums as well.What on reanalysis we will find encoded there (in the . To this extent, the proposals for improvement in Beaver : – envisage a move in the right direction. . In Sternberg  I have already developed an argument formodeling pragmatics on poetics with a view to an alliance: such reorientation would have more constructive effects on the newer domain than weaning it from its inherited dreams of formalism. Sternberg • Factives and Perspectives 139 lexis along with the structure) and enabled (in the inferential process, visà-vis anything from ambiguity to absurdity) invites extension to the rest. Even so, the argument needs to take some shortcuts; I will try to compensate for them where possible by referring to my earlier studies in factivity (: esp. ff., b, ) and related issues. For extra brevity and continuity, but also for more substantive reasons, I will take know as my paradigm case. It has multiple claims to focal status. Even relative to its own associates, know stands out as one of the central verbs in language, ordinary, artistic, or scholarly. Again, within presupposition theory, hardly any other factive has elicited such intriguing responses. As will emerge, it has been distinguished from, say, regret: either favorably (because amenable, as it were, to entailment-like treatment, perhaps due to its ‘‘cognitive,’’ rather than ‘‘emotive,’’ nature) or invidiously (because unamenable to standard generative analysis). Last but not least, know figures in numerous metalanguages, as well as across language, and therefore widens my interdisciplinary theme beyond the fields alreadymentioned. Epistemology now obviously enters the arena to meet, inter alia, narrative theory with its polarity of omniscient versus restricted telling. Obvious or not, don’t these new joiners operate at a different epistemic level, higher, more abstract or strategic than do the mini-presuppositions borne by the common verb, let alone its various associates? How presupposed knowledge state(ment)s relate to inference-making and -conceptualizing at large, on the one hand, and to knowledge-oriented domains that apparently just cross with factivity at a single juncture, on the other, is a question well worth broaching. 2. The Dead-End Fallacy as against Integration Unlimited: Old Cruxes, Fresh Start To see why the multiform traditional paradigm fails, look at the kind of cases that have remained intractable to all approaches: * () Oedipus regrets killing his father although, in fact, he didn’t kill him. (Invented example, after Gazdar : , ) () At his machine gun at the stern of Hiyo, Chief Petty Officer Mitsukuni Oshita heard the cry ‘‘Torpedo Coming!’’ He began to count. At  he knew the torpedo had missed, and relaxed. An explosion jarred Hiyo. Oshita had counted too fast. (Toland : ) () My title, ‘‘The Fabrication of Facts,’’ has the virtue . . . of irritating those fundamentalists who know very well that facts are found not made, that facts constitute the one and only real world, and that knowledge consists of believing the facts. These articles of faith so 140 Poetics Today 22:1 firmly possess most of us, they so bind and blind us, that ‘‘fabrication of fact’’ has a paradoxical sound. (Goodman : ) () They are the center of everything, those shoes. They are it. I know that, now.Too bad it is not worth knowing.Too bad it is not true. It is not even temporarily true. (Barthelme : ) The historical, philosophical, novelistic discourses in ()–() equally manifest the radical incongruence that the opening asterisked exemplese concocts for its analyst to dissect and dismiss as impossibly anomalous. Each factivizes a statement—one of ‘‘knowledge’’ at that, unlike the mere emotive ‘‘regret’’—then takes away its factuality. The crucial question is not whether you find the glaring inconsistencies (openly in breach of epistemic commitment) easy or hard to settle, or how you go about it. The question is rather one of principle: whether you consider them discourse givens, like any other, and their settlement feasible and imperative—never mind challenging—with the onus of inference on yourself.To the field, the answer is No throughout, in multivoiced chorus. Presupposition theory has only one way of dealing with such factive instances—the short, negative, exclusionary way of (pre)judging them ‘‘unacceptable,’’ if at all producible or imaginable outside heuristics. Operatively, the tag signals ‘‘No further inferential action’’: a dead end. In this a priori ruling out, there is not much to choose between the two main lines of approach, the ‘‘semantic’’ and the ‘‘pragmatic.’’ Nor is it accidental that this short shrift, normative label included, has its equivalents apropos the originary definite reference and other triggers laboring under extreme (con)textual pressure. A rare, if negative, unanimity emerges on a wide front, so that presupposition theory as a whole blames its own failure on the phenomena that lay it bare in the testing. The dead end is the ultimate, by no means the only, proof of a lame start. ‘‘Semantic’’ approaches define presupposition in terms of truth, understood the mathematical way: as a relation between abstract, context-free sentences and the world.10 In these terms, special adjustments are needed to reflect, first of all, the type’s unmatched constancy between affirmative and negative. The semantic definition accordingly formulates a new twofold rule of entailment. Sentence X (e.g., ‘‘Oedipus regrets that he killed his father’’ or ‘‘He knew the torpedo had missed’’) presupposes sentence Y (e.g., ‘‘He [Oedipus] killed his father’’ or ‘‘The torpedo had missed’’) iff in all worlds (or readings) where X is true and in all worlds (or readings) . For details see the overviews cited in note  above and the further references in note  below. Among early exponents, perhaps the most representative is Keenan (), the most interesting, Leech (: –). Sternberg • Factives and Perspectives 141 where X is false,Y is true. So far, so good. But what if the presupposed cum entailed Y is false? By standard logic, this would inevitably make the presupposing cum entailing X false as well. Except that where X is false, Y is true, as above, and by logical definition, an entailment always holds. (Try imagining a syllogism that at times follows, at times fails.) To get around this dilemma, the approach generally abandons in turn the very classical principle of either-true-or-false bivalence. Instead, where Y is false, X is deemed neither true nor false: it has no value, to the loss of all propositional status, or must be assigned a third, ‘‘nonsense’’ value (as Keenan [: ] calls it).11 So, owing to these special arrangements, whatever becomes of the X, the presupposed Y comes out alive somehow: X indeed suffers if necessary the fate of truthvaluelessness or nonsense for Y to live on.Y remains uncancelable—or else the whole logic of entailment breaks down—and its sins (e.g., the nonexistence of the presupposed King of France) will be visited elsewhere. Never mind whether, even as logical abstractions go, the result has any bearing outside the system, on ordinary thought, language, and discourse. Just observe what happens after all the drastic adjustments: in ()– () the presupposition not merely turns out false but gets falsified by an ensuing sentence (e.g., ‘‘He didn’t kill him’’ or ‘‘An explosion jarred Hiyo’’) to downright contradictory effect that would undermine the entailment and with it the system from within. The founding semantic accounts (e.g., Keenan ) never envisage the possibility of such outrage. And Leech (), who does, invents a weaker inference type, ‘‘expectation,’’ and consigns it to ‘‘pragmatics’’ in the hope of averting disaster from presupposition and logical semantics. With its word/world relations weakened to fall below ‘‘the criterion of uncontradictability’’ and outside ‘‘the abstract logical system of language,’’ an expectation ‘‘may be cancelled out by an appended qualifying statement: ‘Few girls are coming, or maybe none at all’ ’’ (: , –, –). Even so, the double switch of label would not affect ()–(). ‘‘Know’’ and factivity in general still belong to Leech’s narrowed, shielded, and fortified presuppositional domain: all the more resistant in theory (if possible) to the idea of ‘‘cancelability’’ or ‘‘contradictability,’’ hence all the less explicable (never mind resolvable) in face of its actualization. On every formal semantic account, the worst nightmare of absurdity would then come true: the entail. The likeness to Strawson’s foundational definition, as cited at the outset, goes with a basic difference. Strawson refers not to sentences but emphatically to statements, which use sentences on particular occasions and alone bear truth value. His analysis is therefore ‘‘pragmatic’’ avant la lettre, only not nearly pragmatic enough. The difference comes to little at the end of the day, though, as it will among his followers on either side. 142 Poetics Today 22:1 ment of an affirmative sentence deduced and denied in the same breath apropos the same world. There, impossibly, Oedipus did kill, the torpedo did miss, and didn’t.12 So-called pragmatic approaches, however vocally opposed to reducing presupposition to truth-conditionality and however divergent among themselves, would follow suit here in outraged judgment of ()–().The extremes meet (as a rule, indeed mix) across an official, long advertised chasm in the very idea of this inference type. To the semanticist, given the definitional sentence/world relation, ‘‘whether anyone actually utters or believes some sentence has nothing to do with whether the sentence makes a particular logical presupposition.’’ To the pragmaticist, the inference is defined by ‘‘the relation [e.g., appropriateness, agreement] between the utterance of a sentence and the context in which it is uttered’’ (Keenan : –). Given the endless flexibility of ‘‘context,’’ you would expect the pragmaticist to declare any such ‘‘utterance’’ adjustable with(in) the context ‘‘relating’’ to it: somehow, the inference at issue (however odd within the decontextualized ‘‘sentence’’) would then always endure meaningfully, rather than in a limbo of the logician’s devising. But this expected pivotal followthrough has never materialized. Instead, the bearings of context(lessness) on the inference’s survival in distress simply polarize between the approaches. No longer marked by its constancy, à la Strawson and his logical semanticizers, presupposition now reverses all the way toward ‘‘defeasibility’’ or ‘‘cancelability.’’ Nor does this fate befall it typically under exigencies so extreme as ()–(). In an approving overview of pragmatic efforts, the type is generally defined as ‘‘liable to evaporate in certain contexts, either immediate linguistic context or the less immediate discourse context, or in circumstances where contrary assumptions are made’’ (Levinson : ff.). For example, in ‘‘I didn’t know that he was the King of France,’’ the presupposition expressed by the that-clause would then drop out (unlike its analogue in ‘‘I didn’t know that he was the Mayor of Paris’’) because of what we assume about the world.This new ‘‘crucial property’’ accordingly becomes a ‘‘touchstone’’ of theoretical adequacy, to the wholesale rejection of the ‘‘semantic’’ approach, which cannot and will not accommodate it as a matter of principle (ibid.). In effect, this property also doubles as a touchstone of discourse acceptability, to the summary rejection of ()–(), which will not cancel into consistency: the presupposed tenor of the object-clauses never evaporates, the way it should by that definition, but jars against the sequel. . For broader devastating critiques of the semantic approach, see Wilson ; Gazdar : –; Levinson : –; Van der Sandt : –, –. For attempts to defend and revise it, see Seuren ; Burton-Roberts . Sternberg • Factives and Perspectives 143 Unsurprisingly, the reversal of touchstone carries its own penalties, which all issue from the original sin, as it were, of non sequitur. I mean the omission to follow through the utterance/context interplay, and so to opt for principled adjustability (what I will call ‘‘integrability’’) in lieu of possible, essentially unpredictable defeasibility. To the theory as well as to the inference type, the choice is between life and death. To cut a long and multilinear tale of woe short, this reversal falls into the specious binarism ‘‘either constancy or cancelability,’’ as if any middle were excluded.Yet, other alternatives (e.g., constancy across well-defined variations) do remain open in principle and, we will discover, fit the case, including the notorious problem cases. Meanwhile, the fallacy in pole-switching leaps to the eye.Why, if something has to give way, should the presupposition ‘‘evaporate’’ rather than the ‘‘contrary beliefs’’ held in ‘‘context’’? And held by whom? Need the presupposer (e.g., ‘‘I,’’ regarding ‘‘the King of France’’ above) voice, let alone endorse his addressee’s belief—or anyone else’s: ours, the group’s, the culture’s—or does the addressee need to infer the presupposer’s belief after the fact(ive), completewith belief-frame?Why should I commit myself to your mental representations, or even to those we supposedly hold together, and why should yourmind count as the reference point for inferring the presuppositions of my mouth? Nelson Goodman in () thus evidently attacks the (pre)supposed ‘‘knowledge’’ of the ‘‘fundamentalists,’’ of the tradition, of the discipline, quite possibly of us readers as well, establishing his own frame of reference, epistemic as otherwise. His diametric opposites in viewpoint, or authority, enjoy the same freedom of self-commitment against the world’s opinion. Such an example already discredits the two basic pragmaticist novelties at once: the preconceived ideas of ‘‘cancelability’’ and/by/in ‘‘context.’’ (More on this ‘‘common (ormutual) knowledge’’ fallacy below.) As it is, with the middle excluded and contextuality muddled, further trouble after trouble ensues, by a kind of chain reaction. Thus, on this approach, presupposition does not so much change distinctive features as lose its distinctiveness vis-à-vis other inference types and to them, and not very uniformly at that. In the process of identity loss, it grows stiffer here in their image, softer there, to meet opposing exigencies. On the one hand, if only for survival where it does survive, presupposition usually gets assimilated (after the semanticists) to the hardness of entailment, at least in affirmative sentences: so, among many, Gazdar (), from whom () derives, and the just-quoted Levinson (). In factivity, say, the main clause (‘‘Oedipus regrets . . .’’/‘‘he knew . . .’’) would then entail the that-clause, just as ‘‘The present King of France is bald’’ would the existence of the referent—which should make either entailed presupposi144 Poetics Today 22:1 tion absolutely indelible, durable across contexts (as logical ‘‘worlds’’). On the other hand, to accommodate putative nonsurvival (against the semanticists), presupposition now comes to share ‘‘cancelability’’ or ‘‘defeasibility’’ with Grice’s conversational implicature, where this feature actually originated.13 As Grice (: ) notes, such implicata may be canceled, ‘‘explicitly’’ or ‘‘contextually,’’ wherever it becomes evident ‘‘that the speaker is opting out’’ of the Cooperative Principle.We need only extend the feature, as it were, to an analogous, if not overlapping, domain of inference. But how does one opt out (or otherwise get rid) of presupposition if the very language used encodes it—as does the sense of ‘‘know’’ above—not to mention the binding semantic entailment associated with it? Predictably, the available Gricean aids to riddance, such as they are and where they are at all extendible by analogy, would fall short. (Examples such as []– [] maximize discordance, without enabling their utterer to ‘‘opt out’’ of commitment: they rather bring out the inadequacy of any implicature-like toolbox to the job.) One must look elsewhere for rescue. So presuppositionalists have invested untold wit and work in the attempt to devise stronger or subtler rules and measures of canceling than apply to the analogue. But not even an optimist (e.g., Beaver []) would claim that any adequate, far less uniform set of the desired ‘‘projection’’ algorithms has resulted, or that the very way to the set is undisputed. After fifty years, would-be projectors still wonder and quarrel about how the jumps that they intuit between presuppositional survival and nonsurvival (so called) are to be formalized across the various data, triggers, constructions, environments. Among the ensuing counsels of despair—too numerous to list—some have even attempted to push the intertype analogy to its logical finish and deny presupposition in turn its linguistic encoding as well. In turn, because Grice (: –) further characterizes implicature as boasting ‘‘a high degree of nondetachability’’: it is impossible ‘‘to find another way of saying the same thing, which simply lacks the implicature in question.’’ Many paraphrases co-implicate. By analogy, Levinson (: , , –) argues, if presupposition is cancelable, it cannot remain detachable, ‘‘tied to surface structure’’ or to the ‘‘conventional meaning’’ of triggers: factive predicates, temporal clauses, counterfactuals, and the rest. Instead, given the same truth-conditional makeup, the same presuppositions will arise via ‘‘general pragmatic’’ rules of inference. Never mind that throwing presupposition out of ‘‘detachability’’ would . Or worse, share noncancelability with conventional implicature, yet leave room for blocking or filtering the deductions (an alternative popularized by Kartunnen and Peters []); the outcome amounts to cancelability under another guise. Sternberg • Factives and Perspectives 145 take away its last distinctive feature, or that Grice himself leaves certain implicatures detachable.14Never mind even the empirical disproof that factives are unparaphrasable (e.g., know into firmly believe) without loss, except when intersubstituted among themselves (know = be aware). Just consider this ultimate flattening move in its own terms. It goes to tilt, if not abolish, the balance between the inference’s two aspects on the pragmatic view— the verbal and the contextual.The underlying drive is transparent enough. With the ‘‘ties’’ to language severed, presupposition would become amatter of content-in-context alone, and so hopefully easier to maneuver into some general discoursive lawlikeness. But the results never justify the hope— quite the contrary, as one might predict. If anything, context is far less reducible to order than the language system, and hence will only grow more unruly in the absence of linguistic inference triggers to balance and delimit its infinite variability. In reason, the less or looser the strings attached to the inference—the greater its ‘‘nondetachability’’—the less regularly and foreseeably ‘‘projectible’’ it turns. We need not therefore look more closely at the available pragmatic accounts to diagnose the hopelessness of the enterprise as conceptualized and pursued from the outset.15 The surface details and variants that loom so large in the analysts’ own eyes pale beside the common unmotivated selfdivision: between language code and utterance context, entailment and implicature, survival value and evaporability. They all want to have (or distribute or reduce) presupposition both ways, the hard (nothing less than the hardest) and the soft, with the inevitable outcome of falling between the stools. Methodologically, they want not just to formalize the inference into automatism—along with the logical and linguistic semanticists that they nominally oppose—but to formalize the very open-ended play of context that their opposite numbers at least reasonably neutralize altogether as unformalizable. Declaring presupposition ‘‘nondetachable’’ and/or epistemic ‘‘background,’’ on top of everything else, merely carries the impossibility to an extreme in leaving nothing but the open-ended variable to be formalized. Across all existing treatments, thewould-be pragmaticists run against the pragmatic spirit of language in use, context, discourse. As a functional, not formal, system, language use needs a theory that will systematize variations accordingly: by the roles they play and the differences they make within the . Or, for thatmatter, that the very idea of (non)detachability suffers fromvicious circularity, as persuasively argued in Van der Sandt : –. . The main lines are still the two originally suggested in Kartunnen and Peters  and Gazdar . Beaver  gives an updated overview of how they have since been criticized, amended, supplemented, developed, even crossed. More specific references will appear below as necessary. 146 Poetics Today 22:1 encoded limits unique to, say, presupposition. In this shift of ground, contextualizing the encoded attributes of presupposition transforms from an obstacle to systematizing (how to freeze the fluid, bind the boundless, automatize inferential life?) into its guideline, even rationale: as distinct from both essentially contextless entailment and purely contextual implicature. The sharpest measure of the gulf between these two orders of theory, as well as of the self-division within the pragmatic one, is what I call the dead end fallacy. Whenever the cancellation account will not yet save the data from apparent inconsistency, as in face of ()–(), they get labeled ‘‘unacceptable’’ and thrown out: canceled, actually, along with the recalcitrant presupposition they signal. Nullifying the whole instead of the discordant part: where the softer way fails, the hard takes over. As theQueen of Hearts lays it down, either you or your head must be off. ‘‘Hardness’’ again connotes the emulation of hard, especiallymathematical, science. Wonderland apart, the expedient evidently originates in the formal binarisms of pragmatics’ mother disciplines: the logician’s ‘‘valid/ invalid,’’ the linguist’s ‘‘grammatical/ungrammatical,’’ the semanticist’s ‘‘meaningful/meaningless (anomalous, nonsensical),’’ each specializing a variant of the overall ‘‘un/acceptability’’ or ‘‘ill-/well-formedness’’ divide. The same binary judgment, with the tag(s), now carries over to presupposition on all approaches. Except that ‘‘semantic’’ presuppositionalists might display a consistency of sorts in extending the acceptability judgment, and here the short dismissal, to entities of the same rarefied order, namely, sentences. But their pragmatic rivals impose the extension on the corresponding entities outside the language system, the presuppositional utterances and utterers that use the sentences in discourse. (Compare speech acts, likewise born in ordinary-language philosophy, likewise context-bound, and likewise subjected to a variant of the mother dichotomy, this time ‘‘in/felicitous’’ or ‘‘in/appropriate.’’ No wonder this label is co-applied to presupposition as early as Austin : , –, , and often since.) The unreasonable jump between levels, or systems, would now rationalize the disqualification of whatever resists both the preserving and the canceling formulas ‘‘projected’’ by theory. All the more unreasonable, the disqualifier’s show of reason, since presupposition itself uniquely lies between those levels—encoded in one, hence uncancelable, discoursed in the other, hence always contextualized and integrable, even given ()–(). Resistance then becomes a test of adequacy and, where outlawed, a clear negative verdict.16 . As I will below link factivity to reported discourse, note the parallel argument concerning the preconceived exclusion of data in formalist (e.g., ‘‘generative’’) approaches to the latter: Sternberg • Factives and Perspectives 147 But perhaps contradiction makes such an outrage as for once to justify disqualifying without further ado—indeed to disqualify itself—regardless of how and where it occurs? Traditional views of it imply as much, Strawson’s among them: Suppose a man sets out to walk to a certain place; but, when he gets half-way there, turns round and comes back again. This may not be pointless. He may, after all, have wanted only exercise. But from the point of view of a change of position, it is as if he has never set out. And so a man who contradicts himself may have succeeded in exercising his vocal chords. But from the point of view of imparting information, of communicating facts (or falsehoods) it is as if he had never opened his mouth. He utters words, but does not say anything. . . . Contradicting oneself is like writing something down and then erasing it, or putting a line through it. A contradiction cancels itself and leaves nothing. (: –) Extended to statements of ‘‘The King of France is bald’’ kind, this equation of logical absurdity with nonsaying (‘‘as if he had never opened his mouth’’) yields the judgment that they are neither true nor false: the question does not arise (to repeat his favorite formula) because the statement in effect never did, or not as one of fact. If anything, this makes the absurdity worse in (a) sense than direct contradiction, which at least necessarily boasts a propositional content and truth value (false)—the very attributes imputed to the example by Russell and denied by Strawson in the change to presupposition. The majority of Strawson’s heirs, who take ‘‘Oedipus regrets that he killed his father’’ in () to entail as well as presuppose that Oedipus killed his father, would deem the ensuing ‘‘although, in fact, he didn’t kill him’’ downright contradictory—Russell style—but withmuch the same exclusionary effect on the whole. Strawson’s metaphors for the unacceptable duly recur, too. They even sharpen, if possible, from the ‘‘as if ’’ unuttered and ‘‘words’’ uttered without ‘‘saying anything’’ into the ‘‘unutterable’’ (see, e.g., Miller and Johnson-Laird : ; Lyons : ff., ) or, on the addressee’s part, the ‘‘unreadable/uninterpretable’’ (Van der Sandt : , ).Whatever its locus and typology and subrules, presupposition then assimilates to the cross-disciplinary ground rule of ruling data in or out a priori, by formal fiat.17 ‘‘By itself, grammatical analysis can at most yield grammatical features; and it is predictable that when pulled out of its circumscribed limits to handle context-dependent discourse, it should only expose its own inadequacy’’ (Sternberg b: , ; cf. also the detailed critique of Banfield  inMcHale ). Here, with our inference type, the empirical disproof only assumes a sharper theoretical edge because the victim of exclusion by fiat proves even more systematically explicable along lines unknown to its excluders. . Here is an echoing counterpart on the level that Strawson would be the first (actually, was among the first) to distinguish from statements, namely, the abstract ‘‘sentences’’ of the 148 Poetics Today 22:1 Given this alignment, presuppositional dead-enders (nonstarters by fiat) join dropouts (in context) on a scale of incongruity. It only remains for the theorist to construct the machinery from which every possible utterance will emerge marked as acceptable (‘‘utterable’’) or unacceptable, and if the former, with the right survivors rightly interpreted in the right context. Nothing approaching such formalization has been (nor, as I argued, can be) devised, but hope springs eternal regardless. Miller and Johnson-Laird (: ) report at second hand on a computer program that ‘‘returns an error’’ when faced with presuppositional violation: if genuine, the article would long have made history by now. In the most developed neo-Gricean theory, Sperber and Wilson  [], its equivalent is envisaged as a thing of the future, yet of such a range that presupposition (‘‘background entailment’’) will undergo its vetting, or vetoing, along with all other communicated obliquities. This ‘‘formal deduction system’’ aspires ‘‘to model the system used by human beings in spontaneous inference.’’ At every fork, if all goes well, only the stronger of any two incompatibles will survive the processing: The deductive device has the power not only to read and write assumptions [factual representations] in its memory, but also to erase them. Let us assume that when two assumptions are found to contradict each other, if it is possible to compare their strengths, and if one is found to be stronger than the other, then the device automatically erases the weaker assumption.When an assumption is erased, the device also erases any assumption which analytically implies it, and the weaker of any pair of assumptions which synthetically imply it; this procedure applies recursively until no more erasures can take place.When such a procedure is possible, the contradiction is eliminated at the root, and the deductive process can be resumed. (Sperber and Wilson : –, –) Admittedly, though, ‘‘such a procedure’’ whereby to break the impasse and resume inferential work is not always ‘‘possible.’’ Nor does it greatly matter whether one further idealizes the model of spontaneous inference by having it, somehow, sometime, overcome the if ’s to enable the exclusionary choice along with its whole chain reaction. Meanwhile, the pressure on the addressee here-and-now to do something in real time about irreducible contradiction does not appear to trouble the model-builders, while the possibility that the addressormeant something by it, or that themeaning language code. ‘‘On then amThursday furiously bed ’’ exemplifies ‘‘strings of recognizably English word-forms . . . which are neither grammatical nor ungrammatical. . . . The questionwhether they are grammatically well-formed or ill-formed does not arise’’ (Lyons : ).Unutterability cuts across orders of language (use/system) and of value (grammatical/epistemic). When it comes to the infringement of an existential presupposition, the idiom duly occurs: the offender ‘‘fails to express any proposition at all’’ (ibid.: ). Sternberg • Factives and Perspectives 149 is inferable, escapes notice altogether. So do the amply attested facts that humans have managed to address the exigency quite well, that their interlocutors (and ours) have relied on this ability to find a way out, and that the communicative process has advanced not only through but via the horns of the dilemma.The worlds of art and quantum mechanics and everyday talk alike thrive on contradiction.We do better and other than the fancied regulator of our spontaneous doings: the model, actual or ideal, builds upon false premises that literally end in deadlock. Were not articles of faith so liable to bind and blind, as Nelson Goodman reminds us, all this pseudo-logic of the terminally unacceptable would be hard to believe.Whatever holds for ‘‘sentences,’’ however andwherever they exist, ‘‘utterance’’ and ‘‘unutterability’’ make a contradiction in terms— nonetheless so where the metalinguistic pairing appeals to the contradictoriness of the language used. The former clash alone is unresolvable in reason. A piece of discourse once uttered, how to unutter it? Granted, it can be struck out without trace, but that is another thing altogether: applicable only to private writing or thought, never a matter of record, contact, address, acceptance. In utterance, vocal or written, even if retracted or repaired—a third thing—the discourse will have been communicated nevertheless, with all its infelicities: withdrawal will only (re)draw the addressee’s notice to it, especially to the infelicities, most of all the contradictions. And if left on the table with them all intact—a fourth thing, least imagined by the theories at issue, presuppositionalist inter alia, yet no less widely evidenced—how can the addressee help ‘‘accepting’’ it/them as the official record, for better or worse? 18 As no viable discourse system or model rejects any token out of ‘‘acceptability’’ before the event, nor does the addressee after it on grounds of ‘‘unreadability,’’ not even in extremis. Some way into readability always arises. Come to that, the sharper the encountered offense against reason, or the rules, the higher the interest, the challenge, the gains, maybe the stakes of reasoning it out—and the losses that punish resolution failure.19 With all due respect to the law of contradiction, we must not take its breach for a terminal disease, much less one with immediate effect; and a fortiori lighter . The last two discourse-with-discord cases will soon be refined and exemplified in terms of my own approach to inference as multiform integration. For now, recall the vast body of work on the genetic fortunes of texts (e.g., the Bible) and the upsurge of interest today in conflictual, ambiguous, or undeterminablemeaning, respectively; see alsoMcHale : – on postmodernist writing ‘‘under erasure’’ and van Peer  on ‘‘mutilated signs.’’ . Even by the negative definitional measure of ‘‘relevance’’ in Sperber and Wilson  [], the cost of processing should at least be weighed against the cost of nonprocessing. 150 Poetics Today 22:1 offenses such as presuppositional inconsistency.They all rather make, even mark, our starting point for inference in the living reality of communication. More generally, such breaches and offenses fall under the all-embracing principle that I would call the Law of Reciprocity: whatever is expressed, or imaginably expressible, is by the same token explainable. A strict (teleo)logical necessity, if you think about it.Why would the speaker give utterance to the absurd ‘‘The King of France is bald’’ or to ()–()? That this is the right question to ask, the one we addressees must and do in reality ask, has been clean forgotten by the would-be pragmaticists. So has been the infinitude of inferable answers, many of them not even concerned to rationalize the statement as such: ‘‘He’s insane,’’ ‘‘He’s experimenting with iambic trimeter,’’ ‘‘He’s alluding to the Strawson/Russell debate’’ . . . Whatever your answer, you have thereby explained his expression, contextualized his absurdity. (Here, you even literally ‘‘reciprocate’’ by inferring some closure from the troubled presuppositional inference.) In our communicative affairs, this Law far outranks that of contradiction, to the extent of using it as one premise and trigger for making sense of the expressed—on a par with, say, its very opposite, redundancy.20 Inversely, our explanatory resources vis-à-vis discord far outreach the unilinear rationale of (or born of ) logic proper, single-mindedly intent on whether the givens and their derivations compose a noncontradictory propositional whole, a frictionless representational tenor, ideally a followable chain. In presupposition, as in all other spheres and disciplinary guises, the artificial rigor of formalism is therefore a rigor mortis, out of touch with this or any discourse world. Literary study tends on the whole to the opposite extreme. Its normal practitioners will find an institutionalized dead-end fallacy hardest to believe. Copingwith absurdities of diverse kinds, in diverse ways and texts, has been their daily bread formillennia. As already noted, however, inference theory in the field lags sadly behind the mass of data studied. Collective knowledge on view does not even match individual (if you will, intuitive) practice and competence, far less those of the best readers over the ages. The obstacles to ongoing conceptualization here are too many to number, and themselves typically underexplored. But they include amnesia, factionalism, overemphasis of difference, lust for novelty, alongside objective factors such as the complexity and the intrinsic value of the artwork, hence . My pairing of the informational extremes alludes to the fact that redundancy has elicited from presuppositionalists much the same dead-end verdict: a rider like ‘‘and he killed his father’’ in () would then be as unacceptable as its actual inconsistent mate. See note  as well as example (). Sternberg • Factives and Perspectives 151 of the understander’s operations, and the sheer, increasing magnitude of the task—all so endemic as to be taken for granted.The vastness and variety of the heritage associable with the issue, even if kept down to the fraction bent on the handling of textual problems, accordingly make the ‘‘state of the art’’ difficult to outline. Any random selection, though, will reveal it as the babel within the babel of inference study. Consider one series of landmarks: the Rabbis’ array of interpretive middot, Aristotle’s Poetics on actional and verbal resolution, the medieval levels of signification, Homeric and Biblical low and high criticism, William Empson ( []) on types of ambiguity, Monroe Beardsley on literary ‘‘explication’’ in his unusually empirical and meaning-oriented Aesthetics (: esp.  ff.), Roland Barthes’s () book-length exercise in post-structuralist reading . . . These mark a broken line, rich in fresh inference-related departures and terminologies, thin in backand crossreference, let alone synthesis, with Beardsley a shining exception that proves the rule. Even on a front yet narrower and of more direct concern to us than assorted additions to readerly equipment—the sense, or sense-making, of representation—much the same positive/negative balance emerges (Sternberg a). Hence, if you want to focus the existing disciplinary resources on a new inferential crux like presupposition, you scarcely know how to begin, where to turn, what would align with what. The field might envy its neighbors their comprehensive, updated textbooks and surveys, which orient the inquirer for better or worse—except that, as the foregoing analysis testifies, orderliness is not all. Their chapters on the subject, like the given references themselves, just mirror-image the literary critical imbalance: entailment, implicature, and presupposition unwittingly isolated from the rest, as though there were nothing else, and the first alone tolerably conceptualized or adapted after the logicians. We should not have to make the choice between the imbalances. In the study just cited, I formulate a research program designed finally to subsume all our work and ways of inference, literary or otherwise, under the master heading of ‘‘integration.’’ Or the other way round, the program is designed to break inference down into the types, factors, forces that endlessly compose in practice. They compose in our reading of a discourse or a genre or just an excerpt, in our assorted reading styles, in our encounters with ambiguity, in our hesitancy between intent and giveaway, in our quest for a certain meaning or effect or configuration: wherever and however we integrate parts by appeal to some whole (not necessarily a propositional whole, not even when presuppositions, say, are to be integrated). Following from the Law of Reciprocity—what is expressible is explicable— the program would translate it into operational terms: 152 Poetics Today 22:1 What informs the multiform activity I have called integration is not so much a sense of purpose as a rage for order. If in other words such a theory (like the interpretive activity it systematizes) always works with an opposition, it is with the all-inclusive one directing the whole process of reading: between the coherent and the opaque, the fragmentary, the incongruous. So any mechanism that serves to establish or undermine, initiate or terminate, reveal or conceal, resolve or ambiguate a pattern (and thus to determine intelligibility) deserves equal consideration, though its integrative role will of course vary in particular cases. Thus, themechanism of integrationmay be functional (authorial strategy) or genetic (authorial slip), formal (including typographic arrangement) or semantic (from co-reference to world-picture), sequential (like causality) or suprasequential (from rhyme to thematic counterpoint), extratextual (historical circumstance or philosophical system) or intratextual (repetition, built-in probability-register) or intertextual (from allusion to convention). It may be normative or empirical, lexical or grammatical, psychological or ideological, perceptual or perspectival, referential or rhetorical, logical or chronological or analogical, ubiquitous or archetypal or generic or unique. And so on, till the whole repertory of ordering resources and sense-making combinations available to humans has been covered. (Sternberg a: –) From this wide-ranging repertory,Tamar Yacobi (, a, b, ) has picked out five ‘‘mechanisms of integration’’ in her work on narrative (un)reliability; and theymay now conveniently exemplify the principle, with special regard to our business with inferring inference, or reinferring it under pressure. In rough outline, the genetic mechanism integrates odd features by appeal to accidents in the discourse’s production (e.g., Freudian or authorial slip) or transmission (mutilated, censored, badly edited text): to ‘‘noise,’’ in short. The existential logic assimilates them to the unusual (e.g., supernatural) ontology posited by the text, the generic to latitudes and breaches (of standard expression, causality, mental life) codified in the text’s family.The functional logic makes teleological sense of incongruities, by reference to the ends (shock, amusement, quest for closure) they fulfil; and perspectival integration attributes them to the lapses of themediator as speaking, thinking, writing subject (e.g., an unreliable narrator ironized by the author or the everyday quoter). Let me now show how these universal mechanisms (inter alia) operate (inter alia) on our inference about or toward inference, most specifically about presuppositional inference at the dead end, as it were. Among this set of resources, none has suffered such theoretical neglect in inference study—and across disciplinary lines—as the genetic. Logicians do at least envisage and name the possible misapplication of entailment Sternberg • Factives and Perspectives 153 rules, if only in warning against a bad end where a good one might and should come as swiftly. Their pragmatic and literary counterparts scarcely recognize the human liability to mischance in producing and transmitting discourse—or to noise in receiving it—because of the ideal model of communication assumed. Some visionary variant of error-free ‘‘competence’’ filters out the bleaker realities of ‘‘performance,’’ along with their inferential bearings.The shared lust for idealizing is aided and abetted by more discipline-specific forces.Thus the Saussurean langue/parole antinomy, doubling as that between the systemic and the fortuitous, hence ungeneralizable; or the presumptions of rationality, intentionality, economy (‘‘relevance’’) in language use, on which more below; or poetic articles ranging from inspiration to self-conscious artistry.Theorists will accordingly ignore the trials of genesis that they must daily experience as writers, readers, speakers, listeners alert to the self-exposure of others, consumers of textual criticism, heirs to Freud’s psychopathology of everyday life. It is not so much that they theorize against their better knowledge of the ever-open distance between intention and expression, then between expression and transmission, as that they consider the knowledge unworthy of theorizing into a universal mechanism of integrative inference.21 All this has even larger consequences than may appear, because idealizing error out of the system also minimizes the subjective element, hence the perspectivity lying at the heart of discourse and marking off factive as presuppositional discourse.Wide asunder in intentionality, betraying one’s own self and enacting another’s self-betrayal (e.g., via contradiction) yet refer the giveaway alike to the workings of the subject’s mind. As a reminder of discourse life and trouble outside the Ivory Tower, consider a simple pinpoint example. One novelistic character enters ‘‘brown’’eyed; a hundred pages later, she fixes the hero ‘‘with an artless blue eye’’ (Gilbert : , ). Those of us who spot the inconsistency—and our memories are not perfect, either—will easily enough suspend or qualify the model of ideal competence in face of it, at least ad hoc. After all, even Homer sometimes nods. We then refer the clash to the author’s amnesia, oversight, indeed lapse into ‘‘artless’’-ness, the way we would geneticize any ordinary typo, only perhaps with more deliberation. For, in principle, all other sensemaking resources than the genetic are co-available, whether as substitutes or partners. The question is only what fits the text best: the . Contrast, for example, Lyons :  or Sperber and Wilson  []: . Much the same holds for the otherwise richer ‘‘integrational semantics’’ presented in Hrushovski : amid the fortuitous likeness in terminology, its emphasis on semanticizing discounts the understander’s hovering between the authorized, or ‘‘semantic,’’ and the genetic as rival routes to integration. 154 Poetics Today 22:1 appeal to a performance error or to some communicative, competencepreserving design, and if the latter, which. Thus, the color of the eyes might literally have changed from brown to blue—and with it the principle of integration, from genetic to existential— given the appropriate reality key: that of fantastic writing, say, or of RobbeGrillet’s La maison de rendezvous, where characters keep switching attributes. Modern literary fantasy even prefers to open with a realistic-looking world that shockingly twists out of realism thereafter. Hence such otherwise impossible metamorphoses do (and, if posited here, would) compound existential grounding with generic license and with the functionality of surprise dynamism. In lieu of one resolution, three at once; instead of a slip betrayed in the writing, thick sense conveyed in the process of reading. But would the alluring alternative save the text, or rather substitute another, larger incongruity with its very coordinates as a detective novel, realistic inferential key included? The fifth, perspective-oriented logic also applies here in theory (change of color with change of observer? we wonder). It may even cross the genesis, keeping the odd lapse as such in play, but now with an inward turn and discourse role. For the genetic misadventure can itself get dramatized, hence newly integrated from within, in terms of the generator’s self-repair: () She woke, blinking in the sunlight, upon the opening of the attic door andmoved to rise from the bed.A youngwoman, a blue-eyed, brownskinned Israeli, came in. No, a blue-eyed Arab woman, dressed in a blue shirt and dark skirt, like a lithe Israeli. She was carrying, over her arms, some cloth that looked like the black-out curtains of wartime England. ‘What’s the time?’ Barbara said, feeling down into the pocket of her dressing-gown for her watch, and realizing, then, that the young woman was Suzi Ramdez. (Spark : ) () Tight-lipped, she told him that she was going back to Arden posthaste. She couldn’t be expected to hang around with the press and everyone else knowing that she was responsible for last night’s false alarm. Or, to be more accurate, that her husband was responsible. (Critchley : ) () ‘‘You’re awful smart, you know that? I wish I had a college education. . . . Oh, yeah, you’re right. I did have a college education. I must not’ve been paying full attention.’’ (Higgins : ) Observe the explicit markers of self-repair: ‘‘No,’’ ‘‘to be more accurate,’’ ‘‘Oh, yeah, you’re right.’’ Observe also the variations between speech and thought under instant repair; between genuine and pretended (e.g., joking, as in the last case) slips of tongue/mind; or among the existential Sternberg • Factives and Perspectives 155 (‘‘ ‘Israeli’ ’’), factive (‘‘knowing that . . .’’), and counterfactive (‘‘I wish I had . . .’’) presuppositions that get adjusted in mid-discourse, to the limit of reversal. Throughout, however, the mouth or mind in question enacts a presuppositional stammer, usually followed by an advance (leap, groping, outside drive) toward stable commitment. Real-life equivalents doubtless suggest themselves, except that the genetic process unfolds in ()–() at a remove from first-order speaking—and presupposing—and so between inverted commas, visible or deducible.We in turn eavesdrop on the characters dramatized at eventful self-expression, rather than play their addressee. Dramatizing the problematic genesis accordingly involves its, and the lapse’s, relativizing to a certain subject other than the author: perspectival sense-making, in brief, complete with narrative energy. A far cry from the purely genetic kind exemplified at the outset, whereby the responsibility for the clash fell straight on the novelist himself—or on ourselves in our everyday ill-performance. Thicker meaning apart, this last option rounds out the exposure of the dead-end fallacy.Nowhere in discoursewill contradictions and lesser absurdities just cancel themselves out into nothingness.Under the geneticmechanism, though, only one of the contradictory terms survives the inference, the other(s) being ousted in repair for all communicative purposes. Either brown eyes or blue, first or revised edition, whichever suits best. But everywhere else the survivors number the whole discordant lot. Even with perspectivized genesis (as with all nongenetic rationales), both terms outlive the newly established order as joint facts or effects. Instead of either being sacrificed to the repatterning, the ill-sorted pair cohere and cooperate to a definite end in its light (e.g., mimesis of stammering, character portrayal via self-betrayal, emplotment of thought in/and action), owing to their very show of inconsistency. Among the alternative mechanisms, the ‘‘existential’’ integrates the discordant items by referring them to some world-frame, novel or established, actual or possible or fictional. A clan listed in Watt includes ‘‘Sam’s other married daughter Kate aged twenty-one years, a fine girl but a bleeder (1 , and her young cousin husband, her uncle Jack’s son Sean aged twenty-one years, a sterling fellow but a bleeder too’’ (Beckett : ). If you have missed the breach of natural law, a footnote ad loc. will ensure your outrage: ‘‘() Haemophilia is, like enlargement of the prostate, an exclusively male disorder. But not in this work,’’ or rather world. Having flaunted the incompatibility between disease and sex, the gloss resolves it ad hoc to maintain and, logically, normalize the presuppositions of existence within a new objective context. The notorious ‘‘The present king of France is bald’’ can always (a fortiori) return to presuppositional normalcy along such lines, 156 Poetics Today 22:1 as can the most far-fetched hypothetical statement. And the objectifying shift may in turn range from the unique or one-off to the conventional, as in the fantastic tradition. More evidently yet, the same holds for internal reality, which even realistic ontologies in life and art often picture as torn by contradictory drives.Thus the age-old topos of loving and hating at once, or Mrs. Dalloway feeling ‘‘very young; at the same time unspeakably old’’ (Woolf  []: ), or the putative unreason of the criminal mind: ‘‘They’ll lay two completely incompatible things side by side and contemplate them with the most unquestioning content. You can’t make them see that they can’t have both’’ (Tey : ). If they fail to ‘‘see that they can’t have both,’’ then we can have both, too, properly mapped onto their cognitive sightlessness as a form, enclave, discourse of existential alterity. Always a matter of operative context, one text’s (or one reading’s) inconsistency is another’s ontic intricacy, if not sheer difference in premises. Moreover, given the endlessness of possible worlds—variants of the actual one included—the frame of reference counting as objective lends itself to infinite adjustment. In or out of presupposition, therefore, it boasts universal applicability to match its rivals for the best fit. Genres (e.g., tragedy versus comedy) are often defined by their peculiar, stylized ontology as well as by their thematics, verbal register, part/whole coherence, and artistic license. Therefore, while genetic, source-oriented necessarily opposes existential integration—recall the example of the color change—the generic principle may well join forces with it in making sense of the discourse. However offensive against the way of the world, possibly all other worlds, the breach will then have its repair in some genre’s unique world of discourse, discourse of the world. Such is the case with the following piece of Carrollian nonsense, where incongruity rises from means to end by generic law, as well as, ontically and sequentially, from the deviant to the impossible: () But four young Oysters hurried up, All eager for the treat: Their coats were brushed, their faces washed, Their shoes were clean and neat— And this was odd, because, you know, They hadn’t any feet. (: ) The whole play of inconsistency about ‘‘their shoes’’ enacts itself on the presuppositional stage, building up to a direct clash between factives. Observe how the process of inference twists forward. The opening reference to the oysters’ ‘‘shoes’’ jars against our normal world scheme. Yet the very presupposition of existence carried by the referential term (and rendered Sternberg • Factives and Perspectives 157 both more casual-looking and more appropriate by the list that casts the wearers as good children groomed for ‘‘the treat’’) implies the normality of the footgear within this peculiar literary frame of reference. No need, as it were, to assert the existence of the shoes—any more than of the preceding (and easily assimilable) coats and faces—only of their propriety for the treat. Just another item in an ascending series of literalized personification, all explicable by appeal to generic combined with existential license that produces the fantasy Wonderworld. Our objections having been anticipated and normalized, however, the next line at once proceeds to revive them through the backglance ‘‘this was odd.’’ Odder, indeed, than when you think of the follow-up ‘‘odd’’ as merely voicing our earlier sense of incongruity with the order of things. For ‘‘was odd’’ is a factive predicate, hence itself presupposing the truth of what it declares incongruous, namely the shoes’ existence (‘‘this’’). By itself, though, the tension between the presupposed and the asserted is yet resolvable without detriment to either’s claim: many true things are odd, surprising, astonishing, incredible.Which is exactly why these double-edged predicates got factivized in the first instance. But if the asserted ‘‘odd’’ can live with the reference’s presupposition and its own, the ground adduced for it cannot. ‘‘Odd, because, you know, /They hadn’t any feet’’ piles factive on factive, then (after the line break) pits affirmative against negative presupposition.The footgear can no longer remain an oddity but must turn outrageous if you know that the wearers hadn’t any feet, nor can you know about the feet’s absence amid the footgear’s odd presence. With the forces in conflict equally matched, the oysters end up having and at the same time not having shoes/feet. A logical absurdity, an impossibility in terms of any well-formed possible world? Of course, but for that very (un)reason acceptable and welcome, indeedmeaningful (along with the twisted route of inference leading up to it) within the discourse world generic to Carrollian nonsense. The family likeness to the writing of Kafka or Beckett, down to the opposition in immediate juxtaposition, should be evident. Not that the yoking together of ontic incompatibles strictly turns on membership in a definite, preexisting genre. Consider how permanent ambiguities dualize worlds (events, characters, or indeed reality keys) otherwise so wide asunder as those of Biblical and Jamesian narrative; or, within James, those of the ghost story and high realism.22Only, the definiteness of the type (e.g., nonsense, paradox) gives its tokens a local habitation and a name, a standing . For details and comparisons, see Sternberg : –, : ff. 158 Poetics Today 22:1 and recognizable license—if not an exclusive imprimatur—which facilitates our dealings with the absurd. A generic world being made for a purpose—like the escalation of incongruity above—functional sense-making necessarily comes into its piecing together. Communicative teleology motivates representational typology. As I argued elsewhere, this accounts even for the largest ontological polarity of all: between the worlds at rest associated with descriptive discourse and the worlds in motion attaching to narrative. For narrativity lies in a determinate set of functions, whose exigencies regulate the play of temporality along the sequence, complete with the appropriately emplotted gaps and discontinuities (Sternberg , : esp. –, , ,  with earlier references). From the integrator’s side, then, the gaps left in the storied reality point and assimilate to the unique forces of storytelling: the three plot dynamics that I call, for short, curiosity, surprise, and suspense. But far from being confined to any domain, generic and/or existential, say, the integration of elements in terms of underlying goal, effect, meansend nexus cuts across discourse levels and ultimately unifies them all into a communicative whole. On the stylistic level, for example, observe how glaring clashes may find their rationale in a well-defined effect: () ‘‘You have said what?’’ she got out, at last, in a voice as silky as a burnt crust of toast. (Chandler : ) () She yawned again. ‘‘You interest me—so little I could hardly tell you.’’ (ibid.: ) () The thing was a life-size stone job of some god or warrior who’d dressed in such a hurry he’d forgotten to. (Lyall : ) () He looked up at me and finally asked in a tragic way, ‘‘Do you think Philly [his wife] will ever speak to me again?’’ ‘‘Oh yes. I imagine so. If only to say goodbye.’’ (Mortimer : ) Assorted-looking, these are all variations on one stylistic device: rather than canceling themselves by saying one thing and then unsaying it, each utterance veers around for surprise.Throughout, the sequence generates the dynamics of a twisted mini-plot: it lures the addressee (internal hearer and/or us readers) into false understanding (expectancy, security) with an eye to a shock of recognition and belated repatterning. This universal surprise functionality even branches out by genre.The abrupt reversal of the simile (from the prefigured silk-like to an anti-silky term of comparison) and of the expression of interest (to an inexpressible lack thereof ) both typify Raymond Chandler’s hard-boiled detective story, where wisecracking passes for wit and toughness. More comic are the two ensuing inversions—from Sternberg • Factives and Perspectives 159 the statue’s dress to undress, from reunion to parting, from metaphorical to literal ‘‘speak,’’ and from sympathy to twisting the knife—both in throwaway style. If downright assertions and their entailments are functionally contradictable, as these reverse for surprise, how much more so with the less directly binding forms of presupposition: () EDWARD was blowing his mind, under the boardwalk. ‘‘Well my mind is blown now. . . . Those cream Corfam shoes clumping overhead. I understand them now, for the first time. Not their molecular structure, in which I am not particularly interested, but their sacredness.Their centrality.They are the center of everything, those shoes. They are it. I know that, now.Too bad it is not worth knowing.Too bad it is not true. It is not even temporarily true. Well, that must mean that my mind is not fully blown.’’ (Barthelme : ) ‘‘Known’’ for a moment to be ‘‘sacred’’ and ‘‘the center of everything,’’ the shoes revert by swift degrees to their old, normal footing. Unlike the earlier quartet, however, this (re)turn against factive presupposition makes by itself functional sense but no (onto)logical whole: it generates surprise, by another twisted mini-plot, without otherwise reconciling the inconsistent terms. Nor would it (onto)logically reconcile them even if you compound generic or thematic with affective function, overall with sequential effect, and have the discourse bestow an intrinsic value on incongruity, nonsense style. How would attributing two ends to the reversal eliminate or decide the antithesis between the shoes’ sacredness and ordinariness, centrality and marginality? The breach of factive commitment may then double its gains, yet always short of healing. Unhealable by any of the rationales so far examined, the breach still lends itself to closure in terms of the perspectival mechanism. Actually, the entire quoted passage is flanked by the apparent incompatibles, ‘‘my mind is fully blown now. . . . my mind is not fully blown,’’ which together capture and foreground and enclose a peculiar subjectivity at work. Thus the two head-on clashes, the innermost one (‘‘I know’’ as against ‘‘not true’’) and the outermost one, explain each other by reference to a single (il)logic, even a single trajectory: the zigzag movement of a psyche under the influence of drugs. Aristotle, if equal to the culture gap and the low life, might identify it as peripety cum rediscovery and self-discovery, with appropriate surprises along the way. ‘‘Blown’’ for a time, perhaps only wishfully so, Edward’s mind does or would experience (‘‘know’’) the transformation of the shoes; yet, ‘‘not fully blown’’ after all, the reality principle gains control over his mind again 160 Poetics Today 22:1 and negates by degrees the hallucinatory insight into things.The imagined ‘‘knowledge’’ first loses its value (‘‘not worth knowing’’), then its validity (‘‘not true’’), then its validity for the duration (‘‘not even temporarily true’’): a process, not a point, of return to humdrum existence. In face, in place, even in light of an apparently unstable, contradictory ontology, we trace a dynamic viewpoint on it whereby to motivate (i.e., unify) its shifting images and artful operations through a low epistemic, (re)cognitive adventure in progress; instead of a broken, there emerges a delusive factivity, withdrawn by the subject once he comes to his senses. As the mind’s subjective experience comes full circle, so does our outline of sense-making logics. For () loops back to the stammers followed by corrective touches, even ordeals that were vocally or wordlessly enacted as early as ()–(), exemplifying perspectivized (hence narrativized) genesis at work in less unusual circumstances. And yet, however striking the likeness, it extends much further, because instances of perspectival integration share a wider common denominator than self-repair. Even at its most discernible, this family likeness need not run to a whole trial-and-error process undergone by the subject. It just involves some error (e.g., factivizing an untruth) that we coherence-seekers attribute, and thus relativize, to some fallible subject—as distinct from the appropriate higherorder subject (quoter, narrator, author, e.g., Barthelme in [], Spark in []) taken to know better, with ourselves, and counting as objective.The latter, we then infer, has the former’s perspective expose itself in and through its deviant subjectivity, with orwithout self-correction,more overtly or less, for artistic or ordinary pragmatic ends. Of these, needless to say, the commonest is irony, built into the discrepancy in viewpoint, epistemic or otherwise. Nor indeed does the launching of such operation require so much as the surfacing, or suspicion, of error. Any gap, blur, excess, oddity, dissonance, ill-formedness, peculiar turn of phrase ormindwill do for trigger. (Whether it justifies perspectivizing, even whether it should trigger a quest for integration at all, is another, empirical matter, outside the scope of discourse theory.) Only, the less salient the clues to perspectival inference (compare [] or [] with []), the more does some rival logic suggest itself—or none whatever, as is usual in dead-end presuppositional orthodoxy. In illustrating a research program where every dead end counts as a beginning for endless inferential movements toward sense, however, this quick outline also generates (to some, would deepen) a doubt about the particular movement in question here, indeed about its very particularity. Or, at least, it highlights a desideratum regarding the argument’s sequel. As already generalized and now, I hope, brought out, ‘‘integration’’ encompasses Sternberg • Factives and Perspectives 161 the whole repertoire of ordering resources and sense-making combinations available to humans. This follows from the Law of Reciprocity, since the ability to explain whatever is expressed, or expressible, hinges on one’s being as fully equipped at the receiving (‘‘integrating’’) as at the transmitting (‘‘disintegrative’’) end. So, were integration at large our topic, then stretching and variegating and filling out the range chosen here would now be demanded, along the lines telegraphically ennumerated at the outset. One might also proceed to concentrate on those lines, the way I have done elsewhere on integration of and via time, space, analogy, narrative functionality, quoted discourse, stereotyping, for example, or Yacobi on (un)reliable telling. As it is, the set of mechanisms just picked out rather urge the need for finer discrimination; only thereby may we suit the inferential procedure to the inference type under analysis—especially when gone wrong on the surface and left for dead by theory. Those five integrational logics are ever-available, variously combinable, extendible all the way from word to work, applicable to each and every trouble spot in discourse. By the same token, however, their very universality must restrict their differential power, as crosscutting will always blunt the cutting edge. Left at that, they will hardly capture the specialness of any special case—like presuppositional trouble and resolution, for example— but will instead group it with all others: under the heading of genesis, existence, genre, and so forth. We may want or need to invoke them for that very reason, of course. How to make tolerable sense of the clash of factives in the nonsense example without appealing to high-level generic purpose? But then, how to discriminate that (non)sense-making and its beneficiaries from their assorted factiveless (even presuppositionless) equivalents in the same nonsense text and text family, never mind discourse at large? The play of perspectives, I will soon argue, offers the key to the redefinition, understanding, and, as necessary, unraveling of factivity, hence to troubled presuppositional inference in other shapes. Let me therefore emphasize that, unless duly adapted to the specific task, this play is no exception to the rule. At the high level of generality so far maintained, the imbalance between the inclusiveness and the differentiality of resolution shows itself in the perspectivalmechanism as elsewhere. Like the rest of themaster keys, it unlocks too many doors (here, distances too many infelicities away from the author) to single out and identify for us any particular one. For example, compare () with breakdowns of alternative inference patterns, notably entailment and implicature: () My beard is resented because my philosophy is resented. Philosophers wear beards; Julian wears a beard; therefore Julian is a phi162 Poetics Today 22:1 losopher andmay share with that subversive tribe sentiments hostile to the superstitions of the Galileans. (Vidal : ) () This wasMr.Hackett’s attitude towards things that pleased him.He knew they were not his, but he thought of them as his. He knew they were not his, because they pleased him. (Beckett : ) () ‘‘It’s long,’’ said the Knight, but it’s very, very beautiful. Everybody that hears me sing it—either it brings the tears into their eyes, or else—’’ ‘‘Or else what?’’ said Alice, for theKnight hadmade a sudden pause. ‘‘Or else it doesn’t, you know.’’ (Carroll : ) The syllogism in () boasts standard form and order: major premise, minor premise, conclusion. Yet the argument is evidently invalid, since the conclusion about Julian’s being a philosopher (much less the rider vilifying him by association) is not deducible from the premises. Or, the other way round, they fail to entail it. What exactly has gone wrong? The simplest label is the fallacy of accident: even if all members of the philosopher class happen to wear beards, sharing the inessential attribute of beardedness will not yet make anyone (e.g., Julian) a philosopher.Hence the deductive chain breaks, regardless of the truth or falsity of the constituent propositions. Q.E.D., the logician would conclude, as will the logical detector in us readers and text analysts. As such, though, we must proceed further, the demand for our own logics arising exactly here. To us, the real question is not where but who the breach of deductive argument comes from, and why.The non sequitur, we go on to infer in our turn, is perpetrated against rather than by the overt teller in this historical fiction, the emperor Julian. If anything, he cites it to explain and deride another non sequitur, that is, the putative ‘‘resentful’’ causal linkage between his beard and his philosophy. All along, he elliptically argues, the one term issues from the other— the effect from the ‘‘because’’ of resentment, the syllogistic consequent from the antecedents, the latter from the former chain—only in the mouths, and perhaps the minds, of his Christian enemies, ‘‘the Galileans.’’ The false entailment sequence is theirs accordingly, virtually twisting forward between inverted commas. They offend against reason to destabilize a philosopher-king ‘‘hostile’’ to their ‘‘superstitions’’; all the more freely so because the art of reasoning (like philosophy and rationality as a whole) stands for the pagan Greco-Roman culture to which the so-called apostate would restore the Empire.Whether trained or dabbling in the hateful art, they never scruple to wield it in their holy propaganda war.Theirs, in short, are not logical but psycho-logical rules applied with theo-logical license. His ‘‘superstition’’ being the true faith to them, his ‘‘philosophy’’ their ‘‘subSternberg • Factives and Perspectives 163 versive tribe,’’ his ‘‘Galilean’’ their Christ, his gods their idolatry, the ends of anti-Julian rhetoric justify the means. Conversely, so far from sharing the train of fallacies, the philosopher king wants his reader to detect and savor them, in the hope that the irrationality in rational form will boomerang on its originators and publishers. Little wonder he so followably (‘‘because,’’ then the syllogism, complete with ‘‘therefore’’) orders the aspersions cast on him into inferential or argumentative sequences that do not follow. He indeed beats the Christians at their supposed game of adopting the other’s abominable weapons. To this extent, at least, in turning antagonistic words, thoughts, and ploys against the viewpoint that has generated them, Julian reveals himself an heir to ‘‘that subversive tribe.’’ By the end of the inferential process, then, we find (i.e., arrange) two minds and two drives in scalar opposition—two warring perspectives contextualized as high versus low, ironist versus ironized—below the surface of one illogical discourse.We figure out not only where and who the unreason comes from but also why. A double ‘‘why,’’ actually: why incurred by the original subject(s) and why transmitted to us by the rational narrator of the, and their, discourse. The basic explanatory principle carries over to (), with four main variants. Here we reconstruct afresh two minds in hierarchical opposition, yet not those of two speakers: a subject’s thought (‘‘Hackett’s attitude toward things’’), rather, gets exposed by an other-minded teller. Moreover, the exposure is gradual and tortuous. It rises from the strangeness of the ‘‘but’’ (a jealous or a self-deluding psyche? we wonder) to the non sequitur of the ‘‘because’’ (whereby either inference about Hackett’s psyche must undergo drastic revision to suit with the newly revealed irrationality). In turn, however mind-boggling, the non sequitur shifts offenses from deductive (il)logic—an entailment failure, as in the Julian syllogism—to (im)probability: it takes an abnormalmind to forge a causal chain that grounds awareness of nonpossession in pleasure. Last, the chain starts—or, as befits the twisting ascent of the insideview, restarts—with a factive, ‘‘He knew they were not his, because. . . .’’ The ensuing non sequitur therefore drives the reader back to question the factive statement itself. Are we to understand that Hackett really knew it or that he only imagined he knew? Does the narrator dissociate himself from the subject’s mental causation as a whole, or withdraw commitment from the effect alone, subjectifying (relativizing, ironizing) the ostensible knowledge of nonpossession?Whatever one makes of this ambiguity—and we will come back to the issue—the resolutions hardly vary in essentials. They in164 Poetics Today 22:1 volve constructing some perspectival duality to accommodate a series of dissonances, odd factivity among them: not even at their head in either text order or scale of importance. But downright unreason itself coexists with other triggers for perspectival removal, as for integration in general.Thus () manifests neither a breakdown of entailment nor even an improbability. If anything, the disjunctive proposition that the song ‘‘either . . . brings the tears into their [the auditors’] eyes, or else. . . . it doesn’t’’—with factivity (‘‘you know’’) again thrown in— is all too logical: so much so as to become necessarily, ‘‘analytically’’ true, hence tautological. For once, therefore, the surface failure in coherence has nothing to do with any textbook logic but is purely discoursive and accordingly referable, as illogic, to another rationale altogether. The discourse here changes offenses against meaningfulness from the polar extreme illustrated by the earlier cases, namely inconsistency, to another extreme, that of redundancy.23Yet the informational value of either may still appear to remain nil: as the first ostensibly contradicts itself to leave nothing, so the second predicts itself to add nothing. Necessarily true in all possible worlds, including Wonderland, how can the tautology be otherwise than uninformative? The opposed routes—semantic as well as disciplinary—would then lead to the same dead end. Our example eludes sense-making even on the best known approach to inference in pragmatics, Grice’s ‘‘Logic and Conversation’’ (). According to it, the disjunctionwould come under infringements of the firstmaxim of Quantity, ‘‘Make your contribution [to the talk exchange] as informative as is required’’ (), and so possibly generate some implicature. I for one find Gricean implicature, along with the maxims and the Cooperative Principle behind it all, much too ill-defined to establish an inference type on a parwith entailment or presupposition—or to cover the entire range of our hypothesis making throughout discourse. But let’s assume the machinery for the sake of argument. In its terms, such nonconformity to the maxim as the White Knight’s opens four possibilities (Grice : ). But none of them applies to him, because all presume the speaker’s control and the reference (in effect, the smooth assimilation) of the noncompliance to his own intentionality. Speaker-based variants of functional reasoning, you might call them from . I have elsewhere written on how these two extremes (in the shape of gaps or ambiguity versus overtelling) join offenses, and ultimately forces, even in genres where you would least expect it: narrative with a strong truth claim and the discourse of the law. See Sternberg : esp. –,  (esp. Index under Gaps and gap-filling, Redundancy), and, with special regard to factivity, . Sternberg • Factives and Perspectives 165 our higher ground. And that presumption of the utterer’s choosing among meaningful alternatives fails here, as often elsewhere, all along the line. First, the Knight does not ‘‘quietly and unostentatiously violate amaxim,’’ least of all with a ‘‘misleading’’ intention or outcome. Nor, second, does he ‘‘opt out’’ from the rules, if only because he ‘‘contributes’’ to the best of his ability. Nor, third, is he ‘‘faced by a clash’’ between the demand for informativeness and some other maxim. Nor—least of all—does he ‘‘flout’’ the maxim of Quantity itself, ‘‘blatantly’’ going against it with a view to generating an implicature. The poor Knight is as little capable of such devious exploitation as of hidden misdirection; and even if he were, neither ploy would explain his aberrant discourse, to Alice or to ourselves. By a new route, we arrive at the usual dead end. An inference (call it implicature if you will, albeit by default only) does await computation here.Yet the lines between which it lurks and the logics whereby it operates or emerges are undreamt of in Grice’s philosophy as such, because radically other and wider than speaker-oriented. I refer not just to the compounded generic and functional integration, whereby all nonsense makes sense in Carroll, or Beckett, but also specifically to a fresh perspectival variation. The Knight’s ‘‘sudden pause’’ between the two disjuncts, underlined by Alice’s nudging for closure, evidently heightens expectancy to redouble the anti-climactic effect of the (dis)closure. But does theKnight himself maneuver for this effect on Alice, or is he maneuvered by Carroll into springing it on us readers? Either fork would result in an insideview of his exigency, character, subjectivity. If himself the maneuverer, to Alice’s discomfiture, he becomes Hatterlike in antisocial perversity, indeed uncooperativeness: a reading inconsistent with his behavior as champion and dialogist at once. If (far more probably) himself maneuvered by Carroll the author, then the disjunctive nonsense effect has its motivation in another perspectival inference, casting him as a benevolent bungler happily transported to a discourse frame beyond his ken. The abrupt pause now reflects the speaker’s bafflement. He has no idea how to finish and, under Alice’s pressure, comes up with the most ‘‘logical,’’ tautological disjunct: instead of adding anything new about the song, it betrays the predicament of the singer to-be as a dialogist and, more obliquely yet, his effectiveness as a fictional creature on this very ground. A piece of arrant nonsense in lifelike, first-order terms, the Knight’s is (re)staged within and for a mimetic art of nonsense. If Carroll specializes in communication trouble, all the way to breakdown, it is less his pushing the universal ills to an extreme than his trading on them at the characters’ expense that makes the difference from the 166 Poetics Today 22:1 immediacy of talk in the real world. Were a similar mishap to befall any of us, our conversational partner’s inference of it would, certainly might locate what has gone wrong in our mind, without proceeding to make it right by appeal to some wirepuller behind the scenes.Yet the giveaway figured out in the process remains a giveaway, and as such informative about its utterance-origin, hence integrative andmeaningful, though hardly communicative, because unintended and unfunctional. It takes fiction or more generally quotation, with its double originator, to make the giveaway communicative, too, on a level higher than the immediate, other than the illperformance born of ‘‘genesis.’’ As always in genetic reading—and never in Grice et al.—the inference maps the troublesome discourse givens onto the speaker’s self-exposure rather than self-expression; the latter, if any, further entails the vantage point of the artful exposer in control of the whole show—ironic perspectivizing, that is. But then, this very open-ended range of applicability shows the principle that I invoked to be too strong for our needs here, a counterpart of the weak logical and pragmatic alternatives entrenched in the field. If all is grist that comes to the perspectival mill—all inference types among the rest of comers—wemight as well acquiesce in themainstream lines of analysis that would reduce (factive) presupposition to entailment and/or the mixed bag of implicature.24 This alignment might even plug various glaring holes in either reductive approach. Strategically, I have just exemplified how to transcend the narrowness of speakerand/or intentionality-based inferencing. More to the immediate crux, take an apparent presupposition failure, like inconsistency within a knowledge statement. Confronted with it, our array of mechanisms will generate sense—indeed, an embarrassment of riches, as above—where the unitary entailment analysis legislates a dead end and implicature theory lacks the power to break through on its own ground or ‘‘deductive device.’’ But the advance in integrability would then proceed on such a wide front as to merge presupposition afresh with neighbouring kinds of inference, even less extricably, if anything. To have the best of integrative worlds—and to prove either assimilation wrong—our universal resources must now be specified in keeping with the crux itself, the toolbox geared to the target and the task. Having shown the various ways of rescuing factive inconsistency from sheer dead-end nullity, that is, we must go on to show how, why, and to what effect factivity can . Among these attempts, already exemplified above, the most piquant is Grice : – : in a study dedicated to Strawson, Grice argues for the assimilability of the dedicatee’s inference type to his own implicature and at a pinch to Russell-style entailment. The notes on factivity there (ibid.: –), by the way, are particularly weak. Sternberg • Factives and Perspectives 167 be saved from the inconsistency that ostensibly threatens its very defining attribute: from the breach of commitment to the truth of the proposition expressed by the subordinate clause. Does the coherence we figured out between ‘‘know’’ and ‘‘not true’’ in (), say, illustrate the discoursewide principle that applies everywhere or a subdivision of it adapted to the particular bond-violating incoherence arisen there? If the latter, what becomes of the analogous factive incoherences otherwise settled in ()? And what of the other, nonfactive incoherences analogously settled in ()–()? The answer requires a closer look at the distinctive properties of factivity. 3. Factivity as Quotation The key to factivity—by extension, also to a host of presuppositional issues and triggers—lies in the distinctive perspectival play built into it.This play derives in turn, as I have long argued (, b, ), from its quotational nature, with component viewpoints to match, especially the quoter’s and, or against, the quotee’s.This perspectival interplay, I would further argue, subsumes, interrelates, and dynamizes all orientational parameters— including those usually lumped together under ‘‘context’’—so as to enable a principled yet flexible account, and where necessary, resolution, of factivity’s workings. But let me develop the arguments in order. A verb like ‘‘know’’ entails quotation: discourse about discourse, one subject’s report of another’s utterance or, most evidently here, thought (e.g., knowledge state). It therefore shares, and indeed richly manifests, all the universals that define this protean form of representing (or, better, representing) theworld at a remove. But then, so do its nonfactive equivalents, ‘‘believe’’ for example.The peculiarity of ‘‘know’’ is accordingly found less in than within quotation, whose high common denominator does not yet suffice to tell the verb-types apart. For distinctiveness, we need to align the factive’s specifics as a variety of second-order, bifocal discourse and as a presuppositional trigger (commitment, inference). In this section, we will explore the former crux: why quotation attaches to the verb-types at issue, how it maps itself on them, and where the presuppositional variable must enter to draw the line. Though a first step, it should bring out the rudiments generally missed as well as clear some misunderstandings out of the way.25 Compare two simple (hence, for once,made-up) examples, opposedwith regard to factivity: . Including the strange obstacles to the very treatment of such obvious mental predicates as reportive or as hallmarks of indirect report: see note  below. 168 Poetics Today 22:1 (a) We faced each other in silence.He knew that I had got the evidence. (b) We faced each other in silence. He believed that I had got the evidence. In either case, the discourse traceablymoves from first to second level, from speaking proper to quoting. It goes from the self (‘‘I’’) representing theworld in his own name, with us hearers/readers as immediate addressees, to the self penetrating the mind of another (‘‘he’’) to mediate and thus re-present what this other has supposedly already represented (‘‘knew,’’ ‘‘believed’’) to himself, namely: the evidence gathering done by his former interlocutor and current quoter. Broken down into more rigorous terms, either movement exhibits and traverses in due sequence all the three necessary elements of quotation.26 What may be called the overall frame of discourse that the speaking-I shares with his own audience (the implicit ‘‘you,’’ here ourselves) leads the way. Shares, that is, in the double sense of commonality and communicativeness, as exemplified by the opening stage direction. There ensues the reporting clause with its cognition (or, elsewhere, utterance) verb, or the transformer. Via this intermediary, self-expression (still oriented to the frame on the deictic axes of time, now/then, and person, ‘‘I’’/‘‘he’’) modulates into quotation (bearing on the third-person knower/believer). Finally comes the quote itself, or inset, which purports to transmit the other subject’s original thought, what ‘‘he’’ knew/believed. Here our addressor has altogether shifted roles and voices from speaker to quoter in (re)imaging for us the quotee’s mental image of their interrelation at the time. Thus far, the examples run parallel. The parallel extends to the choice made out of the quotational repertoire, that is, to the surface form assumed by the deeper must’s of ‘‘transformer’’ and ‘‘inset.’’ Both examples combine the syntactic embedding and the deictic one-centeredness that typify the quoting form of indirect discourse. Observe how the transformer/inset nexus surfaces, or doubles, as a passage from main clause to object, that-clause, and how the anchorage in the ‘‘I’’ (established as early as the frame) persists throughout.27 Con. Necessary, I mean, to the very constitution (hence, reversely, understanding) of quotational discourse as such, yet not necessarily as manifest (i.e., actualized) on the surface as in indirect discourse. E.g., both the direct and the free indirect schemas may dispense with an overt quoting context (‘‘frame’’) and/or quoting clause (‘‘transformer’’), which we then mentally supply to (re)constitute the appropriate schemas.The quoted utterance or thought (‘‘inset’’) alone demandsmaterial surfacing, if only for the reason that it can trigger the repair of its hidden fellow elements, but not vice versa. . Likewise with the alternative set of main clause predicates, which take a subject rather than an object complement, e.g., ‘‘It mattered/seemed to him that I had got the evidence.’’ Sternberg • Factives and Perspectives 169 versely, among the same quoting repertoire, neither examplemight in principle opt instead for the direct form. ‘‘He knew/believed: ‘You have got the evidence’ ’’ jars at such a profound level as to deter even literary experimenters. Directness, where transformer and inset enjoy both syntactic and deictic independence of each other—two sentences, two zero points— is unavailable to these mind-quotations. Nor to them only, since the rule constrains nearly all mental predicates, the philosopher’s so-called verbs of propositional attitude. Among the likewise constrained are suppose, realize, observe, forget (with think the odd verb out). However you name them, the limitation visibly and literally raises their common denominator—as high as the surface form.What with the unconstrained utterance verbs (say, exclaim) at the other pole, ranging at will between direct and nondirect style, the family likeness widens in (a)–(b) from the matter to the manner of re-presentation: from inside-viewing transformer (‘‘He knew/believed’’) to twice subordinated inset.28 The new light thrown on our two paradigm cases is already easily enough measurable.The key of quotation once supplied, the traditional balance of equivalence and difference between them undergoes a sea change, even on the fundamental level of analysis, along with the analytic tools themselves. Across the line of factivity, the overlap multiplies in extent and areas while the discrepancy shrinks on further reanalysis. As to the overlap, it far outreaches the narrow correspondence—in ‘‘complementation,’’ ‘‘semantic representation,’’ ‘‘propositional attitude,’’ and the like—drawn by earlier theory, startingwithKiparsky andKiparsky’s classic ‘‘Fact’’ (), the most intensive and influential work on the topic. Before showing why their analysis fails, let me emphasize the comparative reasonableness of their general assumptions, intent on differentiating, for better or worse, where successors have produced a mishmash. The Kiparskys never confuse (factive) presupposition with entailment and/or implicature, but rather want to mark it off from nonfactivity (e.g., ‘‘He believed that . . .’’ in [b]) that takes analogous surface forms.They locate the difference in ‘‘the speaker’s commitment’’ to the truth of what the embedded clause expresses (an echo of Strawson on existential presupposition) rather than in appropriateness to some muddled ‘‘context.’’ Nor would they dream of severing factive presuppositionality from its anchorage in the language—the factive Here the transformer still controls the that-inset, across the variation in the latter’s grammatical role, as well as across the factive/nonfactive boundary. . The exceptional freedom of think probably derives from the immemorial modeling of thought on speech, if not vocal utterance: the soul at self-discourse. 170 Poetics Today 22:1 main-clause predicate, or ‘‘trigger.’’ On the contrary, the trouble is that they would overconnect, attempting to package the distinctive, commitmentbearing lexis with the grammar, the higher-clause verb’s sense with the overall syntax: as if the factive type were sui generis on two levels at once, an ideal linguistic match. In their own words, they undertake ‘‘to explore the interrelationship of syntax and semantics in the English complement system. Our thesis is that the choice of complement-type is in large measure predictable from presupposition’’ as a ‘‘semantic’’ variable, so that ‘‘a series of systematic differences’’ reveals itself ‘‘between the factive and non-factive predicates’’ (ibid.: ). What actually reveals itself is the defeat of the matchmakers’ interlevel undertaking: their very limited success amounts to a boomerang effect. In overpolarizing the ‘‘types,’’ the neater, fuller the twofold binarism the Kiparskys would impose, the more numerous the actual intertype continuities that emerge—syntactic and semantic, deep and superficial, built-in and optional, given and resistant to the linguist’s idea of systematizing. In turn, I will argue, the continuities pinpoint the real divergence operative on the factive/nonfactive axis. That Kiparsky andKiparsky do glance at the phenomenon of quotation, only to exclude it from the issue under analysis, highlights the oversight. For they prelimit the term to ‘‘direct quotes’’—associable with neither ‘‘believe’’ nor ‘‘know’’—and byway of false transformational conjecture at that: ‘‘what one quotes are surface structures and not deep structures’’ ().29 So factives and nonfactives are denied a priori the attributes, resources, indeed the very name of quoting, along with all surface-preserving value. Both alike remain mere ‘‘semantic representations’’ expressed within ‘‘the complement system’’ (): a tag as little unifying and as little differential as the philosopher’s bracketing the two predicates under ‘‘propositional attitudes.’’ An aliveness to the multiple forms of quoting radically changes the picture, indeed the paradigm. It at once brings out the marked common denominator of representing as re-presenting—semantically as otherwise— and apropos a pair like ‘‘believe/know,’’ the still higher commonality of re-presenting thought.This in turn has constructional implications, among others. Either re-presented thought necessarily assumes some frametransformer-inset shape peculiar to quotational, second-order discourse. . On the underlying direct speech fallacy, as I call it, see the arguments and evidence in Sternberg , a, b, : –, , now widely accepted. If you wonder why the obvious-looking factivity/indirectness tie still needs to be argued and differentiated— against continuing opposition, maybe—such attachment to ready-made disciplinary labels yields at least part of the answer. Sternberg • Factives and Perspectives 171 That constructional peculiarity sharpens further in the movement toward the actual discourse as produced and encountered, given the categorical absence of directness from the quotational repertoire elsewhere open to the quoter.Within the common indirectness of the thought re-presented in the ‘‘believe/know’’ examples, the underlying quotation-wide construction evenmaps itself onto a determinate surface structure, identifiable by the inset’s twofold subordinacy to the transformer: syntactic (as that-clause) and deictic (as frame-oriented, or ‘‘backshifted,’’ in time, space, and, above all, person). Prior to such low-level specificity, the issue of clausal, let alone object (‘‘that . . .’’), complement never arises at all here—literally does not break surface—and even so it arises in amarked realm, company, and role: among other features of indirectness (the equally specific deixis included) that advertise and enform the multifold family likeness between the ‘‘know’’ and the ‘‘believe’’ sets. Nowhere outside indirect re-presentation is the object clause found, nowhere else distinguishable as such from identical-looking that’s embedded in sentences.30 The embedding of the inset (i.e., the represented speech or, here, thought) in the form of object clause exhibits a cross of syntax, deixis, and mimesis unique to indirect reporting but not to factive indirectness: on the contrary, the transformer in the higher clause may equally draw its verb from the nonfactive set. The cross makes, or makes over, a bond rather than a barrier. For now, this advance fromdepth to surface, from themimetic to the constructional to the formal properties, should be enough to clinch the point. In uncovering, interlinking, and explaining all these properties, the approach via quotation already establishes its superiority to the traditional Kiparskytype line of analysis. So relocated and reconceived, the factive/nonfactive continuity thickens, sharpens, defines itself to an extent and on levels beyond the reach of such blanket-terms as ‘‘complementation,’’ ‘‘semantic representation,’’ or ‘‘propositional attitude.’’ Inversely with the areas of syntactic difference that would separate our two examples according toKiparsky andKiparsky.These nowdramatically shrink, because the criteria for delimitation will not bear scrutiny. Here, of . One must guard against overextending indirect report to all sentences with complement clauses, which would drag in the nonreportive ‘‘It is a fact that . . .’’ and the like. On the other hand, and with even less reason, Cohn (, taken up in Fludernik ) would exclude most such sentences from indirectness: she wants to associate the linguistic form with a certain representational function.This arbitrary package deal would in effect consign factive reports (inter alia) to the miscellany of Cohn’s ‘‘psycho-narration.’’ It also worsens and, as it were, vindicates the prevalent tendency to diminish the extent and the art of indirectness, factive or nonfactive, among the forms of thought report. 172 Poetics Today 22:1 course, we move down to the next lowest levels of specificity: those below clausal complementation, the forking into object versus subject clause included (see note ). Even at such levels, though, the presupposition-laden and the presuppositionless sentences turn out to elude formal dichotomizing. In other words, they resist the attempts to correlate semantic with syntactic difference into a fixed package deal. The resistance, total or partial, shows all along the line that the linguists would draw; unity replaces, or at least moderates and complicates, the alleged divergence. The respective correlations, hence the overall antithesis, prove on scrutiny to be weak, double-edged, never univalent. Take the attempt tomark off factive from nonfactive constructions by the placement of the embedded clause (my inset) between the beginning and the end of the sentence (relative to my transformer).With subject clauses, accordingly, ‘‘extraposition [end placement] is optional for factives,’’ while ‘‘obligatory’’ for nonfactives. Thus, That John has come makes sense (factive) *That John has come seems (non-factive) where the second sentence must become It seems that John has come. (Kiparsky and Kiparsky : , ) Must it always, and by an iron rule of syntactic/semantic correlation? Just replace ‘‘seems’’ with the equally nonfactive ‘‘is possible,’’ or even ‘‘seems possible,’’ and the subject complement will enjoy the same positional mobility between limits as does the factive equivalent.The attempt to generalize the rule of polar (im)mobility from subject to object complements—our main business—likewise fails with a vengeance.Needless to say, ‘‘That John had comewas widely known/believed’’ is as grammatical across the semantic divide as the extraposed, sentence-terminal that-form, complementing ‘‘It was widely known/believed. . . .’’ But Kiparsky and Kiparsky actually drive a still more ambitious claim, one that postulates a ‘‘general tendency for sentence-initial clauses to get understood factively’’ or ‘‘associated with a factive sense.’’ The complement’s initial positioning would then by itself exercise such metamorphic power as to factivize a nonfactive. In saying The UPI reported that Smith had arrived It was reported by the UPI that Smith had arrived the speaker takes no stand on the truth of the report. But That Smith had arrived was reported by the UPI normally conveys the meaning that the speaker assumes the report to be true. A non-factive interpretation of this sentence can be teased out in various ways, for Sternberg • Factives and Perspectives 173 example by laying contrastive stress on the agent phrase (by the UPI, not the AP). Still, the unforced sense is definitely factive. (Kiparsky andKiparsky : ; also, Horn : ; Seuren ; Grundy : ) The line drawn here between the positional variants, in terms of the speaker’s stand, blurs even regarding the case in point on the given argument. If the that-initial variant admits of an alternative, noncommitted reading—with ‘‘contrastive stress’’—the desired rule to the opposite effect already weakens accordingly. Every utterance being understood as well as produced in context, moreover, whence the ‘‘normality’’ or ‘‘unforcedness’’ of the factive sense? Why privilege, or encode, one intonation contour over another? And beyond the tell-tale concession, once you substitute other nonfactives (‘‘held,’’ say, or ‘‘denied’’ or indeed ‘‘believed’’) for ‘‘reported,’’ the alleged forking by placement evaporates altogether. So do all other attempted package deals between grammar and semantics, (linguistic) form and (discoursive) force. Interestingly, on the Kiparsky list of factive predicates that take object clauses, there appears ‘‘be aware (of )’’ but not the more common and pregnant near-synonym ‘‘know’’ (). Why omit the epistemic exemplar, with its manifest commitment to the truth as well as interdisciplinary network? The reason for this selectivity gradually emerges from the ensuing notes or silences on know’s syntactic intractability, which goes a longway towardmarring the neat twofold ‘‘division into factive and non-factive predicates’’ that comes with the list. Among these divisive rules, for example, the most central or thematic one alleges: ‘‘Only factive predicates can have as their objects the noun fact with a gerund or that-clause’’ (ibid.: ). But then, while ‘‘He was aware of the fact that I had got the evidence’’ is doubtless grammatical, ‘‘He knew [or ‘realized,’ or the possible factive ‘anticipated,’ or the counterfactive ‘pretended’ or ‘wished’] the fact that . . .’’ is not, anymore than is the nonfactive ‘‘He believed the fact that. . . .’’ (As though to multiply confusion, ‘‘know’’ displays the extra peculiarity of allowing instead the adverbial, rather than objective, sequel ‘‘for [or, as] a fact.’’) Kiparsky andKiparsky come to admit as much, if only in the question-begging argument that such ‘‘exceptions’’ are ‘‘semantically factive’’ yet ‘‘syntactically non-factive’’ ( n. A). Nor do they ever account for this exceptionality or acknowledge its range and seriousness.31 . The same holds for other instabilities they admit, some not yetmentioned: e.g., regarding the use of gerunds () or of the accusative and infinitive construction () or of the subjunctive (). In another, semiotic framework, Eco and Violi () likewise find know ‘‘very embarrassing,’’ though for a different reason: as a ‘‘verb of propositional attitude,’’ along with be aware of and believe. Such is the embarrassment that they treat it under this larger, nondis174 Poetics Today 22:1 These omissions vitiate their fundamental explanatory hypothesis, whereby the polar syntactic/semantic correlations alleged issue from the respective depths.Though factive and nonfactive sentencesmay look superficially alike, ‘‘presupposition of complements is reflected in their syntactic deep structure.’’ At the deeper level, factivity would be appropriately, indeed literally represented in the noun heading the object clause—for example, ‘‘I regret [I’m aware of ] the fact that John arrived’’—and preservable or deletable in the movement to the surface form (ibid.: ff.). But how would the putative base-structure generate a ‘‘know’’ either mandatorily doing without ‘‘the fact’’ as immediate object or reinforced with ‘‘for a fact’’ as a nonobjective complement? 32Or, more generally and damagingly still, the other way round: the allegedly factive-specific deep (and, at will, surface) head noun may consist in ‘‘the possibility’’ as well as in ‘‘the fact’’ and will then extend to nonfactives.This oblivion to the entire range of modal choices is impossible to patch up. Most damagingly, within the factive set itself, how do we reconcile the hypothesized deep/surface ‘‘the fact’’ known, as it were (or ignored, deplored, resented), with its nonfactuality to the speaker in examples like ()–() above? If such factive usages refute all theories to date, here they lay open a contradiction in terms, because of the extra commitment to truth falsely trumpeted by the definitional (on the hypothesis) and otherwise superfluous head noun. This last disproof outreaches the horizons of the Kiparsky approach, which never envisages presupposition failure (or implicit ‘‘cancelation’’/ ‘‘defeasibility’’). Within its limits, a commitment is a commitment, and therefore needs formalizing vis-à-vis the noncommittal sentence, rather than explaining, let alone in apparent breach. But then, were the movement between bond and breach taken into account, what would become of the sought factive/nonfactive polarity?What would justify the anchorage in the formation of decontextualized sentences rather than in context-bound statements—quoted, hence twice-bound statements about states of mind at that—where alone pledges take or lose effect? The theoretical implications ensue to reinforce our earlier analysis. A discourse matter par excellence, how a factive arises, works, plays, channels inference, stands to its opposite number, and so forth, are issues that transcend the linguist’s along with the logician’s antidiscoursive occupatinctive rubric (–), apart from that of ‘‘factive verbs’’ such as regret (–), by appeal to the desperate argument that the propositional variety ‘‘transcends the normal use of natural language’’ (). . Nor does the incorporation of ‘‘the fact’’ within Montague grammar escape trouble: see Van der Sandt :– on Delacruz . Sternberg • Factives and Perspectives 175 tion, beginning with their alien idea of how givens reduce to a system, even what givens to what systems. The binary linguistic formalism having predictably collapsed, we may now explore the commissive, presuppositional force activated by factivity as a branch of second-order discourse. 4.Whose Commitment toWhat? Perspectivizing Factivity How exactly does factive differ from nonfactive in the commitment ‘‘to the proposition expressed by the subordinate sentence’’? The answer has generally seemed too obvious, or unitary, to be worth pursuing. Even apropos the classic ‘‘know/believe’’ exemplars, one must throw a bridge over otherwise removed fields to locate the difference in both the degree and the durability of the commitment.33 Return to () and you will see how this juncture already compounds differentials. As factive, ‘‘know’’ encodes a truth bond so strong that any attendant denial of the presupposed knowledge statement (e.g., ‘‘. . . but I hadn’t in fact got the evidence’’) incurs absurdity. It also uniquely survives the negation (or questioning) of the predicate in the main clause itself (‘‘He didn’t know . . .’’). The ‘‘believe’’ counterpart, on the other hand, remains deniable in either form: the subsequent withdrawal of the belief claim with impunity or its advance lifting from all concerned via the predication of disbelief. The misgivings voiced by revisionists about the presuppositional negation test can wait. However tacit, the consensus on the relative scaling enables us meanwhile to proceed with the analysis and comparison of affirmative utterances. Granting this differentiality of factive commitment, however, still leaves further questions open, either mysterious or worse than unresolved. For example, where does the commitment (here, to the truth of the embedded presupposition) spring from? How do we infer, grade, verify it? And who undertakes it? Of all questions, ‘‘Whose commitment?’’ most needs to be reopened, since none has been so generally and damagingly foreclosed. Who else, as it were, should bear the commissive load if not ‘‘the speaker’’? . Epistemologists traditionally privilege the first variable, discourse analysts the second, with little contact for once between the philosophical and the pragmatic approaches. Even J. L. Austin’s anatomy of ‘‘know’’ as performative in ‘‘Other Minds’’ (: –), his most influential work among philosophers, has left no mark on presupposition theory. As a result, both disciplines have lost. For example, epistemologists (including Austin) have never related their paradigm ‘‘know’’ to other factives, in or out of negation; while presuppositionalists would shrug off ‘‘I believe . . .’’ as a nonfactive, in disregard for its evident, if low, commissive force vis-à-vis ‘‘I know’’ and the ostensible evaporation thereof in ‘‘He believed . . . .’’ Yet see Lyons : – for the attempt at correlating the factors within epistemic logic. 176 Poetics Today 22:1 On this very ground, indeed, some would exclude categories like ‘‘verbs of judging’’ (e.g., accuse) from the presuppositional domain. ‘‘The implications are not attributed to the speaker, so much as to the subject of the verb of judging’’ (Levinson : , also  n. ): to the accuser, for example, whose hostile attitude toward the accused is ‘‘implied’’ without being endorsed by the verb.34 On pain of exclusion, nothing must threaten the speaker’s monopoly. But what if the speaker of the factive discourse as such proves to be a respeaker, re-presenter—indirect thought-quoter, say? The commitment to the given re-presentation then turns double to include the quotee as original representer and now the overt subject of the factive verb that introduces and governs his re-presented thought.We need sharper tools to capture the duality. For now, observe how (a) enacts two knowers—of whom, moreover, the thinking ‘‘he’’ rather than the vocal ‘‘I’’ is officially saddled with the epistemic load. By contrast, (b) exhibits a single believer only, yet the part and its epistemic posture again falls to the ‘‘he,’’ now re-presented for a change in ‘‘my’’ altogether noncommittal voice. This refines afresh the difference between factive and nonfactive statements but also underscores the copresence in them of at least two attitudes, like-minded or otherwise, toward the embedded proposition. In likeor other-mindedness, ‘‘the speaker’’ must live with the subject.The wider and sharper the commissive difference, themore salient and inevitable the common duality in viewpoint. Factivity is a key variable and variety of indirectness, while bi-perspectivity, or what I term perspectival montage, is a constant of indirect as quoted discourse. The former hinges on the choice of quoting predicate, the latter inheres in quotation’s criterial two-in-oneness: discourse about discourse must yoke together the respective discoursers. Given a factive affirmation in indirect discourse, like (a), the built-in perspectival montage flexibly extends from the transformer, which surfaces here as main clause, to the inset as that-clause (or its otherwise embedded, e.g., infinitival, equivalent). And given the universality of such montage across quoting forms—the indirect among them—it remains to discover . Within factivity proper, compare how Kiparsky and Kiparsky (: ) deem exceptional ‘‘situations where two egos are involved, as in the case of an actor describing his part.’’ The exception is actually the built-in rule. Conversely, and idiosyncratically, Ann Banfield (: –) would rather exclude the speaker: ‘‘Factives in main clauses in narrative with no overt first person seem not to imply the truth of their complements. . . . We read these sentences only as the private realization of their subject of consciousness.’’ This reversal of the usual one-sidedness is of course simply false. Also, in the light of example (), compare how the fixture on the speaker’s univocality, as well as intentionality, hampersGricean implicature analysis. Sternberg • Factives and Perspectives 177 what singles out the (de)composition of the factive variety, and why.With this in view, let us reconsider the clause-length segments in order. The transformer has already been defined (indeed named) as a transitional link in the chain of quotation: intermediate between quoter-occupied frame and quotee-evoking inset, it is necessarily two-faced. For the extra precision required here, we need to dovetail componential with perspectival analysis: to unpack the quoting/transforming verb by reference to its meaning in the lexicon as well as to its mediacy in the chain of discourse. Exemplary as ever, know decomposes into three relevant nuclear semantic features or factors that in turn divide (unevenly) between the transformer’s two joint perspectives, the quoter and the quotee. Let us unpack these montages in tandem. Among the verb’s sense-components, one mentalizes the report, another intensifies themental state(ment) reported, yet another validates it. I will name and assign them accordingly.35 First, there is evidently the thought component, or mentalizer, which encodes the factive predicate (together with the factivized—here, ‘‘known’’—inset itself ) as a reflex of a psychic event: an insideview, in short. Denoting the quotee’s secret life, the mentalizer opposes the predicate to all communication verbs (e.g., say, tell, answer) while grouping it with other verbs of cognition, perception, or emotion (believe, think, suspect, assume, presuppose, observe, feel, etc.) Yet know differs from the latter, too, in encoding a higher confidence on the quotee’s part in whatever his mind represents to itself—the highest possible among quotational predicates.The difference now accordingly gravitates, or extends, from the ontic axis of subjectivity (public/private, outer/inner) to the epistemic rank order (certain/uncertain, necessary/ impossible). Intensified by this sense component, the verb uncovers a thought entertained with certitude. The built-in presence (or absence) of the intensifier explains why the quoted knower’s (‘‘his’’) mental re-presentation in (a) deters the kind of attenuating follow-up that the quoted believer’s exact counterpart in (b) tolerates: ‘‘. . . but he wasn’t really sure,’’ say. Here precisely the verbs’ respective locations on the scale of intensity make the difference.36 In this re. I have opted for the terms mentalizer, intensifier, validator, rather than ‘‘mentaleme’’ or the like, because of their helpful forkedness.Their reference can alternate between the semantic particle and the discoursive party who brings it to the verb: exactly the two issues correlated here. . Therefore the claim that ‘‘ ‘X knows S’ presupposes S and asserts that X believes S’’ (Beaver : ) unduly lowers, or de-intensifies, X’s mental state to equal that asserted in ‘‘X believes S.’’ 178 Poetics Today 22:1 gard, though the twinning of mentalizer and intensifier produces a unique composite that resists paraphrase, the nearest equivalent to (a) would be (c) He held true strongly [with all his heart/with the utmost assurance] that I had got the evidence, as distinct from the mere ‘‘He held true that . . .’’ of (b). The explicit adverbial booster tacked on to the one predicate would then do duty for the semantic component built into the other. But does the paraphrase of (a) into (c) mean that (factive) presupposition, like Grice’s implicature, is ‘‘nondetachable,’’—triggerable by synonymous wordings—and therefore reducible to the fellow inference? Not at all, if only because any such paraphrase will lack the original’s key component and will accordingly leave the inference or commitment (here, the that-clause) deniable. Even amidst such possible twofold likeness in indirect discourse predication, (c) might yet reasonably finish, no less than might (b) itself, with the counterstatement ‘‘but he was wrong’’ or ‘‘but I hadn’t.’’ So might finish even quotes heralded by nonfactive verbs that, like ‘‘know,’’ comewith a built-in intensifier: ‘‘felt sure,’’ for example, or ‘‘was convinced.’’ The amenability to a counterstating sequel no more depends, then, on the form than on the very occurrence of the reinforced start. An outrageous tail end to (a), the same ‘‘but . . .’’ pinpoints at last where ‘‘know’’ as factive parts company with nonfactives, however intensified: in the presence of yet another semantic feature, which encodes the quoter’s endorsement of what the quote expresses.This validator loads the verb with the commissive force of presupposition.37 The know/believe dividing line thus consists in neither of the respective transformers’ earlier features: the one (mentalizer, for thought predication) shared, the other (intensifier, for thought reinforcement) sharable in amanner, and both oriented to the thinking subject, the knower or believer. It is rather the factor of validation and with it the quoter’s own epistemic attitude to the truth of the thought—all the way between full endorsement and none—that makes the presuppositional difference. The built-in, semanticized validating component explains why the quotermay not contradict the putative knower (or the discoverer, regrettor, resenter) without absurdity while freely contradicting the believer (or emoter, sayer, guesser, observer). It also explains why no amount or manner of (c)-like, or ‘‘sure’’-like, intensification would produce a comparable result outside factivity. For such a possible analogue, the added transformer-strengthening ersatz must shift . My ‘‘validator’’ bears some resemblance to the ‘‘neustic’’ in Hare  []: –, : – as the ‘‘nod of assent’’ to the speech act performed: see also the development by Lyons (: ff., ff.). Sternberg • Factives and Perspectives 179 its bearing (origin, viewpoint, responsibility) fromquotee to quoter. It must, in other words, turn from first-order (‘‘subjective’’) intensifier to secondorder (‘‘objective’’) validator of the thought: from re-presented part(icle) to re-presentational posture. (d) He held-true, rightly, that I had got the evidence. As (c) approaches (a) in intensification, so (d) does (a) in validation, either nonfactive substituting a makeshift surface phrase for the appropriate fixture in know’s deep semantic make-up. Even so, negate either paraphrase, and the analogy will collapse to re-establish the lexis-boundness (‘‘detachability’’) of presupposition. By the same token, inversely, adding those very makeshifts to the factive (a) would incur redundancy (literally, surface on top of implicit endorsement) rather than yield novelty value (quoted belief elevated in quoting gloss to the status of factuality). Again, the opposite adverbs, ‘‘incorrectly,’’ for example, would make nonsense of the factive (comparable to the sentential follow-up ‘‘but he was wrong’’) while simply and informatively invalidating the nonfactive. Most revealingly, the underlying logic holds even where the surface upor down-grading is alike informative in both predicate types. Regardless of whether and however you modalize the factive predicate out of its basic categorical form, the validation will remain unaffected.Thus ‘‘He perhaps, or certainly, knew that I had got the evidence’’ vis-à-vis ‘‘He knew . . .’’ in (a): the modal adverbs qualify or overinsure ‘‘his’’ knowledge of the fact, not the factuality of the knowledge statement itself in the reporting (otherqualifying, overinsuring) I’s own eyes. The modalization, and with it the novelty value, bear either way on the insideview, rather than on the state of affairs viewed from the inside but always validated from without, by ‘‘myself.’’ Moreover, not the validating feature alone but the entire indirectdiscourse transformer (‘‘He knew’’) comes from the mediating activity of the quoter who shares the overall frame with us. The quotee in (a), as in (b), would originally, at least normally, refer to himself as ‘‘I,’’ not ‘‘he.’’ Nor, whatever his certitude, would his original private expression predicate of himself (his ‘‘I’’) knowledge, any more than belief: he feels the cognitive state rather than states the cognitive feeling.38 On the plain indirect reading, then, the quotee never originated the threefold factive predicate, any more than he necessarily verbalized the ensuing statement. It emanates from the quoter, who selects it in transmission . In case we do assume such original self-predication, (a) will read as free indirect discourse, on which more later. 180 Poetics Today 22:1 out of the repertoire of quoting verbs in order to characterize in advance, for our benefit, the discourse act to be transmitted. So in (a) the quoting I’s transformer evidently serves typical expository, pre-contextualizing ends. The grammatical subject having identified the quotee or subject of discourse as ‘‘he,’’ the three-component verb proceeds to locate ‘‘his’’ discourse beforehand on the speech/thought axis, on the hierarchy of certitude, and on the range of truth. The mentalizer in ‘‘knew’’ establishes the knowledge claim to follow as an originally interior representation; the intensifier encodes the original representer’s wordless confidence in it; and the validator stamps it with the re-presenter’s own approval. In rough paraphrase, their meanings compose into something like, ‘‘He held true, strongly and rightly, that. . . .’’ Much the same threefold nuclear composition underlies the kindred predicates be aware, resent, regret, etc.39 Across all differences in the quotee’s state of mind—between cognitive and emotive interiority, say, or between more or less salient intensity—the quoter’s endorsement neverwavers. (The same goes for counterfactives likewish, except that they have an invalidating semantic particle instead.) A series of conclusions follow: () Although the entire composite verb has been supplied in mediation, two of the components bear jointly on the mediated discourse (or subjectivity), one on the mediator’s discourse about that discourse: as befits the factive version of the double-facing, doubly oriented transformer in its specificity regarding presupposition. () If the validator is built into the (here, factive) language, so is the presupposition: the validated quoting sentence itself then necessarily presupposes the factuality of the quoted sentence. The quoter just actualizes and officially assumes the presupposing as the user of that language schema, complete with binding verb, in discourse. This already lays the ground for reconceptualizing the issue against either of the traditional approaches (or, in the usual hybrid theories, any of their mixtures): () Against the ‘‘semanticists’’: locating the presupposition in the sentence, between the (transformer and inset) sentences, enforces no break with discourse and its anchorages and role-players, but a principled continuity. Nowhere more evidently so than apropos factivity, whose very intersentential construction as well as enunciation—though uniquely manifest in language—perforce bear on discourse about discourse about the world. Hence the smooth advance from second-order sentence to its second-order speaker . So, when Grice (: –) translates regret into ‘‘think and be anti,’’ he leaves out its factive differential vis-à-vis, say, fear; and cf. note . Sternberg • Factives and Perspectives 181 as carrier of the presupposition. Nor does the (inter)sentential locus go with translating presupposition into ad hoc and unearthly entailment rules, or any substitute.The only ‘‘entailment’’ involved concerns the strict necessity of the linguistic basis: given the inherence and the meaning of ‘‘validator,’’ the encoding of (factive) presupposition in the language system is a logical truth. () Or the other way round, now against the ‘‘pragmatic’’ approach: given such encoding, (factive) presupposition is uncancelable in discourse, whatever the ‘‘context,’’ just as is the semantic feature +male throughout the uses of ‘‘boy.’’ Once inbuilt, always inferable. (For that matter, if you recall the inference types to which pragmaticists would assimilate presupposition: it no more lends itself to canceling than does entailment on all accounts and conventional implicature onGrice’s own theory. As to conversational implicature, it is precisely defined as unencoded, ‘‘nondetachable,’’ hence cancelable.) What becomes, then, of presupposition under (con)textual pressure, notably from inconsistency? For an answer, we have to bring in the rest of the factive design, where the presupposed statement constitutes the quoted, ‘‘inset’’ object. The perspectival montage necessarily carries over from transformer to inset: from the reporting to the reported clause, and within indirectness, from the matrix to the embedding. I want to reemphasize this necessity, as a universal with factive specifics and corollaries, because it has not just been ignored but in effect categorically denied in various quarters, and because so much hangs on the issue. In the specific framework of presupposition reviewed here, this necessity already derives its rationale and its hallmark alike from our earlier analysis. By definition, a factive affirmed in indirect report not only enforces but also distinctively enforms such continued montage on at least one perspectival axis, namely, the informational, world-imaging, epistemic. For the very make-up of its transformer heralds a two-party statement in the inset—a meeting of minds apropos, say, the obtaining of evidence, twice upheld in (a)’s re-presentation. The rule applies throughout factive presupposition, where the affirming presupposer is the indirect quoter cum inferred validator of the thinking (e.g., knowing) subject’s position. If nonfactivity bi-focuses, then factivity equi-focuses, as it were, the state of affairs thus obliquely transmitted. On top of this definitional imperative, moreover, come the universal licenses of reperspectivizing attached to second-order discourse. Whether taken willy-nilly or freely or guilefully, they enhance the quote’s maneuverability between the viewpoints. (Which also redoubles its ambiguity, since 182 Poetics Today 22:1 we understanders and decomposers don’t usually know beforehand, if ever, which composite issued from which licenses.) Qua re-presented thought as against speech, for example, the inset here need not have been originally verbalized, either.The transformer’s combination of quotee-oriented mentalizer and intensifier with quoter-oriented validator may accordingly always stretch to the that-clause in another guise—the quotee providing the mind-stuff, the quoter the words. First the mentalizer built into the quoting verb, then the mentalese translated into the quoted clause, with an analogous doubling of input and outlook. (Except, to repeat, for the quantum leap in ambiguity: the material to be divided no longer consists in three semantic elements—lexicalized, hence predetermined at that—but in an entire newly-made sentence.) Again, regardless of the original’s (full, partial, zero) verbalization, there is the gamut of more or less faithful re-presentation. The inset image can always gloss, rephrase, shorten, expand, despecify, enliven, weight, subvert the original in transfer, to the limit of misre-presenting. For all we can tell, ‘‘he’’ in (a)may have really thought anything from ‘‘The bastard has nailed me’’ to ‘‘The liar has got nothing.’’ And if we can’t tell—the epistemologist in us wonders—how could the quoting inside-viewer and guarantor of that hidden knowledge? In short, the less medium-preserving, verbatim, scrupulous, warranted, or just the less determinable the echo of the subject’s original thought, the larger the quoting-I’s share in the montage beyond the inherent validating minimum. Given that the perspectival montage willy-nilly extends to the inset, another two of the axiomatic assumptions about factivity need to be rethought together: that it involves commitment to ‘‘a proposition,’’ or merely a proposition, and nomore than one commitment, the ‘‘speaker’s.’’ Neither axiom can possibly be true, since quoter (alias ‘‘speaker’’) and quotee co-occupy the inset discourse that expresses and subsumes the factivized proposition. The expressive manner as well as the expressed (inter alia, propositional) matter, therefore, must issue from both of them—jointly or also apart, harmoniously in all regards or otherwise—and lie between them in the text we encounter. As the one purports to re-present the other’s (here, inner) representation, the only problem is how to break down the finished composite and assign the proper share, as best we can, to either contributor. Where does, or did, the line pass between the original discourse and the second-hand discourse about that discourse found in the inset? Evidently, the inset’s perspectival montage is harder to break down (though, as will emerge, only relatively speaking) than its componential antecedent and equivalent in the transformer: to the extent that linguistic performance Sternberg • Factives and Perspectives 183 enjoys by nature much greater freedom than the language (e.g., lexical) system. Whether the licensed performance has its own systematicity, that is rather the question; and factivity, partaking of both language and discourse in its very composites, again suggests a principled affirmative answer. At the quoter-maximizing extreme stands drastic misquotation, garbled out of truth to the original, possibly even reversed. Here is an insideview factivized with gusto but (on omniscient authority) counter to mental fact: () [A wife bent on treating her husband to a pedicare session:] ‘‘I enjoy it. It’s such a nice cosy thing to do. And you know you love it yourself.’’ Freddie knew no such thing. (Amis : ) The reality bumps against the pretense to like-mindedness or what I called equi-focusing: equal delight, awareness, self-engagement. The putative quotee would (a less henpecked husband than Freddie will) deny outright the knowledge statement about his state as one unknowable of and unthinkable by himself. He would justly disclaim the whole lot, from stylistic manner to presupposed matter, from the uncongenial idiom to the untrue substance of ‘‘love’’ for wifely foot treatment. So, where authoritatively informed or incredulous on empirical grounds, will we hearers/readers ‘‘cancel’’ the presupposition in ‘‘context’’ under threat of inconsistency with the facts?Of course not, if only because the presupposer remains bound towhat she factivizes: the perspectival setup is itself a context (indeed the master context of discourse, though still unappreciated as such).We rather assign it all to the (fallible, wishful, inventive) mind where it exclusively (all too consistently) belongs. The nominal inset then comes to reflect, and reflect on, the frame. It likewise reflects on the flatteners of indirect to univocal, in effectmonologic, discourse (see references in Sternberg a and b passim, : ff.) and particularly of factivity to the epistemic outlook of the factivizer as ‘‘speaker.’’ But even a demonstrably invented or inverted quote entails a perspectival montage, owing to the head-on clash between attributed and original subjectivity, between the like-mindedness (equi-focusing, harmony) pretended and its negative truth-value. The resulting dialogue among the minds in play only specializes in frictions. A univocality of sorts, () thus comes with a vengeance: against the quoter’s wish to speak for two, against the quotee’s wordless denial, against the author’s and our knowledge of the respective states of knowledge. Within this network of ironic oppositions, themonopoly on voice ismultiply perspectivized—to a greater extent, if anything, than in usual, more harmonious double-voicing, factivized or otherwise indirect. 184 Poetics Today 22:1 With comparable effectiveness, our simple example also lays bare a set of further misconceptions. I refer to the fallacies of ‘‘common, or mutual, knowledge,’’ of samesaying or samethinking, and of fictionalizing. One bears directly on (factive) presupposition, yet has wider consequences for discourse; the second would generalize over indirect and the third over quoted discourse at large, yet both receive their ultimate quietus from factivity that goes counter to fact, as here. Of these, the first two require only summary mention by now.The widespread idea of ‘‘common/mutual knowledge’’ is among the worst fallacies in pragmatics, especially its inferential branch, and maybe the strangest. In general, such ‘‘knowledge’’ comprises the ‘‘background assumptions’’ shared by the parties to the discourse as the ‘‘context’’ within or against which the ‘‘foreground’’ of the utterance operates. Regarding our inference type, the ‘‘speaker’’ would then ‘‘presuppose’’ what ‘‘speaker and addressee mutually know’’ (e.g., the truth of the factive complement, my inset) to the highlighting of the novelty (e.g., the factive predication, my transformer). And vice versa: the speaker would allegedly never contradict the mutually known, hence such contradictory presuppositions evaporate. This would hold even on less extreme variants, whereby the commonality does not necessarily reside in the foreknown but embraces the ‘‘consistent’’ (Gazdar : ff.) or the ‘‘uncontroversial’’ (Grice : ) between the parties. Accordingly, the common epistemic ground defines ‘‘context’’ and at times (as late as Kay : ) ‘‘presupposition’’ itself, down to the limits of its survival in ‘‘context.’’ (For an approving overview, see Levinson : esp. – passim.) This dogma has been progressively (and, given its strangeness to anyone outside ‘‘pragmatics,’’ a fortiori to narrative theorists, unsurprisingly) refuted ever since we encountered it: apropos the muddled equation of such ‘‘commonality’’ or concord of awareness with ‘‘context,’’ dead against example () from Nelson Goodman’s sneer at fundamentalist metaphysics. Nor does its exposure stop here. Essentially, though, the fallacy betrays a multiple confusion about the worlds, the voices, the workings of discourse, and presuppositional discourse above all. For the confusion gets worst confounded where the very terminology abets it, with the result that analysts bracket the strict technical sense of presupposition (as commitment/inference type) with the loose, ordinary usage (as whatever goes without saying). Thus, the discourse parties involved freely oscillate in current analysis between the interlocutors and the ‘‘average’’ (‘‘common’’?) hearer/reader, ourselves—as if their reality, knowledge, posture, truth bond were answerable to (hence inferable from) some extradiscoursive, even universal norm Sternberg • Factives and Perspectives 185 of ‘‘context.’’ An obvious utopian absurdity. But why should the interlocutors proper fall under any norm of joint awareness and goal? They may at will and often do use, say, the factive complement not to reiterate old information (shared, loosely presupposed) but to spring on the addressee new, uncommon, discordant, controversial matter, including ‘‘knowledge,’’ whether Goodman’s in () or the wife’s in (). All the more so because, as factivizers, they are quoters rather than just speakers, hence commit themselves in presupposing knowledge to like-mindedness with their quotee rather than any actual or possible addressee. From the addressee’s side, in turn, uncancelability ensues, regardless of how contradictory the putative ‘‘context.’’ Even thoughFreddie combines the roles of quotee and addressee, the fact that he ‘‘knew no such thing’’ cannot yet erase, but only ironize, his wife’s presupposition that he does know. (Our reanalysis of negation will indeed deal the last blow to this unearthly fallacy.) Her truth claim about their joint knowledge is one thing, made and retractable by her alone; its truth-value is another thing altogether, judged in relation to the overall discourse frame. Here, omniscience itself explicitly falsifies the whole knowledge state(ment)—through an insideview of the quotee/addressee alleged to know and co-falsifying the allegation—without in the least affecting the knowledge claimant’s own pledged view to the contrary. Another dogma of commonality bears on the indirect quoter vis-à-vis the subject quoted, as if the image here equals the original proposition. Donald Davidson (: –) famously puts it in terms of ‘‘samesaying: when I say that Galileo said that the earth moves, I represent Galileo and myself as samesayers,’’ for ‘‘an utterance of mine matches an utterance of his in purport.’’ Likewise with ‘‘samethinkers,’’ though Davidson does not officially extend the term to them or call the relation indirect discourse (ibid.: – ). But this idea ofmatching recurs across terminological, conceptual, disciplinary lines. Thus according to John Searle (: –), the indirect reporter ‘‘is committed to repeating the speaker’s propositional act’’ (or to propositionalizing what the thinker believes). Such reporters are ‘‘not samesayers, but same proposition expressers.’’ As with the philosophers, so with linguists and poeticians.These would frequently even bracket the reputed univocality of indirect discourse with its being an ‘‘interpretive’’ mode, which paraphrases or translates someone else’s expression into one’s own words (Banfield : ) to generate a single content-minded viewpoint. That indirectness must and/or does paraphrase, however, is simply a weaker version of the fallacy that directness reproduces, and I have already exposed it as such (especially in Sternberg ).Three aspects of the fallacy are of vital concern for us. By immemorial license, there need not be any original for the quoter to match—paraphrase, (re)propositionalize—as in 186 Poetics Today 22:1 future-oriented, hypothetical, and negative quotation (‘‘He will/might/did not say [or think] that . . .’’). Even if there supposedly is one, as alleged by the quoter, we often cannot tell whether it actually was andwhat it represented; the ensuingmontage then leaves the original proposition undecided, at best inferable. Again, even in the optimum conditions, when we have access to the original and the quoter stands not just tacitly but lexically (‘‘know’’) bound to its truthful paraphrase and very truth, as in (), what happens if we find it misrepresented, indeed ‘‘countersaid’’ or ‘‘counterthought’’? Would we exclude the ‘‘counterdiscourse’’ from indirectness and/or factivity? Surely not—any more than we would the quoting of nonexistent or ambiguous originals—on pain of confusing re-presentational form and function, epistemic pledge and value.The nonfactuality of a claim as high as a knowledge statement (highest on the scale of alleged samethinking, if you will) only recoils on the factivizer more powerfully than on any analogous indirect quoter. The thirdmisconception in effect begins where the second ends, with the collapse of any rule of ‘‘samesaying.’’ Throughout her book The Fictions of Language and the Languages of Fiction (),Monika Fludernik adopts to good purpose my exposure of the age-old reproductive fallacy in direct, a fortiori otherwise reported discourse, and its consequences for a viable theory (see note )—with one unhappy twist that she herself considers so important as to merit titular dignity.Whatever falls short of one-to-one reproduction (and every second-hand image of utterance or thought does, along with every other image) thereby becomes ‘‘fiction.’’ Under this category Fludernik subsumes (especially in chapter ) my array of anti-reproductive quotational forces and forms, whether constrained, licensed, or possible.The array includes despecifying (e.g., an original summarized or compressed), deconcretizing (one quote for many originals), stylizing (for intelligibility or generic effect), modalizing (a prospective or hypothetical quote), giving utterance to dumb reality, memory lapse and so forth, as well as the pressures always exerted by the frame on the inset (liable to reverse even a verbatim echo). From these modes of deviance, she concludes what never follows in reason nor indeed is the case in practice: ‘‘[A]ll linguistic speech and thought representation relies on a mechanism of schematization and typification which is independent of actual speech and thought process and can be analysed in terms of a fiction ‘manufactured’ by means of language, by means of linguistic devices’’ (: ). Even ‘‘mimesis in oral [real-life, as distinct from literary] language would then be of the same quality as in fiction—not imitation, but invention and projection’’ (ibid.: ). If it did follow—to single out one outcome—this conclusion would necSternberg • Factives and Perspectives 187 essarily rule out factivity altogether from language use. As factive quotation must both assume a nondirect form and image a nonverbal (‘‘thought’’) object, it cannot even reproduce the original to the extent possible for a direct quotation of speech. And if the nonreproductive amounted to the fictional, what would become of the distinctive commitment to the truth expressed by the inset quote? The compounded form/object exigency would alone disable the presuppositional force. One need not add such licenses as paraphrasing or modalizing to see how this odd fictionalizing twist reduces to absurdity, with large consequences—the largest, since they bear on the very ground rules of representation. Nowhere in reported discourse, or for that matter in discourse at large, does this category mistake so glare as in factivity (and with it the pernicious influence exerted even on rational scholarship by today’s lust for assimilating everything nonfactual, or just about everything, to fiction). And (), because it marks such a radical deviance from the original, highlights the non sequitur. Does the wife manufacture, invent, project? In a sense, yes, but this sense has nothing to do with fiction. Her ‘‘You know,’’ as a factive, binds her to the truth of the ensuing report on the husband’s state of mind. When that report proves nonfactual, indeed counter to fact, she does not at all stand revealed (amid continued self-binding) as having fictionalized but as having falsified the matter in error. She simply got it wrong. Nor would she turn fiction-maker if she were a deliberate ‘‘inventor,’’ speaking against her better knowledge of his ‘‘loves.’’ In epistemic commitment, then, the question of fictionality never arises, because the values applicable to the endorsed (e.g., knowledge) statement polarize, or atmost range, between the factual and the false alone. Checked against the facts, a factive truth claim elicits in reason either a positive or a negative truth judgment, and if the second, counts as either a lie or a miss.40 What my repertoire of antireproductive forms establishes, inter alia, is the flexibility of such values or judgments, rather than their wholesale suspension, let alone abrogation. According to the frame’s rigor or permissiveness, the same quote (e.g., a knowledge claim reported in shorthand) may modulate between the true and the untrue: between the operationally equivalent or nonequivalent to the original. Whenever assigned, deviance integrates according to the logic of functionality—the way the text resolves itself in context—not of fictionality. Fictionalizing never comes into it, on pain of a contradiction in terms, a breach of (onto)logical premises. How else would the domain of suspended truth migrate to the axis of truth, assuming an existence discontinuous with its own nature? . For a more developed argument see Sternberg : ff, , , : ff. 188 Poetics Today 22:1 Nor is a factive voiced or thought by a fictive character exceptional, for the ground rule will then operate within the fiction, as it does in () and numberless counterparts, at a proper remove that we easily effect. Once posited, a fictional world enacts its own truthsayers, falsifiers, mendacious or mistaken, and fictionalizers: however imaginary, the boundaries among them are essentially as well-defined, and our ability to tell them apart sometimes as challenged, as in real life. And the same holds for all indirect, all quoted, all first-order discourse (i.e., all the way from the presuppositional to the assertive, from the re-presented to the representational) under such epistemic commitment, genuine or mimetic. Last, if re(-)presentation in all these varieties has a definite generic, even ontic bearing, it concerns narrativity rather than fictionality. As an image of the world, no matter how static-looking, representation implies time, change, and so (hi)story; as a discourse event, it further implies the participants’ viewpoint(s), itself time-bound, on the (hi)storied world. Representation, as an image of a world-image and a discourse about discourse, by nature multiplies these latent objective cum subjective narrativities. And deviant re-presentation, factive or otherwise, maximizes in turn this highlevel energy: it bristles with the tales of perspectival discord that we have begun inferring, and will progressively replot and generalize, to make sense of incongruous data. Among the rest of its manifestations, this logic cuts across the varieties of montage in factivity. At the other, quotee-favoring pole, opposed to (), we encounter near-verbatim echoes of the original thought—or what counts as such—reported with the same factive transformer. Look at this scene of cross-examination between a prosecutor and the passenger of a hit-and-run driver: () Counsel: You knew that your friend [the accused] had lied about it? Smith: That he’d said he hadn’t been in an accident—yes. Counsel: That he’d lied about it? Smith: Well—yes. (Cecil  []: –) Note how this refutes anew, from the opposite direction, the idea of ‘‘fictionalizing’’ in quoted discourse, the confinement to univocality and samesaying in indirect discourse, and the premise of epistemic ‘‘mutuality’’ in presupposition. Either interlocutor here schematizes the original thought— neither even pretending to reproduce it, as they might in the direct form— but their quarrel nevertheless strictly concerns a matter of fact, with grave legal and practical implications for the accused, possibly for thewitness too. Re-presentations that vary from the discourse-event and from each other Sternberg • Factives and Perspectives 189 do not yet entail fiction, any more than does one made of whole cloth. Further, not only could the legal, as well as the marital, scene happen in real life; it even eschews the omniscient insideview associable with imaginative literature. Again, the assumption of epistemic commonality or consistency regarding the presupposition now breaks down the other way. If the requirement provesmuch too high in (), absurdly so, here it proves too low.The clash of wills embodied in the variant quotes (and heightening the narrativity latent in every variance into a tense reportive drama) starts, rather than ends, with an agreement about the bare propositional content of the witness’s knowledge at the time. That the accused denied to the police his involvement in the accident forms common ground, to which either party binds himself: the one having factivized it in point-blank address (‘‘You knew that . . .’’), the other assents in elliptic resumption (‘‘That . . .’’). The quarrel between reporter and self-reporter turns instead on the mental expression of the reportee’s knowledge—on the original’s hidden linguistic surface, especially the operative topic and verb. The prosecutor at factive reporting is not content with letting the facts speak for themselves, as it were, which would in effect make him alone responsible for the wording. He wants the record to show the ugliest (here also the genuine) version of the happening and the thinking: eliciting assent to ‘‘your friend had lied’’ will most incriminate the accused driver and render the hostile witness’s ‘‘knowledge’’ of it, then as now, guiltier to suit. Inversely, to avoid these penalties, the witness at self-reporting would vary into admitting only his awareness (by elliptic implication at that) of his friend’s negative statement (‘‘he said he hadn’t been in an accident—yes’’). Far from a theoretical nicety, he so minimizes the desired common ground that his apparently obliging ‘‘yes’’ amounts to a half-‘‘no’’ in practical juridic terms. It takes repeated pressure to compel him to abandon the euphemism for the plain whole truth, re-presenting the deed, the thought, and the operative language in their original ugly unity. Indeed, having once assented to the bare-bones proposition, he cannot help admitting the other half, namely, that his mental idiom fitted his knowledge as his knowledge did the action. It would be idle for him to object that he knows the idiolect of his own mind best. The weaker-sounding ‘‘Well—yes’’ now carries the stronger meaning, ‘‘Yes, my friend did lie and I knew it to be a lie.’’ Factive validation, then, accommodates disharmony—propositional or otherwise—between the validating and the validated perspectives in the given composite quote (as well as between either and the truth). Sheer reportive inversion, as in (), only carries disharmony to the either/or ex190 Poetics Today 22:1 treme. Transmitting the original thought essentially ungarbled yet leaves room for such divergences (and implicit tales of divergence) on the way to the finished inset that the montage can always get thrown out of ideal, literally equi-focal balance, to the privileging of one contributor.When balanced, the inset reflects total harmony, with endorsement to suit; when unbalanced, the harmony may consist in the propositional gist, exclusive of the rest, a fortiori of (say) the language and the value judgment, which accordingly turn divisive. Depending on which of the perspectives they fall to, the quoter’s presupposition will then outreach or abstract the quotee’s proposition. In radical imbalance, the factive presupposer will stand committed not just to more or to less of the inset, respectively, but to all of its aspects or to one alone, what it states about the world. Balance or imbalance? And if the latter, tipped which way beyond the shared propositional minimum? These are the basic possibilities that the factive indirect report leaves open for us unpackers, or disambiguators, to test in context. Conversely, the extent and inferentiality of thismaneuvering ground already militate against any ‘‘projection rule’’ whereby to foretell the presuppositions that a complex sentence takes over (or drops) from its clauses—any ‘‘formal’’ part-to-whole inheritance law. Since the early s, presupposition theory has endeavored to codify such a rule; but the quest is hopeless, becausemisguided, in light of the variability attaching to the data. How will you legislate the projection from component (e.g., inset sentence) to whole when the component turns out a predictably unpredictable mixed quantity, so that you never even know in advance what to project from any single given exemplar: the factual content or also the wording and/or the norms? The difference among such heritages is still deeper than appears, if only because the wording may carry further, nonfactive (e.g., existential) presuppositions that hover between the twinned viewpoints.The idea of reducing all possible variants to unitary rule therefore goes against the grain of factivity and discourse at large. In what ensues, its hopelessness will only worsen as the room for ambiguity, and proportionally for the contingencies of inference, widens. The worse ‘‘projective’’ legislation fares, the stronger the claim of the opposite idea, whereby theorizing means generalizing and explaining the interplay of unity with variety as a functional dynamics: the lines and rationales of variation are alone predictable, the outcome unrelatable to them before the discourse event. Thus, the imbalance within the that-complement need not even run to either extreme. In between, there range endless possibilities of combining (or, from our side, disentangling) the viewpoints that necessarily go into the inset. Every viewpoint boasts a variety of distinctive features—expresSternberg • Factives and Perspectives 191 sive, emotive, informational, ethical, logical, metaphysical, aesthetic, and so forth—which make up its unique subjectivity. The two-in-one montage can therefore issue from operations both along and across all the respective perspectival lines. Not only is the quoter free to compose his own peculiar style with the quotee’s unequal awareness or other-minded judgment—and inversely—but the divergent styles, awarenesses, judgments are themselves mixable.41 The interplay of reference will best exemplify in miniature how and why factive discourse may variously throw the quote out of perspectival balance without losing (or ‘‘canceling’’) the presupposition that lives on quoter/quotee agreement. For one thing, reference constitutes a meeting ground of the sundry perspectival axes: the choice of terms for reality-items bristles with expressive, cognitive, normative differentiae, and accordingly guides the unpacking of the inset’s extrareferential components as well. For another thing, the vexed issue of reference engendered presupposition and revisiting it within factivity now enables us to correlate the two major presuppositional triggers—along lines and with gains that already suggest themselves. Not only does reference, owing to themultiple perspectival features it implies, help with the decomposition of factivity’s twinned perspectives, and vice versa. Factive bi-perspectivity in turn points a larger moral about reference, when reference is in need of integration above all. Negatively speaking, whereword and thing conflict, as with ‘‘The present King of France,’’ standard approaches go against communicative reality in dooming the reference either to presupposition failure or to untruth, a dead or a bad end.These approaches hardly seek to integrate the clash, not even via the shift in perspective that will normalize the reference at a remove: the speaker possibly doubles here as quoter to mediate another’s (uninformed, biased, or just odd) referential viewpoint in its own terms. Factivity encourages such hypothesizing, inter alia, by its necessary discoursive compositeness, and so brings home the general rule: we must attentively infer all propositions, as we do its specific presuppositions, from the given discourse, which may always equivocate between voices. Here is the forked tongue of reference in action: () ‘‘I saw the ambulance drive off and there was police cars all over the place, so I knew something had happened.’’ (Porter : ) () Charlie Heller’s name is not exactly a household word in America. . The openness of suchmultiform crosses to indirect quotation, factivity included, has been amply demonstrated in Sternberg , , b, , ; see also McHale : – , : –, particularly for expressive mixture, and Fludernik : ff. 192 Poetics Today 22:1 Still, enough people are aware of what he did (without necessarily associating it with his name) to make a book on the subject worth while. (Littel : ) () The guilty knowledge that I was going to rob Mrs. Joe—I never thought I was going to rob Joe, for I never thought of any of the housekeeping property as his— . . . almost droveme out ofmymind. (Dickens n.d. []: ) () She had not mentioned the matter to [her husband], as she knew him to be a man of the most ridiculous integrity, who would never have agreed to her signing the [bogus] form. (Cecil  []: ) Throughout, the peculiar referencemakes sense if oriented to the quotee, so that the choice of linguistic terms for the referent captures assorted features of the putative knower’s subjectivity—generally defective in some way and measure. Although better informed about what happened, the self-quoter in () levels down the knowledge to ‘‘something’’ so as to match the quoted self ’s awareness at the time. Likewise in (), except that now the quotee (‘‘people’’) credited with ‘‘awareness’’ doubles as the addressee. The thirdperson masculine pronoun ‘‘he’’ best refers to the hero in question, since its emptiness has the virtue of open-endedness. In the absence of the higher, nominal common ground between the parties to the tale about him, the pronominal will do. It strikes an epistemic balance between the author’s informed ‘‘Charlie Heller’’ and whatever pseudonym with which readers will associate the referent (conceivably including ‘‘the present King of France’’: not such a bad cover for an ex-spy whose ‘‘pet name’’ for his ex-employer, the CIA, is ‘‘Byzantium.’’ Joking aside, the logic of other-minded, perspectivized reference does newly apply to the philosopher’s crux). More overtly and evaluatively, next, the grown-up Pip adheres to his young experiencing-I’s crowding out of ‘‘Joe’’ by ‘‘Mrs. Joe’’ as object of ‘‘rob,’’ hence as owner. He thereby reflects in miniature the boy’s warped view of the family circle, dead against the balance of natural kindness and affection he received, as well as topsy-turvy by contemporary Victorian standards. To crown the incongruity, the scale latent in the very names chosen—‘‘Joe’’ and ‘‘Mrs. Joe’’—polarizes with the statuses actually accorded to the respective name-bearers by the little subject. Ownership, or the sense of it, depends on whether the candidates for reference exercise brutal power at home: ‘‘guilty knowledge’’ indeed, beyond that of the coming robbery and the self ’s ken at the time. Driving the inside-viewer’s judgment of the ‘‘knowing’’ subject’s illSternberg • Factives and Perspectives 193 judgment to the limit, () again encapsulates it, now in a mini-portrait: a woman’s attributing to her spouse ‘‘the most ridiculous integrity.’’ The loaded attributive reference is presumably hers and hers alone, because for the narrator to share it (even to the extent that the older Pip does the halftruth about robbing ‘‘Mrs. Joe’’) he would have to co-‘‘ridicule’’ the spouse’s attitude toward forgery along with her. Rather, he does underwrite everything—the faultless educated language, the husband’s image as a man of integrity, the likelihood of the man’s veto on the crooked deal—with the pinpoint exception of one value-laden adjective. This exemplifies afresh how the inferencing about presuppositional inference, in quest of the best fit, may end up with a focus on the quotee, and howwidely the focus ranges (inter alia) in scope: from the entire that-clause, as in (), to a part that jars within an otherwise shared, equi-focal inset. In factivity as elsewhere, wemust not expect any of the categorical (all or nothing, either/or) outcomes long favored by dogma across disciplines, legislated with the stick of ‘‘unacceptability.’’ The inferential dynamicsmay lead to our singling out an element of the whole presupposition (e.g., a referring term) for oppositional inverted commas—as though it were quoted directly and at a remove, if not under protest, amid factive indirectness. By contrast, the following examples enact locally, andmore inferentially, what happened wholesale in (). The quoter obtrudes his own referential viewpoint (and all it signifies, consciously or otherwise) upon the inset proposition alleged to transmit the knowledge originating and shared with the quotee: () I said pompously, ‘‘We now know what we’re looking for is the log of the Skadi.’’ Willie said, ‘‘Oh, that ship?Well, sorry, old boy, butwe aren’t looking for the log. Do you mean the deck log or the engine log or the rough drafts of either, or the official log? Not to mention the movements book—what?’’ ‘‘So don’t mention it, then,’’ I growled. (Lyall : ) () [A murder suspect questioned by a detective:] ‘‘We were just good friends!’’ ‘‘Ho! Ho!’’ said Dover sarcastically. ‘‘But you’ve already admitted that not long before she was shot, you and Isobel Slatcher became less ‘good friends’ than you had been. She knew you wanted to marry Mrs. Ofield here, didn’t she? And she didn’t like the idea, did she?’’ Ofield looked contemptuously down his nose and didn’t deign to answer. (Porter : ) 194 Poetics Today 22:1 () What I liked about her, she didn’t give you a lot of horse manure about what a great guy her father was. She probably knew what a phoney slob he was. (Salinger : ) Each of these exhibits some referring term (descriptive, nominal, pronominal) that likewise attracts notice, but toward the other montage-partner. The chosen term always jars here, rather than agrees, with the subjectivity in which it allegedly originated, the knower’s—because (we infer) it gives away the mediator’s divergent, possibly odd view of the thing re-presented. The reference-making ‘‘I’’ in () speaks to the ‘‘we’’-group of knowledge he considers shared—by another role doubling, the quoted mind now extends from the addressee to the speaker himself—only to betray the holes in his own expertise. His definite description ‘‘the log’’ meets with a superior knower’s flat, echoing denial (‘‘we aren’t looking for the log’’), rubbed in by a list of alternative referents (‘‘Do you mean . . . ?’’) that isolates the alazon from the group. He must therefore withdraw the definite article in favor of an indefinite counterpart (pronominalized by the anaphoric ‘‘it’’) that reflects the group’s true state of ignorance. The presupposition of (singular) existence, if not the whole factive report, backfires upon the ‘‘pompous’’ landlubber. Example () confronts us with the proleptic designation ‘‘Mrs. Ofield here,’’ and, of all people, from the mind of a rival who ‘‘didn’t like the idea’’ of her becoming Mrs. Ofield at that. Nor could Ofield himself ‘‘want,’’ or be ‘‘known’’ to want, ‘‘to marry’’ a woman already married to him and bearing his name. The discordant referential predating surely gives away, not the knower’s jealous, but the knowledge attributor’s insensitive, mind. Ofield may well look down his nose, if only considering the questioner’s bad psychology and general illogic. Still, though there was no such thing in the germane ‘‘known’’ world—any more than there was, or is, a King of France—all concerned manage to identify the referent of ‘‘Mrs. Ofield here.’’ The prolepsis once noted, integrated, and discounted—relativized to the ill-quoting detective—everything claimed by the factive insideviewmay therefore be true: the jealously of the quotee, Isobel Slatcher, even Ofield’s suspected role in her killing.Whatever the truth, subjective or objective, no failure of reference comes into it, far less makes it indeterminate a priori. In (), finally, we infer a direct counterpart to (), since the loaded predicative expression changes subjective poles.Whence the jump from the daughter’s absence of speech (not giving you ‘‘a lot of horse manure’’ about her father) to the conclusion regarding her harsh attitude (knowing him for ‘‘a phoney slob’’)? An insideview that reads lack of enthusiasm into her silence might appear in order, but such unfiliality as expressed, let alone Sternberg • Factives and Perspectives 195 factivized, in the reference demands harder evidence.The style along with the judgment in the thought-quote rather bear the trademark of Holden Caufield, who co-ascribes them to the quotee he ‘‘likes,’’ typically jumping from liking to like-mindedness about phoneys. But even Holden—our tale of report concludes—has enough doubts (and decency) left to modalize the factive, so as to qualify one of the joint viewpoints. Her ‘‘probable’’ gets marked off from his own presupposed knowledge, implicit but categorical down to the wording. This range of perspectival interplay translates into one of commitment, and multiple commitment, too. About the truth bond attaching to the factive verb itself (e.g., how the predication of ‘‘know’’ asserts an intense mental state in the alleged knower) enough has been said for now. The transformer’s assertive bond, however, finds a more complex sequel in what the inset presupposes: in the obliquities of indirectness. Here, to begin with, the talk of ‘‘the speaker’s’’ commitment to the ‘‘proposition’’ expressed by the dependent clause mistakes the lower commissive limit for the whole range, freezing it in the process. (Unsurprisingly, because how else would the desired ‘‘projection rule’’ evermaterialize, or remain imaginable at all?) The factivizing indirect quoter stands committed not to but at least to that proposition, or its gist: he may presuppose anything from the state of affairs re-presented by the inset to the embodied representation itself—depending on the elements he would appear to share (‘‘validate’’), obtrude, distance away in quoting.The transformer only foretells that he presupposes, rather than what, nor will the inset itself tell us, if ever, unless unpacked first into the component viewpoints. Where any given frame-occupant stands vis-à-vis the inset, between the lower and the upper limit, is therefore a matter of inference. Each presupposes, hence guarantees, exactly what we hearers/readers work out as the presuppositionmost appropriate to the (con)text—especially to the quoter’s viewpoint when disembarrassed of its twin. The validator underwrites as much or as little as we understand from the data on the operative rules. The symmetry between these unders highlights a type-distinctive inferential nexus that underlies the communication (if you like, the hermeneutic circle) of factivity, with effects that will keep branching out as we proceed. Presupposition being encoded in the language, any trigger signals a commitment obliquely made through the discourse and waiting to be made out from the discourse. Hence again the exemplary role of factivity. In its commissive as in its perspectival aspect at large, factive presupposing entails not just discourse about discourse, but also inference about inference—toward that attributable to the second-hand discourser’s own frame. Entails, I in196 Poetics Today 22:1 sist, because nobody can escape the exigency, not even those would-be formalists who misconceive (factive) presupposition in all the ways already outlined and still to emerge.They only do the inferencing badly or, to judge from the performance of communicators in life and literature, ill-theorize what they do quite well. At the same time, in the same inferential process, the indirect quotee also gets committed, albeit the way his viewpoint as a whole gets communicated—at second hand. The very idea of second-hand commitment will sound bizarre unless you appreciate (as presupposition along with speech act theorists, inter alia, do not) the wider, multifold network of relations in which inference necessarily operates. Here, it is all too easy to forget both the inseparability of the two parties to the quoted discourse and the artificiality of isolating any commissive act in which either figures (e.g., the ubiquitous sentence-length example) from the rest. Under such amnesia, the quoted subject fares worse than the ‘‘speaker’’—to the point of exclusion—because he apparently cannot be held accountable for the words, much less insideviews, least of all knowledge state(ment)s, attributed to him by another. Where if not with the attributor—the factivizer, above all— does the whole onus belong? But think of yourself, of your friend, of a protagonist you have just followed throughout a (hi)story, and the deceptiveness of this appearance will reveal itself. All subjects, historical or imaginary, bear their records of commitment with them, and these are diverse as well as incremental. Only part of the record consists in what the subject himself has voiced with whatever assurance on record. The other part consists in whatever has been ascribed to him, with whatever intensity and validation, by quoters of the most diverse authority—not least relative to that he enjoys about himself. It can possibly underrank, possibly approximate, possibly exceed, even transcend his own self-awareness, self-disclosure, self-confidence, with the allauthoritative narrator vis-à-vis the narrated figures as limit case. In epistemic engagement, therefore, the recordmakes or breaks, and our piecing it out from all the evidence available (re)constitutes, the subject. Once constituted, however provisionally, it becomes the measure against which we test for consistency each new engagement spoken or quotational, autonomous or attributive or attributed. Even when encountering a subject for the first time—in life, fiction, or artificially isolated example— we infer the record as best we can, if only from the probabilities of the very state(ment) that launches its buildup in the given (con)text.The above discussion has already established that an oddity, a fortiori inconsistency—the quoter’s or the quotee’s, the validator’s/presupposer’s or the mentalizer’s/intensifier’s—may arise on the shortest acquaintance; and Sternberg • Factives and Perspectives 197 how would it arise without some past (discourse, norm) to jar against? Inversely, thewealth and universality of the resources available for integrating it (e.g., by way of attribution to another, more fitting viewpoint, the quoting or the quoted subject’s) testify to the pressure exerted on us understanders by the threat of any commissive breakdown. If we keep in mind that the commitment at issue here is one to consistency—the epistemic face of this discourse-wide basis for sense-making, or ‘‘integration,’’ and not necessarily attended with other pledges—its secondhandedness no longer rules it out from the inferential arena. In (), for example, not only the judge, the jury, and the eavesdropping reader but the hit-and-run driver himself must somehow set the record straight: integrate his own persistence in denial with the barrister’s/witness’s knowledge claim that he had lied to the police about his involvement in the accident. The trouble is rather that the exigency no longer seems peculiar to knowledge and other presupposed claims, either, but assumes universal jurisdiction. Unlike the quoter’s commitment, the quotee’s in our examples may not appear specifically factive at all. The less so if you go by the sweeping generalizations (or, where silent on the topic, their corollaries) that have bedeviled theory in the disciplines concerned—about indirect quoting or propositional attitudes, as if theywere reducible to unity.Doesn’t this rule of self-accord hold, youmay then wonder, for all thinking (cognizing, perceiving, emoting) subjects?Doesn’t it hold for all speaking subjects aswell—qua fellow propositionalizers—across variations like, or within, indirectness? In brief, doesn’t one necessarily subscribe to whatever one thinks, or says, a fortiori when one samethinks, or samesays, it with the quoter? Yes, one does, except that necessarily doesn’t mean uniformly. As with the quoter’s own special commitment in factivity, lost upon the overgeneralizers, the tendency to lump together quotees presented in the same (indirect) form again blurs a vital distinction. To revert to my componential analysis, all thought-verbs encode a mentalizer; but the mind’s degree of confidence in what it thinks varies widely outside factivity and rarely ascends to the upper limit of ‘‘intensity.’’ Thus the descending scale from feel sure to believe to guess to daydream: the more subjective (or, if you will, the more perspectivized) the thought attributed, the less binding in regard to the thinking subject as well as the invariably noncommittal thought-quoter. The former may then silently add with progressive liberty the kind of disclaimers always open to the latter. In factivity alone does (or at least can) this variable inverse ratio of subjectivity/perspectivity to self-commitment turn straight and comparatively stable.Within the factive predicate, too, the mentalizer attributes the original thought to the quotee—the subject of the clause and of consciousness 198 Poetics Today 22:1 alike—yet the intensifier now binds the quotee to the thought as fast as the third, validating element does the quoter.What the one’s mind confidently deems true, we gather in advance, the other’s mediacy presupposes to be true and will next transmit with the presupposer’s endorsement. It being the subject’s assured mental representation that the inset so re-presents, he must co-shoulder the appropriate load—anything from the propositional to the whole burden, from ‘‘the facts’’ to the discourse as re-presented—unless we have grounds for inferring otherwise. And inferring otherwise means locating a significant distance, if not chasm, between the quotee and the original subject whose mind the indirect quotation purports to re-present, from world-image and truth commitment upward. Among the grounds for decreasing that quotee’s cosponsorship below the propositional threshold, all the way down to nothing, the predicate’s modalization first suggests itself by its overtness. It signals an element of uncertainty about the original representation, hence amore or less unequal partnership in the given re-presentation, from the surface to the views to the ties involved. Depending on context, the modal qualifies the very mentalizer (‘‘Did X think or believe anything of the sort?’’) and/or the intensifier (‘‘Was X sure, rather than only thinking or believing it?’’), to an extent that can shade off into the minus of negation. Thus, the ‘‘she probably knew’’ in () throws into open doubt not just whether the alleged knower silently refers to her father as a ‘‘phoney slob,’’ but even whether she does or would opt for a softer descriptive equivalent: indeed, whether she thinks, far less knows ill of him at all, but also whether the description on record is harsh enough to cover her knowledge of him. What Holden introduces as ‘‘probable’’ lends itself to reading as anything from an understated to an illusive commonality, from the marriage of true minds beyondhis dreams to a sheerwishful projection of hismind onto hers. This ambiguity about the filial attitude stays unresolved.And in operational terms, as long as it does, she may proceed to express (say, think, factivize) anything about her father without contradicting ‘‘herself,’’ that is, her hypothetical inner selves. No ensuing discourse in her own name can be judged a breach of what we have now registered as her mental epistemic commitment—or, strictly, falsify his qualified commitment about hers. Good, bad, or indifferent, her expression will have been anticipated somewhere along the range of possibilities (including the quoter’s ‘‘probably’’) that apply to her early cognition. Her commissive record on the topic is as openly open as the modal leaves it; and were she a Queen Elizabeth, or a heroine in a James novel with a likewise troubled filiation, the gap’s interest and the pressure for (dis)closure would mount accordingly. Even so, the indeterminacy stops well short of Holden’s own position. Sternberg • Factives and Perspectives 199 While qualifying his friend’s knowledge, he remains unreservedly pledged to the truth of the inset, ‘‘phoney slob’’ and all, as well as to the likelihood of cosponsorship: to the entire given stretch of indirect discourse, in brief, from modalized assertion to categorical presupposition. (Against Wilson : –, who fails to unpack the viewpoints.) This brings us to a still wider and deeper reason for variations in the quotee’s imputed or inferred truth-bond, as for entanglements in perspectival ambiguity. Why does Holden see fit to weaken the predication cum partnership by modalizing it? The constraints of the human condition—of access to other minds, above all—are no doubt at work. Less charitably, he (or, if too innocent, his fellowmodalizers) wants to compound the epistemological decorums of realism with rhetorical gains, or to invoke the former as camouflage for the latter. Mutatis mutandis, qualifying the claim of insight into another’s awareness (discovery, regret, bother . . .) enhances the credibility of whatever qualified epistemic claim one does make about the privileged secret life. And nowhere in thought quotation does realism so press or the enhancement come into such demand as in factivity, whose bare predicate already ‘‘intensifies’’ the truth claim: from the quotee’s mental attitude to a mental certitude on a par with the quoter’s own. There Holden in effect asserts, and duly hedges, nothing less than his knowledge (rather than, say, belief ) of her shared knowledge (ditto) of what he presupposes about the phoney slob in first-order reality. Such multilevel accord among humans can do with a grain of doubt. But modality is only a special case of variation in other-binding. Don’t we always check, against the appropriate norms, the degree and grounding of the quoter’s assertive commitment about the quotee’s like-minded commitment, together with the presuppositional self-commitment about the world itself ? In general, therefore, that degree and grounding need not find any articulation, encoded or otherwise, for us to seek it, if possible to supply it, against the factivizer’s wishes, if necessary. This checking works by appeal to the same two universal axes of point of view as have just been adumbrated apropos Holden: the mediator’s knowledge (from interior to possibly exterior in factive order) and reliability. Local clues apart, the predicative assertion as well as the inset presupposition depend for their truth value on where (if anywhere) our inference locates the factive truth-claimant between the extremes of insight and blindness, absolute trustworthiness and none. Insight versus blindness, I emphasize, not omniscience versus restrictedness, as traditional narrative theory would expect. I indeed reject its whole preconception of discoursive, or even epistemic authority, along with its low naturalistic counterparts in epistemology proper, history study, and 200 Poetics Today 22:1 related fields.42 My Proteus Principle, generally substituting a many-tomany for their one-to-one correspondence between form and function, once again predicts the true state of affairs: the insightful/blind or authorized/unauthorized axis (i.e., the differential epistemic value) has multiple operative surface correlates in discourse, of which omniscient/restricted is only one and not necessarily a fictional one at that. So, when the Biblical narrator, entrenched as divinely inspired historyteller in context, grants us the insideview, ‘‘Saul then knew that God was with David’’ ( Samuel :), all normal divides vanish together—at least in regard to the given propositional meaning or its truth valuation. Here, nothing separates objectivity from subjectivity, ontology from epistemology, first-orderworld representation frommind re-presentation, quoted from original thought, predication from presupposition, selffrom otherbinding. Further, though this omniscient will tell us less than he knows—the whole truth—he guards against misleading us (the way ironists like Fielding or Austen do) in what he tells. Narrative practice therefore goes here with absolute epistemic privilege to maximize reliability. (For details, see Sternberg a: ff., : – and, with specific reference to factivity, .) Validated on the highest authority, therefore, this ‘‘knew’’ establishes as a rocklike fact, and whatever the appearances to the contrary, not only the truth of the inset proposition within the represented historical world but also its truth to King Saul’s viewpoint on that world.Were he to deny it, he would incur self-contradiction, if to our eyes only, privileged for the nonce by courtesy, and God’s. It is beyond the jointly held propositional content that the ubiquitous question arises here: what is whose in the insetmontage? And the answer implied will most often reactivate the very irony-fraught distinctions neutralized at the basic, common level. It will, that is, oppose the quotee’s truth (partial, because humanly limited and/or selective) to the whole truth (e.g., about the God/David alliance) reserved for the quoter’s superhuman frame. On the other hand, omniscient intervention can turn out not only misleading (as with the ironists just named or the equivocators to be discussed) but also virtually superfluous. For all the contrast of these practices to each other—as well as to the Biblical norm—the effect is much the same in the . Heremy analysis complements the attack on the tradition of epistemology inRorty , except that it both reaches further and offers a viable theoretical alternative across communication: a turn from the hopeless (meta)discourse of epistemology, with its prescriptive unitary ideal, to the functional epistemology of discourse, endlessly Protean and atmost contingently regulated in practice. For how this dynamism affects the theory of narrative (un)reliability, see again the Yacobi references as well as the argument below. Sternberg • Factives and Perspectives 201 premium put on our inference. So, in (), ‘‘Freddie knew no such thing’’ rather clinches than uncovers to us the falsity of the predicative (‘‘You know’’) and the presupposed (‘‘you love’’) statement alike. For on the most earthly inspection, Freddie’s wife tops off the ordinary human liability to error with a peculiar (generically comic) mix of blindness, busybodiness, insensitivity, self-delusion that allows of no self-questioning.What with the husband’s commissive record thus far—on ‘‘love’’ included—her negative polarity discredits the entire quotation: factive and fact, quoted and original Freddie, re-presented and represented state of affairs, all draw poles asunder, foreseeably so. Were Freddie brave enough to protest, he would contradict her from the inside with the same epistemic impunity—because with an analogous privilege and weight vis-à-vis hers—as the omniscient teller himself enjoys in the actual negation of the ‘‘know.’’ The responsibility for the entire false insideview (and the inconsistence with the truth dramatized, given, utterable) is hers alone: a single perspective masquerading as a partnership. Conversely, humans sometimes factivize, as well as distance or ironize, the workings of other minds with an authority (and so commissive force) practically equal to an omniscient’s or to a Freddie’s self-knowledge. This would be admitted even by hard-line epistemologists in philosophy, with their fellow dogmatists in pragmatics, historiosophy, and literary study. Except that they would condition the insight on some pregiven ‘‘epistemic warrant’’—anchored in a favorite culture-bound version of naturalism or rationalism, hence alien to the boundless dynamics of (quoted) discourse— instead of attaching, detaching, or refashioning the strings of warrantability according to the operative contextual norms and premises. Although (), for example, bears a marked resemblance to ()—the stronger because the factive’s subject doubles as its addressee—they otherwise contrast all along the line. In our eyes, and the dramatic observers’, the reference made by the prosecutor to the witness’s knowledge commits the latter to the whole insideview, from transformer to inset, from predication (‘‘You knew’’) to presupposition (‘‘your friend had lied about it’’).43 We register the claim as a truth, the factive as a fact (here, within the fictional world).What’s more, and still less ‘‘realistic,’’ its other-commissive force outranks the witness’s divergent expression of self-knowledge in anything from propositional to verbal load; and the witness finally admits as much. The inside-viewee revoices and assumes verbatim the uncongenial commitment placed on him by the inside-viewer from without. Most notable, the outsider’s outknow. Though formally a question, the cross-examiner’s reference has the force of an (elliptical) assertion: ‘‘I put it to you that you knew. . . .’’ 202 Poetics Today 22:1 ing the insider contrasts with the failure of knowledge in () since, ‘‘realistically,’’ an old married couple should have better access to each other’s minds than a pair of strangers. And if the barrister’s insight yet lends itself to motivation by appeal to empirics—native wit, professional experience, the tenor of the evidence—it has equivalents in life and literature that issue from pure intuition: diametric warrant, analogous weight. Pure convention may be stronger yet, qualitatively so: what your authorized spokesmen (e.g., lawyers) assert on record about your mind binds you along with themselves, the way an omniscient would. In factivity, as elsewhere, discoursive authority varies athwart and against as well as with realism, a fortiori the shifting standards thereof. An affirmative indirect quotation once associated with either extreme of authority, then, our advance from the basics to the variables of factivity is essentially straightforward. The factivizer then makes two selfcommitments, one objective and presuppositional, one subjective, assertive, and implicating a like-minded other. Objectively, as presupposer/ validator, he binds himself to the truth of the inset re-presentation (thatGod was with David, that the accused lied, that Freddie loves pedicare, etc.). Subjectively, as quoter of the original representation, mental and intense, he binds himself to the truth of the insideviewof the quotee (Saul’s inner certitude about God being with David, the witness’s about the accused having lied, Freddie’s about his love for pedicare).Hence the subjective bondwould commit the quotee with equal force to the objective state of affairs allegedly represented by him and now presupposed in quotation. What reverses between the extremes is not the truth claims made but their truth value. At the authoritative pole, all the factivizer’s commitments, objector subject-oriented, undertaken or imputed, accord with factuality (i.e., what counts as such, even in fiction); at the opposite pole, none do, least of all the imputed, second-hand one that presumes to view and fetter anothermind.The clear-cut results between truth and falsity, however,may leave a host of secondary questions open to nicer inference. Do the paired images of truth—the quoter’s and the quotee’s—match in depth, extent, consciousness, semiotic form, attitude beyond the propositional? Does the array of untruths yet contain elements defensible within a wider frame or less fact-bound epistemology? Variety in basic unity, local or higher accord in discord. Many (most?) of our inferential processes, however, range and shuttle between these extremes, often forever, without a determinate end-product to show for our trouble. The factivizer being neither here nor there, the hard questions no longer start at the level of relative detail or even stop with the Sternberg • Factives and Perspectives 203 (dis)ambiguation of objective factuality. Instead, like the rest of the equifocused montage, the other’s equal burden and the distance separating the given inset from the original state(ment) also turn ambiguous as a whole. In (), did Isobel subscribe to the view, let alone to the knowledge, of Ofield’s perfidy that the unempathetic detective imputes to her? Confronted with ‘‘I knew something had happened’’ in (), should we (along with the immediate addressee) trust the self-re-presenter’s profession of ignorance at the time? In flattening everything into the bare narrative minimum, from topic to verb, from agent or patient to action, doesn’t he tell less than he knew? In (), epistemic not only crosses ontic ambiguity, as usual, but raises it to the level of genre. Is the presupposition of existence carried by ‘‘Charlie Heller,’’ then that of truth factivized by ‘‘people are aware of what he did,’’ geared to the real or to a novelistic world? With which kind of operative claims, hence ultimately values, are we to associate the subject’s awareness of/and the existent’s doings in the three-point shuttle among the factual, the false, and the fictional? (The author’s Foreword indeed proceeds to entangle the issue.) And so forth. But the extremes themselves, of course, are nomore given (and their constituents are therefore no more ‘‘projectible’’ by rule) than the intermediate equivocalities. Our inferential activity en route to the latent presuppositional inference accordingly stretches from transformer to inset, from the montage of viewpoints in general to its truth-pledging aspect and balance, from truth claim to truth value, from state(ment) to the stating discourse, from whole to particular, toing and froing among them all in search of the best integration.Where do the forces in play converge, where diverge, where hover between convergence and divergence? Everywhere, nowhere, here and there? Yet, whatever our answer, it will rest on one uniquely stable ground: the affirming factivizer cannot without inconsistency disclaim responsibility for some propositional minimum.What happens under threat of such inconsistency is (predictably by now, I hope) further inference designed to motivate, if possible to eliminate it—and the ultimate test of my argument. 5. Shifting Perspectives, Downgrading Presupposers: From (Indirect) Quotation to (Indirect) Quotation within (Free Indirect) Quotation To return to my beginning, since the early s presupposition theorists have agonized over (invented) conundrums of factivity like: () For all I know, Oedipus regrets killing his father, although, in fact, he didn’t kill him. (Cf. ()–(), ()–(), () above.) 204 Poetics Today 22:1 Inwhat is still among the best available pragmatic approaches,GeraldGazdar () observes: The unacceptability of this example can be readily explained if we allow the factive [i.e., ‘regrets’] to entail its complement. The argument goes as follows. The sentence as a whole—assumed to be a conjunction—epistemically implies . . . that the speaker knows that Oedipus did not kill his father, because of the second clause, but the first clause, and hence the sentence as a whole, entails that it is compatible with all the speaker knows that Oedipus did kill his father.These two implications are contradictory, hence the anomaly. (This is an informal rendition of what is a simple proof in epistemic logic.) (: ) 44 I would challenge every step in this analysis, from negative premise (‘‘unacceptability’’) to argument (even ‘‘proof ’’) to Q.E.D. It is, alas, a characteristic tissue of fallacies, oversights, misjudgments, and plain bad analytic habits, some already encountered, some newer to us. Briefly: (a) Left uncontextualized, against the nature (‘‘pragmatics’’) of discourse, the example yet invokes a situation (the Oedipus tale) whose apparent familiarity may predispose our understanding and judgment—or, as here, disorient them through unfamiliar twists. Exactly because the ‘‘speaker’s knowledge’’ plays such a role—as against the subject’s and ours—what is actually known, what foreknown, what unknown? Why, wherefrom, to whom, within what referential frame? Thus, of the ‘‘two contradictory implications’’ alleged, which holds true in the example’s discourse world: Oedipus as traditional ([fore]known) parricide or as an untraditional innocent? If the former, then the factive survives after all, and it is the counterstatement that becomes the problem and demands resolution (e.g., ‘‘. . . he didn’t kill him but merely caused his death’’): the imagined ‘‘unacceptability’’ finds its target, one other as well as more localized than the analyst singles out. Such unstable world-coordinates make a principled difference, if only because of the need to keep apart internal, context-specific from extratextual truth. Seeing howoften the two get conflated, in and out of presupposition theory, we must remember that factivity can bear any relation to external (natural, historical, scientific) fact: . More recent parallels would be the claims that twinning ‘‘John regrets that he lost his job’’ with ‘‘John did not lose his job’’ (Chierchia and McConnel-Ginet : ) or ‘‘Elspeth knows that Fred is happy’’ with ‘‘Fred is, or may be, unhappy’’ (Beaver : –) must deeply offend against felicity, compatibility, truth-conditionality, and the like. Conversely, some would deem a redundant follow-up, like ‘‘. . . and he killed his father,’’ no less anomalous (‘‘ill formed’’ in Bickerton : ff.; ‘‘incoherent and incomprehensible’’ for Van der Sandt : –, –): my counterargument below readily extends throughout to this polar terminus. Sternberg • Factives and Perspectives 205 () Mr. Camilla came to the rescue at this point, and said in a consolatory and complimentary voice, ‘‘Camilla, my dear, it is well known that your family feelings are gradually undermining you to the extent of making one of your legs shorter than the other.’’ (Dickens n.d. []: ) However fantastic Mr. Camilla’s complimentary complement, he stands bound to it as factivizer: the incompatibility of the ‘‘well known’’ with the foreknown laws of naturewill be integrated bymechanisms (e.g., existential, generic, psychic) other than the epistemic. Again, presupposed and ‘‘common or mutual’’ knowledge, strict and loose ‘‘presupposition,’’ are worlds asunder, as they always are in reason. (b) Even more typically, Gazdar overlooks factivity’s quotational status, bi-perspectivity, distinctiveness, let alone variations, hence maneuverability. He officially treats the entire two-part example as a matter of the ‘‘speaker’s’’ knowledge and epistemic consistency. Officially, since his own analysis at times betokens otherwise. Not onlymust he be aware that ‘‘Oedipus regrets . . .’’ introduces a subject and amental state; he elsewhere adverts to them himself in claiming that the sentence likewise ‘‘entails’’ that ‘‘Oedipus knows that he killed his father’’ (Gazdar : ). But the fragments remain uncorrelated and untheorized—not even into a putative coupling of the two ‘‘entailed’’ knowers, which would amount to a glimpse of quoted discourse, or at least of nonunitary viewpoint. So aggravated by incoherence from within, this oblivion breeds further trouble. (c) The appeal to entailment, onwhich the argument rests, is odd, ad hoc, and untenable. In line with all ‘‘semantic’’ approaches to presupposition, Gazdar insists that ‘‘simple affirmative factive sentences’’ both entail and presuppose their complements. Nor indeed, more’s the pity, is this move ‘‘especially controversial’’ (ibid.: , –) in the rest of the field (not even among those who would dispense with the added ‘‘epistemic logic’’). But why conflate two distinct inference types, and in an analysis mounted to distinguish the lat(t)er at that?Why do so in the wake of the semantic accounts that Gazdar the pragmaticist otherwise so shrewdly demolishes? And why the selective forcing of themarriage on ‘‘affirmative factive sentences’’ only? The last wonder is the easiest to resolve: because presupposition opposes all other inference types, entailment among them, in its remarkable survival under negation, and yet Gazdar et al. want to keep it cancelable at need as well. Either way, the survivalist or the cancelationist, its behavior under negation plays havoc with entailment rules, and the desire to have it now this way, now that (as usual) enforces their confinement to the affirmative mode. Sheer or presuppositional entailment here, unentailed pre206 Poetics Today 22:1 supposition there. But if so, the halfway house, stopping short of the very distinctive negational feature, only worsens matters: why compound oddity with selectivity, indeed ad hocness? At the deepest, methodological level, the splicing of inference types betrays the continued stranglehold of philosophy, especially logic, on the very notion of what a theory of discourse should be and do. There the exciting idea of presupposition, as well as implicature and speech act, originated to constitute pragmatics, together with a false (because unattainable and unsuitable) ideal of systematizing language use that still misguides the heirs. In this regard, there is little to choose between traditional ‘‘semantic’’ and ‘‘pragmatic’’ approaches (to, say, factivity).Whatever their differences, either line would square the circle by reducing all inference to virtual automatism: to logical, if necessary quasi-logical computation, to ‘‘projective’’ formulas, to irrefutable ‘‘proof ’’ or disproof, to the either/or simplicities and certainties and in this sense trivialities of entailment. The formalist algorithmic spirit of the enterprise goes against the elusive realities and goaldirected regularities of communication. This hankering after strange gods generates, inter alia, the argument under scrutiny. Odd and selective, the appeal to entailment with(in) presupposition, we hear, ‘‘readily explains’’ the ‘‘unacceptability’’ of cases like (). According to Gazdar, if you take the opening clause as merely presuppositional, then the ensuing counterstatement would render it ‘‘merely infelicitous or inappropriate’’: too weak a judgment on the falsification, or mutual exclusiveness, incurred. Although or because presupposition is liftable (‘‘defeasible,’’ ‘‘cancelable’’) at a pinch, criticizing the speech act apparently does not do full justice to the outrage sensed here. But ‘‘if we allow the factive to entail its complement’’ as well, then ‘‘the first clause, and hence the sentence as awhole entails that it is compatible with all the speaker knows that Oedipus did kill his father,’’ while the ‘‘second clause,’’ hence ‘‘the sentence as a whole . . . epistemically implies . . . that the speaker knows that Oedipus did not kill his father’’: a head-on, either/or ‘‘contradiction’’ emerges. And since entailment, unlike presupposition by itself, is too hard to be ‘‘canceled’’ or ‘‘defeated’’ or judged ‘‘merely infelicitous’’ under whatever pressure, the ‘‘two implications’’ alike survive to contradict each other, breeding and explaining ‘‘the anomaly.’’ For the analyst, it goes without saying that such explanation announces the terminus, like the ‘‘merely infelicitous’’ judgment, only more definitely so: illogic outdoes illperformance.The sharper the anomalousness, the less acceptable its carrier and the deader the dead end. But then, the chain of reasoning snaps at the first link. This factive, or Sternberg • Factives and Perspectives 207 any other, never entails its complement, and we have already discovered why: because it entails quotation, precluding all further entailments.Thus, () does not entail that ‘‘Oedipus knows that he killed his father.’’ In the given quotation, as always, the inset’s perspectivalmontage hovers between quoter (Gazdar’s ‘‘speaker’’) and quotee, ambiguating everything, down to the origin of the referential terms. And if the term used here, as before in ()–(), is the quoter’s, then Oedipus (e.g., at the outset of Sophocles’ play) may well regretfully admit to having killed the old man at the crossroads while denying, and not in the least knowing, that he had killed ‘‘his father.’’ This supposed entailment fails the definitional test of being true in all possible worlds. Inversely, as hinted above, with the quoter himself. Whether he alone or also his subject ‘‘knows that Oedipus killed his father,’’ he need not contradict his own (or either’s) knowledge in the second clause: he may instead be drawing the finer, ancient, relevant line between killing and causing death. (Not to mention as yet possible readings other than indirect thought-report.) For better or worse, the logic of entailment vanishes, along with its binarisms and contradictory impasses, to leave the dynamics of presupposition on its own: never strictly contradictable, the inference always remains maneuverable even, or above all, in face of apparent inconsistency. The dead end opens up into life. (d) By the same token, the very premise of ()’s ‘‘unacceptability,’’ as entailment and/or presupposition, collapses. Small wonder the empirical evidence negates the premise afresh: the concocted putative anomaly has countless equivalents on record, throughout language use. What instead needs to be disentangled and ‘‘explained’’ is itsmultiple acceptability—even along lines that go beyond, or against, the sense-making via indirect discourse on which we have just drawn afresh. (e) With this in view, and in fairness to Gazdar, it should be pointed out that his halfway house of factivity suffers further division at the hands of other theorists. Splicing the two inference types, they yet split their joint applicability not only between affirmative and other sentence types but also between imagined factive subtypes.Gazdar himself names somewhowould mark off cognitive factives like ‘‘know,’’ taken to entail as well as presuppose their complements, from their emotive mates, like ‘‘regret’’ (, ). My counterargument above evidently holds for both. However, ‘‘know’’ being my paradigm anyway, the challenge of developing the positive argument through the allegedly stronger, cognitive branch is welcome. Take first these diverse and contextualized equivalents of () on public record: 208 Poetics Today 22:1 () At his machine gun in the stern of Hiyo, Chief Petty Officer Mitsukuni Oshita heard the cry ‘‘Torpedo coming!’’ He began to count. At  he knew the torpedo had missed, and relaxed. An explosion jarred Hiyo. Oshita had counted too fast. (Toland : ) () She [Mrs. Ramsay] saw his [Mr. Ramsay’s] anger fly like a pack of hounds into his eyes, his brow, and she knew that in a moment something violent would explode, and then—but thank goodness! she saw him clutch himself and clap a brake on the wheel, and the whole of his body seemed to emit sparks but not words. (Woolf  []: ) () I went down to my bedroom, drew the curtains, locked the door, turned off the light, went to bed knowing I should not sleep. I did not wake till morning. (Symons : ) () The E.M.P. [Exalted Military Personage] shook hands rather grandly and referred tome as ‘‘the hero of the hour.’’ He thankedme andRoss and Alice, but I knew there wasmore to it than that.When he began the sales talk, with ‘‘Mr. Ross is most anxious that you should hear this from me . . .’’ I knew what it was. Ross had finally taken over Charlotte Street.What timing! . . . But now the conversation was taking a different turn. Ross, it seemed, wasn’t taking over Charlotte Street. The purpose of the visit was an explanation to me! (Deighton : ) The otherand the self-quotations alike presuppose, literally ‘‘know’’ what the sequel undercuts beyond remedy. Each factive gets immediately falsified by the facts: the ‘‘known’’ miss of the torpedo by the explosion, the fore-‘‘known’’ verbal explosion by the nonexplosion, the certitude of insomnia by nightlong sleep, ‘‘Ross had finally taken over’’ by ‘‘Ross, it seemed, wasn’t taking over.’’ That none of these factive statements, any more than ()’s ‘‘Oedipus regrets . . . ,’’ entails its complement is indisputable. But then, scholarly bias and fiat apart, why should it in the first place? How will ambiguous discourse, made so by its inherent montage, entail anything? Bracketing or replacing presupposition with entailment only confuses the issue.What should rather trouble us here is the undercutting of presupposition proper. Why does the factivizer let himself (or herself ) in for such blatant inconsistency between truth commitment and negative truth value, between implicit pre-validation and explicit posterior invalidation of what the inset re-presents? But this violence does not yet bring the matter to the dead end so widely announced by analysts. It never dooms the presuppositions involved to Sternberg • Factives and Perspectives 209 unutterable or unreadable ‘‘failure,’’ much less presupposition in general to ‘‘cancelability.’’ Various loopholes and counterforces keep the quest for better inference going. For one thing, the discord stops short of the extreme of self-contradiction. Presupposed rather than asserted or entailed, the factive complement incurs the lesser offense of inconsistency with the sequel. As we have seen, under the Law of Reciprocity, even outright ‘‘impossible world’’ contradiction (e.g., []) lends itself to integration; but anything below it enables wider, tighter, and factive-saving possibilities. For another thing, rather than concocting an example and pronouncing it ‘‘unacceptable’’ at will, we encounter here actual discourse that must be accepted, breaches and all, as having been made in and for communication, presumably by a rational addressor. The armchair theorists’ dead end is therefore the addressed inferrer’s beginning, their verdict of inconsistency an added spur put to our reading, or rereading, for coherence: the bigger the threat, the higher the stakes and the more urgent the quest. Faced with the broken factive, we need tomake some sense of it, the best inferable from the (con)text: dismissal on sight with a negative tag is a nonstarter, just as it would be if addressed to ourselves in everyday talk or writing. More so, come to that, since the very breaches look deliberate.The opposition in immediate juxtaposition—in small compass, in the same breath— not only ensures our awareness but sharpens our sense of incongruity to maximize wonder. So does the retrospectiveness of the narrative: thanks to it, the quoting narrator could meanwhile, if he or she would, revise the factive to suit with the fact. (Substituting the unvalidated, hence deniable ‘‘believed’’ would do.) The two extras, pinpointing and after-the-fact(ive)ness, even jointly argue against genetic mischance. Everything suggests, then, that the quoter here courts inconsistency, forcing it on our notice. But to what end, and how to infer it from the given illogical discourse—or, the other way round, how to make sense of the discourse givens in its light? (Always leaving aside the nonperspectival integrating mechanisms, which would explain and organize the tension but not necessarily eliminate it, or not in terms specific to factivity.) The key to the accord of the inset truth claim (e.g., ‘‘the torpedo had missed’’) with the operative truth (e.g., the frame’s ‘‘An explosion jarred Hiyo’’) lies in changing the origin, and so the claimant, of the transformer (‘‘He knew’’): in the redistribution of its three semantic components between quoter and quotee, whereby they will all now fall to the latter, complete with the validator, and the inset statement with them. This new strategic unpacking of the perspectival montage entails in turn a shift in our reconstruction of the quoting schema: from plain indirect to free in210 Poetics Today 22:1 direct thought, ormore exactly, fromplain indirectness to plain indirectness within free indirectness. On the simple indirect reading presumed andpracticed thus far, thementalizer and the intensifier in ‘‘knew’’ belong to the quoted, the validator to the quoting viewpoint. This bifocal partition has worked well for all our earlier cases: whatever the dissonances encountered and the inequalities reasoned out there, they never approached the limit of basic propositional (factive/fact) inconsistency. Now, with the quoter apparently invalidating what he has just validated, the limit case for the first time becomes the source of trouble, multiple trouble, incurring a breach of presupposition and/or assertion, of re(-)presentational, perspectival, and commissive unity at once. Possibly welcome to the nonsense maker alone, it ill-fits the above genres of history writing, of high modernist novel, of realistic crime story alike: even the fallback on poetic license or superfinesse—the evader’s refuge, the literati’s hubris—is as usual blocked. A universal license instead comes to the rescue, namely, the (re)play of perspectives. Encouraged by the quoter’s show of deliberateness, we extricate him and ourselves from this predicament by attributing the validator along with the mentalizer and intensifier in ‘‘knew’’ to the quotee: Oshita, Mrs. Ramsay, the narrating-I’s former, experiencing self. Their minds (so the adjusted inference goes) not just originally held the respective thoughts with certitude but articulated that certitude—into an image like ‘‘I know the torpedo missed’’ or ‘‘I know that in a moment something will explode.’’What the directly quotedwife in () orMr. Camilla in () voiced aloud, Oshita et al. mentally expressed. Become the transformer’s originator and factivizer, the quotee now doubles perforce as indirect committed (self-)quoter within a larger and noncommittal and unadvertised frame of quotation, where the narrator re-re-presents to us the character’s validated re-presentation of the state of affairs (e.g., the torpedomissing) that the subject confidently, if erroneously, represented in private to himself. Noncommittal, this larger frame, because disembarrassed of the validator. The second-order quoting voice (the storyteller at mind report) can now pit factive against fact without inconsistency. The factive being another’s, its factual invalidation rather serves to oppose and expose lowerorder epistemic unreliability: to ironize the limitations of (fore)knowledge betrayed by the overconfident dramatic quoter and quotee. Of this frameoccupant’s two commitments as indirect quoter, only the subjective one remains in free indirectness, except that it now embraces a chain of subjectivity. All the free indirect quoter pledges himself to is that the lower subjects have between them originated, intensified, and factivized the worldimage re-re-presented in their name: that they have committed themselves Sternberg • Factives and Perspectives 211 to the truth or ‘‘objectivity’’ of their imagined knowledge state(ment). As to the authority, the interplay, the finesse, and other variations among the respective engagements—or viewpoints in general—they have already been outlined above and easily extend to the more complex, three-party set-up. With the added frame, the underlying narrative that makes sense of it all grows longer, tenser, richer than in plain indirectness. Noncommittal, again, this larger frame of quotation is here also tacit, unsignaled, being implied by the epistemic discord rather than manifested on the discourse surface—transformer-less, in brief. ‘‘He knew the torpedo had missed’’: a chain or Chinese box of report, variform at that, makes do with a single reporting clause.Yet this lack is no necessary condition of the form and is elsewhere repaired to ensure its visibility as an aid to understanding.45The more overt, unambiguous the free indirect pattern, the lesser the danger of our missing it, which in factivity would amount to mistaking the discourse for simple indirect report, complete with objective endorsement. Formally, the signaling devices meet this need by counterworking the root of the problem throughout ()–(), namely, the quoting clauses given and elided. One device relocates the initial, factive, double-edged transformer; the other provides—likewise inmid-discourse—the absent, higherlevel transformer. Here is relocation: () Mr. Farange’s remedy for every inconvenience was that the child should be put at school—there were lots of splendid schools, as everybody knew, at Brighton and all over the place. (James  []: ; also examples []–[], [] below) With the transformermidposed (or elsewhere endposed), the discourse rules out the simple indirect form, whereby ‘‘everyone knew’’ must precede and subordinate ‘‘there were lots of splendid schools . . .’’ At the same time, the past tense of ‘‘knew’’ points away from a generalized knowledge claim in the narrator’s own voice: it ties and relativizes the parenthetical instead to the Farange milieu, mind, dilemma concerning ‘‘the child.’’ If only by grammatical elimination, then, the free indirect reading suggests itself. For good measure, the parenthetical clause appears with an extra ‘‘as,’’ which strengthens its attachment, validator and all, to what Farange himself expressed: his original midposed appeal to common wisdom (‘‘as everyone knows’’) just got backshifted (into ‘‘. . . knew’’) in requoting. Hence the free indirect requoter—the Jamesian voice—shares no responsibility for either questionable claim, the transformer’s proposition about universal knowl. The coexistence of these variants, the signaled and the unsignaled, has long been established in the research of the free indirect style. (For the most comprehensive survey to date, see Fludernik .) Their intersection with factivity is another thing. 212 Poetics Today 22:1 edge or the inset presupposition about schooling, not even equivocally. What endures in the movement away from ambiguity between nondirect patterns is his subject-oriented commitment alone: to Farange’s having cited with approval ‘‘everyone’s’’ knowledge state(ment), a transparent excuse for getting rid of his inconvenient daughter. Any leftover what’s-whose indeterminacy concerns pockets within a determinate pattern. In other words, given the transformer’s midplacement, we no longer need any fact/factive discord to establish nondirect quotation at a double remove from the world, with appropriate nonvalidation of its objective truth. Such clues to discord here only sharpen the formal noncommitment into effective other-mindedness between quoter and requoter, as the jarring colloquialisms (‘‘lots of . . . all over the place’’) imply other-expressiveness to boot. The alternative signaling device interpolates for the purpose a second, usually nonfactive transformer: () This time, she was convinced, she knew what was coming. He was going to say she must give him up. (James  []: ) () When she got the letter she realized at once that a mistake, as she thought, had been made. She knew quite well that [contrary to what the letter stated] she had not backed the winner of the Oaks. (Cecil  []: ) At last, the enclosing free indirect and the enclosed indirect report have each its proper reporting clause: one nonfactive (‘‘she was convinced’’/‘‘she thought’’), parenthetical, and juxtaposed or independent, the other factive (‘‘she knew’’/‘‘she realized’’), initial, and subordinating. Amid the new arrival’s common midlocation, the differences between the examples affect our reading to a limited extent. In (), the nonfactive-before-factive order precludes what the reverse order in () temporarily allows on the way to the belated nonfactive: the simple indirect, all-commissive formation of the whole sentence (comparable to that of the ensuing ‘‘She knew quite well that . . .’’). Sooner or later, however, the added higher parenthetical disambiguates the form to the same noncommittal, other-committal effect as did its merely relocated counterpart in (). With the indirect factivizing left visibly unvalidated within the free indirectness, its predication and its presupposition may again both turn out false without inconsistency. So they do here: ‘‘she’’ neither ‘‘realized’’ nor ‘‘knew’’ the truth, for no mistake ‘‘had been made’’ nor would a demand to ‘‘give him up’’ be made—except in the subject’s overconfident ‘‘thought,’’ from which the larger omniscient frame vocally distances itself to underscore the irony. As either device signals free indirectness, however, neither quite equals Sternberg • Factives and Perspectives 213 the manifest self-distancing from another’s epistemic bond that attaches to a third possibility, namely, (factive) indirect within (nonfactive) indirect report. Examples, naturally invented, from presupposition theory would be: () Fred thinks Mary doesn’t know that she won. (Beaver : ) () Maria just told Sebastian that she knows that Nixon is dead. (Wilson : ; note the parallel to the initial cognitive verb above or in Wilson’s foregoing example.) Of course, presupposition analysis is hardly aware of the alternative, free indirect patterns. Nor, of course, though the one pattern exemplified by ()–() has received ample notice, do analysts study it in terms of quotation, let alone quotation within quotation. Even in face of such lacunae, it still passes understanding how often this safest and shortest route to noncommitment has beenmisread—on formal grounds too—as preserving the ‘‘speaker’s’’ commitment. If your ‘‘speech’’ about another’s mere ‘‘thought’’ (belief, desire, hope) leaves you unbound to the thinker’s ensuing complement, as in (b), why should his ‘‘thought’’ about the object of a specific (two-clause, thinker-factivized) complement bind you? What have I to do with the presupposition of anyone whom I refused my factive predicate and then consigned to his own? Against logic and language, you and I must allegedly, somehow, carry the burden nonetheless. Thus, among semantic approaches, Leech (: –): a factive having once assigned ‘‘the status of a presupposition’’ to a that-clause, it remains ‘‘unaffected by a neutral factuality in a higher predication.’’ Or, recast into pragmatic terms: given a factive, ‘‘the presupposition should be expected to hold not only at the highest accommodation site, but also locally.’’ So (), which has the trigger ‘‘ ‘knew’ embedded under a negation operator, itself embedded under ‘thinks’,’’ would presuppose ‘‘that Mary won’’ as well as ‘‘that Fred thinks she won’’ (Beaver : , with earlier references). Another false projection rule, in short, and sometimes not even lawlike on its own rationale at that, because ‘‘context’’ may turn it either way, or neither.46 Instead, along with the alternative forms of quoting within quoting, only more so, this one does for a change enforce an iron rule—of noncommitment on the higher quoter’s part. Unlike the looser, juxtapositional grammar of ()–(), say, here the nonfactive directly controls, infects, hence subjecti. Other divides within this form includeWilson’s (: ff.) ‘‘preferred’’ (i.e., implicationpreserving) versus ‘‘nonpreferred’’ interpretation, and Grice’s (: ) ‘‘know’’ (factivity lost) versus ‘‘regret’’ (factivity possibly kept). For accounts that would perhaps get ()–() right,more or less, thoughdefinitely not ()–(), seeKartunnen ;Kartunnen andPeters ; Van der Sandt : esp. ff. How the problematics carries over from thought to utterance verbs in the higher sentence, or frame, will emerge below. 214 Poetics Today 22:1 fies the factive—just as it would any ensuing proposition and regardless of ‘‘context’’ other than the all-determinative manifest cotext (e.g., ‘‘thinks’’). The wider context at best merely implies whether or not the uncommitted indirect requoter nevertheless endorses de facto, contingently, what the indirect quoter (‘‘thinker’’) presupposes in factivizing the state(ment) of the innermost quotee (‘‘knower’’). () There was a question that must be asked. He knew it and he fancied Peter knew it too. (Gilbert : ) () I persuadedmyself that I knew hewas taken. (Dickens n.d. []: ) () She knew or hoped she knew that it was not being a misery or defeatist of her to imagine [the sexist remarks] that could follow her exit. (Amis : ) In upper-level, requoting context, nonvalidation again gravitates at will toward de facto validation of the factivized statement () or toward invalidation () or toward neither (): the question proves to be a must, the taking a false alarm, the nondefeatist imagining a gap.What is built into the pattern (lexis, syntax) is built-in, what contingent irreducibly contingent, and so the most unprojectible of unprojectibles. As the array of signaled variants of quotation-within-quotation are all by nature inference-facilitating, let us concentrate on the unsignaled pattern— especially the range and regulators of difficulty open to its own silent workings. What I call the opposition in juxtaposition throughout ()–() offers the most pinpoint, hence arresting indicator of the shift in epistemic center and commitment toward the fallible quotee. Its high salience there derives from themeeting of at least three forces: the immediacy, the explicitness, and the factual basis of the discord resolvable by the switch to other-minded, otherwise indirectly-quoted presupposition. But these discourse axes remain independent, hence freely graded or crossed. Each factor thereon may itself vary all the way to virtual disappearance, and with it the attached salience, pressure, guidance in the labyrinth of ambiguity. Is less (axes, degrees) better or worse? Opinions would divide between uncertainty-favoring literati and adherents of the Cooperative Principle. Either would reify the protean life of (factive) discourse that the above breakdown quickens anew.This liveliness makes no value judgment on the quoter’s repertoire or choice of aids to inference but only complicates our arrival at firm, if any, judgments as to which of the data belong to whose perspective and commissive record, in what transmissional shape, how reSternberg • Factives and Perspectives 215 liably, and why. Instead of better or worse under variation, the effect is just different in the process and often the product of sense-making. On the first (‘‘immediacy’’) axis, this happens when the inconsistent members—the presupposed and the asserted world-images—draw apart in location: () [In my dealings with Estella,] I was always restrained—and this was not the least of my miseries—by a feeling that it was ungenerous to press myself upon her, when she knew that she could not choose but obey Miss Havisham. (Dickens n.d. []: ) Compared with ()–(), the factive’s divorce from the quoter, Pip the older narrating-I, marks here a change in time, and so in accessibility, to our disfavor.The ‘‘knowledge’’ (mis)attributed toEstella (that their common patroness intends her for Pip, willy-nilly) no longer bumps against an immediate counterstatement. Nor do we enjoy any vantage point of (hi)storied foreknowledge to the contrary (of the kind that [] equivocally assumes). Even the shrewd reader will therefore at most suspect the presupposition’s objective groundlessness, and its origin in the free-indirect wishful thinking of young Pip, but will have to wait for overt retrospective (dis)confirmation, long after the factive. When it comes, though, this (dis)confirmation takes as specific a form as the presupposition to match at last, behind time, the explicit force of inconsistency in ()–(). Yet the explicitness, or givenness, may also lessen. At the opposite limit, we need to dig for the contextual ontic premises against which to judge, (re)organize, (re)assign the factive givens: () Jammy [a newsman out of favor with the police] consigned them all to perdition.What did the Yard want to take it like that for? Everyone knew that what you wrote in the paper was just eye-wash.When it wasn’t bilgewater. (Tey : ) () Wallenstein had the recklessness of one who knows that all things are predestined, that fate is written in the stars and cannot be changed. (Huxley : ) () Jenny Pargiter was the information officer . . . and she read [out her update] without hope, secretly knowing that it was the lot of any woman, when imparting news, not to be believed. (Le Carré : ) () While in the recesses of the Escurial Philip II pondered the consequences of Mary’s death, chafing on the fringes of the Court at Greenwich was a sailorman [Francis Drake] who had known long before Philip did that God’s plan was for war between the two of them. (Mattingly  []: ) 216 Poetics Today 22:1 () She [Emily] lived alone since her father’s death, without even a pet, holding that dogs bring in dirt on their paws, while cats scratch the furniture. As for birds, everyone knows they encourage mice. (Gilbert : ) Whether anything here jars at allmust depend on theworldview that frames the knowledge claim. Do you reconstruct the official quoter (novelist, historian) as sharing with the quotee (Jammy, Wallenstein, etc.) the factivized generalization: about newspapers, things predestined, the credibility of women, the accessibility of God’s plan to the chosen, the linkage of birds and mice? If yes, then you have in effect (regardless of your terminology, even self-awareness) opted for harmonious indirect discourse that enjoys contextual validity.47 If not, as most readers would presumably infer in context, then you switch hypotheses to unreliable, quotee-focusedand-factivized indirect discourse within a higher, free indirect quotation that silentlymocks it.Themockery overtakes whoever endorses the pseudoknowledge, from subject to presupposer to gullible understander. So, true or untrue? Re-presented or re-re-presented nondirectness? Gnomic or ironic montage? I am less concerned to adjudicate between the diametric readings than to establish the implicitness of the very incongruity as well as of the resolution that they attest. The factive, the form, the functionality: all polarize in accordance with the ontic norm taken to govern the represented world, with the operative logic of fact, in short. This brings us to the third axis.Whether immediate or removed, overt or implicit, specific or generalized, the counterforce to the factive hitherto lay in factuality proper: what the (re)quoter holds true about the world. Elsewhere an assortment of alternative, extrafactual inference bases, or combinations, predominate to the same two-edged effect: integrating (the whole) yet divisive (of its viewpoints), simplifying (the matter, the product, the followability) yet complicating (the transmissional manner, the process, the route), disambiguating yet short of certainty and possibly reversible. Again, none of the inferential bases is unique to factivity, in any of its quoting molds, but all uniquely dovetail the shift of the factivizer’s identity and presuppositional load with that from indirect to indirect-within-free-indirect quotation. And the entire coordinated shift again requires no more, and no less, than our puzzling out an extra, higher, unspoken, self-distancing (if only noncommittal, e.g., ‘‘x thought’’) transformer in the name of the overall quoter: . Or even, as with (), harmonious free indirectness that is equi-focused and equivalidated in propositional content but expressively focused on the subject (e.g., ‘‘eye-wash’’ sounds like Jammy rather than the narrator). For the weak bearing of expressivity on truth value that enables here either nondirect accord, see below. Sternberg • Factives and Perspectives 217 () [Amaidservant] had been in a bit of trouble some time back. Lightfingered, that was her difficulty. . . . Mrs. Prendergast was ready enough to keep her. ‘‘It’s all right, Annie’’ she said.You can be sure we shall never throw this up against you.’’ But, of course, the girl didn’t stop. ‘‘It doesn’t give me a chance,’’ she said. ‘‘If anything’s ever missing, they’ll know it was me.’’ (Gilbert : ) () It was their relation [Mrs. Ramsay’s with her husband the philosopher], and his coming to her like that, openly, that discomposed her; for then people said he depended on her, when they must know that of the two he was infinitely more important, and what she gave the world, in comparison with what he gave, negligible. (Woolf  []: ) () The paintings that made Franklin’s life palatable to him, that opened up a channel to a resonant past, Annie knew instinctively to be treacherous impostors. . . . She was true to the culture she was born into, truer than [Franklin] Todd who has abandoned it. She could cry at pop tunes and laugh at Yellow Pages ads. Her sloganeering, her mangled proverbs, her utter incomprehension of irony, her ability to recite ‘‘Buckle up for safety’’ as if it were a PaterNoster, marked in her genuine humanism. (Powers : ) In example (), the clue to other-minded factivitymigrates altogether from the domain of fact to that of reason, the speaker’s and ours. For Annie to validate ‘‘they’ll know it was me’’ in the hypothetical scenario (‘‘If anything’s ever missing’’) would be to defeat her own argument against staying—the reformed thief dogged by her past—as well as to incriminate herself. In () the basis for inference gravitates to socioethics: few do or would take Virginia Woolf as an indirectly discoursing party to the heroine’s low ranking (actually self-ranking in the name of ‘‘they [who] must know’’) visà-vis her husband. And () appeals in turn to aesthetic (mis)judgment. ‘‘True’’ as the subject is to herself and her inherited culture, ‘‘truer’’ than her lover Franklin, pop tunes and Yellow Pages ads do not equal, still less discredit in context his beloved old paintings that she ‘‘knew instinctively to be treacherous impostors.’’ Exactly because the value scheme and irony are fine, beyond her comprehension, the polarizing of cultures invalidates (‘‘subjectifies’’) the fancied knowledge of right and wrong in art. This goes to establish the range of bases for our inference of or between factive nondirectnesses: factuality proper, rationality, sociomorality, aesthetics. (Example [] fromGoodman adds metaphysics, trained on the philosophy of fact, it so happens.) Such bases all constitute norms against which we (re)read the operative form along with the function: their obser218 Poetics Today 22:1 vance in the data here points one way—the simple and validating—while their breach pulls another, more intricate way. Yet I advisedly, and I hope noticeably, omitted at least one further inferential basis, or signal type, with a view to another key distinction regarding factivity. The absentee from my typology that will look most conspicuous to experts in reported discourse is the ‘‘expressive,’’ language-bound indicator of subjectivity, the very type that the field has intensively pursued (to my mind, overprivileged) since the late nineteenth century. True, even in the examples dominated by extralinguistic norms, the inference of the underlying quotational setup is less than, say, purely fact-based. Note how () and () distribute, within and around the factive construction, expressive markers that would appear to suit best the quoted mind at work.These include figuration (‘‘anger fly like a pack of hounds . . . explode . . . clap a brake on the wheel . . . emit sparks’’), loose conjunction (‘‘his eyes, his brow’’), indeterminate reference (‘‘something violent’’), exclamation (‘‘thank goodness!’’ or ‘‘What timing!’’), subject-oriented deixis (‘‘now . . . was taking’’). Inversely, our examples display objectifying markers that point the wrong way, for they bespeak the quoter’s redoubled validation of a nondirect inset that he in all likelihood ironizes.Why does () break the sequence of tenses in ‘‘one who knows that all things are predestined, that fate is written in the stars and cannot be changed’’ after ‘‘Wallestein had . . . ,’’ or () in ‘‘Everyone knows they [birds] encouragemice’’ after ‘‘Emily lived ’’? The explanation that the unbackshifted present indicates the knowledge’s ongoing applicability (‘‘presentness’’) springs to mind. It even draws extra power from the statement’s general nature, associated with the gnomic present, and from the tense’s coverage of transformer and inset alike: the knowing and its object reinforce, as it were, each other’s endurance. If so, the quoter superobjectifies the quote, with supercommitment to match. But these extras, in all likelihood, only thicken the irony, perhaps baiting the trap for the unwary (i.e., indirect) reader, as the subjectivities elsewhere underline the distance from the misfactivizing quotee. Still, I have not deferred mention of the expressive indicator-type just to redress the traditional imbalance in its favor. Still less is its downplaying here grounded in its being liable tomislead (as do the objectifyingmisdirections above). For, by the nature of inference, all bases and signals partake of this liability. Epistemics, ethics, aesthetics, or their reconstructions from the discourse, are notoriously elusive, indeed two-faced in principle, to go by our examples; and why should Annie in () drive a rational argument when Hackett in () travesties logic? My deferral has its ground not in the elusive results of this signal type’s application, but in what eludes its power altogether, namely, to mark off Sternberg • Factives and Perspectives 219 objective from subjective factivity, validated from invalid presupposition. All other bases of inference enable judgments of truth or falsity on the presupposed complement, relative to what the frame deems valid: actual, logical, just, or valuable. Hence the factive’s quotational form and, above all, force hinge on its consistency with the appropriate norm (or what we take as such, rightly, wrongly or disputably, always probabilistically). In contrast, the norm of expressivity cannot by itself yield judgments on this matter: expressive style and epistemic worth vary regardless of each other. A vivid (low, emotive, broken) idiommay go with a valid, as a neutral or wellformed one may with an ironized, presupposition; and vice versa. Nor is either idiom the property and hallmark of some nondirect (any more than of the direct) reportive form. Strong expressivity can therefore encourage a subjective reading, for better or worse, for plain or complex nondirectness, without yet determining factive subjectivity. Even in helping to unpack the free indirect montage between quoter and quotee, it must leave the propositional disambiguation to the workings of other norms. More generally, this defines the relations between truth value and quotational pattern in factivity. Where we hypothesize ordinary indirect discourse—as silently done throughout the foregoing sections—the presupposed complement must be true to the frame or in the frame’s world. But not vice versa, for such truth is by itself equally compatible with free indirectness, and the decision between quoting patterns will then hinge on the best overall fit, expressiveness included. The principle shows as early as Biblical factivity.48 The difference there between indirect and free indirect thought correlates only with that between grammatical subordination and the vivid subjectifying marker ‘‘and behold’’ at the head of an autonomous inset. No other strings are attached, no other differential bases or aids regularly provided. On the contrary, either form of nondirectness selectively matches the factive with factuality, since this narrative poetics of inside-viewing would rather oppose the narrator’s (quoter’s) whole truth to the character’s (quotee’s) truth than polarize truth with untruth. Modern equivalents at times follow suit, even in the reputedly ironic form as harmonized at will by, say, Virginia Woolf: () But this was our way of knowing people, she thought: to know the outline, not the detail. . . . She knew him in that way. She knew he had changed somehow. ( []: ) () And all the time, he knew perfectly well, Dalloway was falling in love with her. ( []: ) () Her father was dying there, Mrs. Ramsay knew. ( []: ) . See Sternberg , b, : esp. –, –, , : e.g., , , . 220 Poetics Today 22:1 All as true for Woolf as for her reflectors. In short, where the presupposed complement is judged false to the frame, we hypothesize free indirect discourse—thereby saving the higher quoter from inconsistency—but the converse does not necessarily hold. To unify the two subrationales. For consistency, where the indirect quoter must, the free indirectmay ormay not, authorize the factive statement (e.g., the knowledge presupposed as well as predicated of the subject). Or, the other way round, from the understander’s own viewpoint on the factive statement encountered: where taken as disauthorized, we must read free indirectness, and where authorized, we may read either nondirect pattern, unless extrapropositional variables (style, transformer location) tip the balance.Throughout, wherever the factive transformer (e.g., ‘‘knew’’) leads the way, the free indirect discourse entails an indirect quote within it, whose lower-order quoter undertakes the same epistemic bond to his inset (now become his own presupposition) as does his first-order indirect counterpart elsewhere and by the same logic. But then, the nonsense artist undermines at will the premise of consistency itself, with the entire chains of reasoning that branch out from it, to suit the (il)logic of an otherworldly discourse and existence. So the quest for the best presuppositional fit again typifies our whole affair with discourse in having its rules, routes, routines yet nothing like formulas, algorithms, package deals, not even at its most orderly.With presupposition uniquely encoded in the language system (e.g., factivity’s validating component), it is unerasable in language use but also unassignable there by appeal to that decontextualized systematics or any equivalent outside language: the inference toward whoever validates the encoded inference must operate with whatever resources and regularities are afforded by the protean discourse system, to which quotation belongs. A busy and risky life, the inferrer’s. Nor indeed do the busyness and the risk end here, since the possibilities of factive ambiguity and multiformity don’t—or the inferential way round, the measures for shifting a troubled factive commitment to a perspective (subject, ‘‘context’’) other and lower than (because framed within) that of the apparent nondirect quoter.To return to the disputedOedipus exemplar (), we have already seen two lines of unpacking the perspectival tangle— via indirect or free indirect thought—so as to make ‘‘Oedipus regrets [that he killed] his father’’ compatible with ‘‘although, in fact, he didn’t kill him.’’ Along neither line would Oedipus regret, or know about, his parricide. On the near-harmonious indirect reading, instead, Oedipus as regrettor stands committed only to his having killed the obnoxious old man; while Sternberg • Factives and Perspectives 221 the quoter of his thought, who knows better and therefore substituted a coreferring term in transmission, endorses the more accurate propositional variant thatOedipus killed ‘‘his father,’’ yet quickly adds a corrective, differentiating, mitigating argument from ignorance. Nuanced validation from a superior viewpoint, factivizing while doubly perspectivizing the shared bare fact: the inferential maneuver takes longer to analyze than to execute. (Compare [] above.) On the unharmonious free indirect alternative, the quoter’s own discourse never presupposes (let alone entails) even that Oedipus killed anyone (e.g., some old man, let alone his father), nor that either he or Oedipus ‘‘knows’’ about any such killing. This Oedipus would merely think he did and does, to the extent of factivizing the crime in his own consciencestricken mind, ‘‘I regret that . . .’’: an emotion that the free indirect quoter transmits with a due backshifting of person (‘‘Oedipus . . . his . . . he’’) but no propositional blessing. Quite the contrary, as the belying (rather than nuancing) follow-up makes explicit. The frame’s double-edged attitude, on the indirect-thought reconstruction, now sharpens into polar epistemic irony. The example’s unfortunate quasi-historical reference may predispose you for or against either reading, or both.To make sure, therefore, recheck those lines of counteranalysis against my own varied illustration above, or against more neutral examples invented by the same opposition: ‘‘John regrets killing his father,’’ ‘‘John regrets that he failed’’ (Gazdar : –, –, ), ‘‘John regrets that he ate all the pudding’’ (Levinson : ), ‘‘John regrets that he lost his job’’ (Chierchia and McConnel-Ginet : ). And while you’re about it, check () itself for yet another, third omnipresent possibility of integration, with a twist in the factive’s reference between outer and inner quoting. Like all its analogues, the example at issue may in principle equally read—what with its typical divorce from context—as the re-re-presenter’s free indirect speech-quotation of another’s (here, the confessional Oedipus’s) indirect thought-quotation that enacts and validates himself as regrettor (here, Oedipus the emotive subject).We then fill out the givens into ‘‘Oedipus, so he declares, regrets that he killed his father,’’ and accordingly neutralize or extricate them afresh out of inconsistency with ‘‘although, in fact, he didn’t kill him.’’ Oedipus’s silent regret is again twice mediated but now also twice voiced and addressed. First, Oedipus the self-penetrator confesses and presupposes his regret, in the overt nondirectness attached to factivity; then, hiddenly and noncommittally, the higher frame transmits that confession to ourselves. The chain of quoted discourse now runs from the original subject as 222 Poetics Today 22:1 experiencer to the indirect vocalizer cum factivizer of the experience to the free indirect re-presenter of the vocalized re-presentation and re-representer of the experienced original. The first in line may think, ‘‘I wish I left the old man alone’’ or ‘‘A terrible deed, killing one’s father’’ or something unverbalized to this effect. The second is alleged to voice ‘‘I regret that I killedmy father’’ or its near equivalent in self-disclosure and -binding. The third never commits himself to the factuality of the deed, or even of the thought, but just to the second’s voicing, which he backshifted into the distal free indirectness ‘‘Oedipus regrets that he killed his father’’—and whose presupposition he can therefore (again) belie at will in his own name.49 In the process of inference, of course, we work out the chain in reverse: from the artful primary quoter in our own frame (deemed free indirect re-quoter, hence objectively commitment-free, on pain of absurdity) to his vocal quotee, who in turn plays indirect self-quoter and misfactivizer (but no self-controverter) vis-à-vis the ultimate, regretful quotee. Allegedly plays, again, since we only have the higher-level quoter’s bound word for it. And although the word about another’s word—as against the thought discussed thus far, including that other’s own putative sorrow—essentially lends itself to empirical checking, it remains subject to the norms of authority in force.Whatwe encounter here is always a chain of representations that successively claim factual status, one that only changes transmissional pattern and length in distanced, second-order factivity. The shift of inferential ground (hence of setup, perspective, commitment, effect) need not, then, end with the form of quotation, between indirect and free indirect thought-report, but may advance to its object, between thought and speech: private versus public (mis)factivizing on the reportee’s part. The same joint metamorphosis is open to the cognitive branch of factivity—often pronounced stronger than the emotives, because it runs to entailment, or otherwise different—with know at its head. For another change, let us explore and test the juncture against genuine usage. What we need to test is my claim that the shift of the requoter’s object from quoted thinking to uttering mainly affects the product rather than the process of factive inference as theorized so far (and the shift from emoting . He could also, or instead, freely go on to deny (as he could endorse) the utterer’s selfpredicated regret: ‘‘Actually, Oedipus neither feels regret nor killed’’ or ‘‘. . . feels regret but didn’t kill’’ or ‘‘. . . doesn’t feel regret that he killed,’’ always with the exception of the inconsistent ‘‘. . . doesn’t say he regrets,’’ which would make epistemic sense only if the earlier ‘‘regrets’’ predicated thought. On the latter free indirect understanding, of course, the analogous shielded exception becomes ‘‘. . . doesn’t feel regret’’:much the same range of deniability, hence irony, with the variation appropriate to the given transformer’s movement between exterior and interior Oedipean subjectivity. Sternberg • Factives and Perspectives 223 to cognizing, not even the product). To keep those objects distinct, hence under maximum control, we will also open with signaled utterance, comparable to ()–()’s signaled thought in its status as a given. () Mary, who had been a silent witness to the confrontation, said to the boy [who had just joined the police] that he knew his father like many miners regarded the police as enemies. (Symons : ) () [Asked for the royal dowry by the council of Henry VIII, the Spanish ambassador] Fuensalida took alarm, and when alarmed, he blustered. The council knew very well, he shouted, that the Princess [Catherine of Aragon] had brought the plate and jewels with her in . No, said Fox, softly, they had not known that.They had assumed that Catherine’s personal ornaments . . . would belong to her husband. (Mattingly : –) The attribution of knowledge to the quotee (‘‘the boy,’’ ‘‘the council’’) and in () also the proposition declared known are false: even immediately if ‘‘softly’’ falsified (‘‘No . . . they had not known that.They had assumed’’ the contrary, instead. For good measure, these factive and nonfactive denials keep up the same pattern of free indirect speech-report as the denied factive). Yet, factivity notwithstanding, who will in context mistake the falsity for truth?Whowill saddle the immediate subject (the alleged knower) or the highest-level quoter (novelist, historian) with either untruth, inconsistency, breach of pledge (never mind reason and resolvability)? Hardly anybody, I trust, beyond the presupposition analysts who maintain apropos ()–() that a that-clause, once factivized, keeps (‘‘projects’’) its factuality intact (‘‘uncanceled’’ on their view, unshifted on mine) regardless of any higher nonfactive embedding. The disproof of such immunity now extends to our vocal reframing. The context here makes a wider speech-frame explicit, via an added, higher, dialogic transformer, preposed or interposed within the knowledge report itself. The factivized thought was beyond doubt uttered between parties to a dialogue, which enacts the thought-quoting factivizer as addressor and the quoted factive subject as addressee.With ‘‘Mary . . . said to the boy that he knew his father . . .’’ (in []), the indirect thought-quotation comes under a superordinate, higherlevel but nonfactive indirect speech-quotation, and with ‘‘he shouted’’ (in []), under a free indirect one. Given those nonfactive higher-level transformers, both the utterance and the origin, hence the onus, of the knowledge predication cum statement trace back to where they properly belong: to the sayer/shouter as first-level unvalidated quotee. It is those speaking quoteeswho in turn (mis)quote, and (mis)factivize via their own lower-order transformer ‘‘knew,’’ what their conversational partners allegedly deem a 224 Poetics Today 22:1 fact in their state of epistemic certitude. Wherever the discourse spells out thoughtwithin speech-report, factive within nonfactive mediacy, indirectness within plain or free indirectness, all at once, the responsibility for the truth of the entire lower-level inset-within-inset (‘‘knew that . . .’’) is univocal: it rests neither with the highest frame nor with the lowest, thinking subject but with the intermediate speaking quotee (Mary, Fuensalida), who doubles as mind-reading quoter. On her or him alone falls the commitment to both the subjective (mis)predication and the objective (mis)validation of knowledge, to both the other’s state and the joint statement. Factively noncommittal, again, the re-quoter may assent or, as here, object de facto, contingently alone, and the ‘‘real’’ truth may itself likewise shuttle, depending on the overall setup, his own authority included. That the predicated state and/or the validated statement in () or () turn out false has to do not with any immutable law of discourse form but with the contingencies of fact in and information about the discourse world. () Aunt Sybil . . . wrote poetry. She was poetically superstitious. She said she knew she would die soon after my sixteenth birthday, and did. (Nabokov  []: ) () I . . . had to endure for several minutes the inquisitive commiseration of the campmistress, a sluttish worn out female with rusty hair. Dolly she said was all packed and ready to go. She knew her mother was sick but not ‘‘critically.’’ (ibid.: ) In the first example, the speech-reporter distances himself more perceptibly from the reported knowledge state(ment): the noncommittal ‘‘said’’ both immediately precedes and grammatically controls the reportee’s double epistemic commitment (subjective and assertoric, then objective and presuppositional). In fact, both women died, Aunt Sybil exactly when ‘‘she said she knew she would,’’ Dolly’s mother before the campmistress ‘‘said’’ Lolita ‘‘knew her’’ to be ‘‘sick.’’ Of the two factivized quotations-within-quotation, then, the one more distanced from the higher quoter proves true to the frame and the facts alike, the one less distanced untrue.Yet either outcome just happens to be the case, since the narrator happens both to know and to foretell either woman’s end on reasonable authority.Were his narration otherwise—uninformed, uncommunicative, unreliable—we would have to feed his self-distancing into our own reanalysis of the perspectival montage, possibly with a different or suspended or wrong judgment on the factive’s truth value as a result. By now, however, all this should be expected frommy earlier systematizing of nondirect thought-report in factivity.Themaster rationale, complete with its bases and branchings of inference, persists across reported objects. Sternberg • Factives and Perspectives 225 In the resolution of the loweras of the higher-order ambiguities of factivity, the aids to inference also cut across quotational matter andmanner. Let me just quickly exemplify how one key variable recurs, namely, the co-presence or absence or definiteness of a nonfactive extra transformer: compare ()– (), ()–() with () Henry, his arm thrown securely around the fretful [Spanish] ambassador’s neck, dragged Chapuys through a long political discussion. This business of Milan would surely mean war. Chapuys did not think so. Could he have permission to go to [Catherine’s house at] Kimbolton? Oh yes, it would! And the French would give anything in the world to secure his alliance. Francis knew he could not win without England’s help. (Mattingly  []: ) To bring home the point, this example complicates matters not by omitting all signals of modulation from the frame to a speech inset, but by weakening them relative to the norm above. All other things being equal, this is enough to make a difference in identifying the object and what it involves. ‘‘Discussion,’’ as well as the ensuing expressive markers, alert us to the free indirectly reported exchange between the eager Henry and the impatient ambassador. This dialogic understanding organizes everything— assertions, exclamations, questions, answers—until we come to the factive presupposition. Does ‘‘Francis knew . . .’’ fall under the same free indirect vocal pattern, with the difference that the speech indirectly quotes Francis’s thought in turn? If so, as factivizing speaker cum insideviewer on his own, Henry possibly misreads Francis’s mind, whether wishfully or with the intent of misleading the Spanish addressee to his own advantage. However, the same insideview alsomakes sense as a break in the free indirect dialogue: an aside, addressed by the history teller to ourselves, whereby he indirectly re-presents and validates from within Francis’s mind the claim just voiced by Henry to his own auditor. At this juncture, a modest extra aid to reading the knowledge statement, like the parenthetical ‘‘he shouted’’ in ()’s discourse or ‘‘as Henry affirmed,’’ would spare us all the trouble. Instead, even the expressivemarkers now disappear. Only when we learn for certain that Francis never reckoned on Henry—ergo, the entire statement is false, public, free indirect—does the fork to have been taken establish itself. Compared with the plain indirect reading, this yields not only a longer, tenser, richer (hi)story of event and discourse event but also one more exterior, interactional, because it assigns the ‘‘knowledge’’ to a vocal as well as subjective mediating agent. First alternative (hi)stories, then the victory of the tangled contender. The 226 Poetics Today 22:1 historian disdains the line of least effort preached by Grice’s ‘‘Cooperative Principle,’’ Sperber and Wilson’s Relevance theory, or Lessing’s aesthetics of transparence.He rather opts to the last possiblemoment for the pleasures and profits of compound factive ambiguity. 6. Presupposition Retested and Redefined Before drawing together the threads to redefine presupposition as both a type and an exemplar of inference, I must keep two promises of extending the analysis to key variables. One concerns negation, the other nonfactive presuppositionality. Fortunately, such is their alignment with the regulating dynamics that an outline will suffice to exhibit and motivate it. Within our new paradigm, then, how does this inference type fare under negation? Introduced at the outset as the originary but moot definitional feature of presupposition, and occasionally glanced at since, the ‘‘negation test’’ now falls into pattern with the corresponding affirmative statement. Inversely, that unity in variety completes the break with earlier approaches to the topic, whether uniform (that of Strawson or the logical semanticists) or divided (as usual among pragmaticists). To cut short a long analysis and a longer story of dispute, correlating two facts about negation so dynamizes its workings as to resolve all traditional issues of inference (non)survival by appeal to the same principle that has brought us thus far: constancy within well-defined perspectival limits. One fact has to do with negation’s built-in denial of some representation coming from a viewpoint other than the denier’s, or here, the denying quoter’s; the second fact lies in negation’s scope-ambiguity, the multiplicity of propositional as well as perspectival targets that it can operate on within any given stretch of text and thatwe inferrers need to disambiguate in context.50Either factor accordingly joins an invariant (other-mindedness, plurality) with a variant (some mind, some target): this redoubles the negation’s indeterminacy (denial of whose what?), yet by the same token opens up the range of candidates to enable its uncanceled integrationwith someone’s disapproved outlook on something. By itself, and conceptualization apart, neither of these factors is quite new, not even regarding this inference type. In a way, one was already sensed by Strawson himself, though he fails to correlate it with presupposition, so that its mate also disappears from view there. . I advisedly call it a ‘‘fact’’ because the persistent claim for univocality (surveyed inBurtonRoberts : ff.) lacks any empirical reality: confronted with a negative operator, we cannot possibly foreknow, and so must always figure out, which of the relevant affirmations it is likeliest to negate. Sternberg • Factives and Perspectives 227 Generally, Strawson treats ‘‘not’’ as language’s prime operator or ‘‘device’’ for the ‘‘explicit exclusion of a predicate, or the explicit rejection of a statement,’’ whether ‘‘one’s own or another’s’’: There are many very different kinds of occasion on which our primary concern is with the explicit exclusion of a predicate; e.g., when we wish to contradict a previous assertion; or to correct a possible false impression; or to express the contrast between what had been expected, feared, suggested, or hoped, and the reality; sometimes, when we are answering a direct question; sometimes, when we grope towards the right description by eliminating the wrong ones.What is common to such cases is that they create a need or a motive for emphasizing a difference rather than a resemblance. (: , , ) Latent in the concern with ‘‘emphasizing a difference’’ is a sense of negation’s tense bi-perspectivity, to which I will return. Meanwhile, observe that exactly when it comes to presupposition—‘‘The King of France is not bald’’—Strawson forgets what he generalizes here about this ‘‘device.’’ To use his own terms, if ‘‘not’’ serves for the ‘‘exclusion of a predicate’’ or the ‘‘rejection of a statement,’’ then the above exemplar would either ‘‘exclude’’ the predicate ‘‘bald’’ (internal or narrow-scope negation) or ‘‘reject’’ the entire utterance (external or wide-scope negation). But he now ignores this scope-ambiguity (along with its further differential, in effect perspectival, correlates regarding the negator vis-à-vis the implied affirmer). Indeed, he must ignore it to sustain his very definition of this inference type as one that remains true whether the utterance bearing it is true or false—hence as constant under negation. According to his lights, if the utterance of both ‘‘The King of France is bald’’ and ‘‘The King of France is not bald’’ presupposes the referent’s existence, then the negation must be internal or narrow-scope (governing ‘‘bald’’ alone, i.e., its predication of the King) rather than external or wide-scope (nullifying the whole proposition, the royal existent included). The one-scope-one-reading exigency follows from the approach, which rules out all presupposition denial (and ‘‘cancelability’’ at large). Therefore, as many have objected, Strawson cannot handle perfectly acceptable instances under obvious (or so they allege) wide-scope negation: like ‘‘The King of France is not bald, because there is no such person’’ or ‘‘John doesn’t regret that he failed, because, in fact, he passed.’’ Nor can the neo-Strawsonians.Within their logico-semantic approach, the opening parts of these ‘‘sentences’’ entail theKing’s existence or John’s failure, just as would the corresponding affirmatives. And given that entailments are uncancelable without anomaly, either sentence must become contradictory— represent an impossible world—once the negation assumes a wide-scope 228 Poetics Today 22:1 meaning under the pressure of a follow-up that denies existence or asserts success. A false prediction, an explanatory impasse.51 Inversely with the ‘‘pragmatic’’ alternative. Its exponents generally assimilate presupposition to entailment, too, but draw the line at the negative sentence and so may freely ‘‘cancel’’ its inferences in face of contextual discord. Nowhere more so indeed than regarding putative wide-scope negation of the kind just exemplified, whose sequel (‘‘there is no such person’’ or ‘‘he passed’’) overtly denies the presupposition (of the referent’s existence or failure). We have already witnessed the oddity of now reductively semanticizing, now distinctively pragmaticizing this inference type. Nor need we go into the variants of such change of the type’s identity and the approach’s ground. Judging by results—a fortiori with the results themselves judged by formalist ideal—the verdict is inescapable. As the pragmaticist turn toward entailment for presuppositional survival breaks down, so does the jettisoning of entailment for cancelability at large and particularly under negation. Here the approach must acknowledge the frequent survival of inferences regardless but cannot explain and predict (‘‘project’’) the jumps between survival and alleged cancelation (not even on the readings known to it: aliveness to, say, free indirect discourse would aggravate matters again). Just compare example ()’s ‘‘Freddie knew no such thing’’ with ‘‘Gloag knew and Mrs. Henn-Christie didn’t know that the new by-pass had already been approved’’ (Gilbert : ): why should one factive putatively lift, and the other keep, the epistemic bond under a single operator? And how to foretell from the analogous-looking inputs, algorithmically, which will do what? Negation tests this mixed approach in turn, as well as the inference pattern itself, and again shows it wanting even by its own standards. Hence, as even a recent formalist survey concludes, ‘‘the failure of any pragmatic theory to account in anything like a satisfactory way for the projection and discourse properties of presuppositions of negated carrier sentences’’ (Seuren : ). No aprioristic rule has emerged after fifty years: another predictable collapse of the whole idea of formally predicting discourse behavior. For the key to an integrated treatment of negation both within and along with presupposition, we need only extend the foregoing argument. Under negation, as elsewhere, given the built-in (e.g., factive) trigger, the . For criticisms along such lines, seeWilson : – passim; Gazdar : ff., –; Van der Sandt : –, –; Grice : ff. Burton-Roberts () sharply criticizes the critics in turn. Yet his own amended semantic theory only relegates the awkward negations to pragmatics: on top of methodological hybridity, this incurs other substantive (cancelationist, projectional) difficulties of the kindwe discuss next.The usual fate ofmixtures from either side. Sternberg • Factives and Perspectives 229 inference-cum-bond remains uncancelable. And it remains so, moreover, due to its linguistic encoding, rather than to any logical entailment, which never comes into it. (Hence one is spared both the semanticist’s hopeless clinging to and the pragmaticist’s hasty retreat from deductive logic in negative forms.) As always, again, what pragmaticists intuit and misconceptualize here as cancelation (all the way to overt denial) is in reality nothing but a reshuffle of commitment among the parties involved: withdrawn (‘‘canceled,’’ if you will) from some viewpoint, it yet adheres and/or migrates to another. In brief, ‘‘constancy under negation’’ again means not constancy of perspective and truth bond (‘‘the speaker’s,’’ as if the only role involved) but quite the opposite: its adjustability via inference to someone else in play, for overall consistency. As always, finally, this new extension of the principle—not least the what’s-whose inference toward inference— has its own specifics and variants. Actually, these partly lurk already in the suggestive excerpt from Strawson cited above.52 Recast into my own terms, Strawson’s ‘‘we’’ negating/excluding/rejecting speakers will become quoters and his ‘‘not’’-device for negation/ exclusion/rejection will encode a marked, though flexible, perspectival opposition to some quotee. The element negated (excluded, rejected, contradicted, corrected, unrealized, eliminated) then gets distanced from the negator toward a viewpoint that does or did (if actual) or would (if imagined, ‘‘possible’’) affirm it. In short, we may generalize, negating is counterspeaking. A binary yes/no, pro/contra structure of quotation ensues, with the encoded negative operator doing duty for quoting verb or inverted commas as well. The options for negative transformer accordingly range from the implicitness of simple ‘‘not’’ through the manifestness of ‘‘not, as X says/thinks/knows, . . .’’ (or ‘‘not, pace or contra X, . . .’’) to the juncture ‘‘X does not say/think/know that . . . ,’’ our present chief concern. In ascending order, these options bring home why such reconceptualizingmatters, here above all: presupposition thereby reveals itself as themeeting ground of at least two independent language-specific quotational structures, factivity and negation. The two indeed converge on the last of the above transformers; but even when they don’t, their re-presentational kinship suggests a uniform inferential rationale.The basic family likeness even sharpens in the parallel between such apparently unrelated phenomena . His insight there has been variously echoed since by presuppositionalists as well as negation theorists: Fillmore (: –) and others even call some of the denials at issue ‘‘semiquotations.’’ For extensive references and an attempt at synthesis, see Horn . Across all differences, terminological or substantive, however, negation hardly lends itself to these piecemeal cancelationist accounts, any more than to their Strawsonian and neo-Strawsonian opposites. 230 Poetics Today 22:1 as negation and conflictual (e.g., ironic) free indirect discourse: both turn out to involve a thorough reshuffle of viewpoints (attitudes, commitments) under threat of inconsistency. So any threatened clash between propositions translates, and resolves itself, into one between perspectives on them. On this inclusive quotational ground, moreover, negation’s scopeambiguities in turn find their rationale. On the face of it sui generis, they essentially specify or thicken the ambiguity attaching to the negational as to any other discoursive montage, and accordingly channel or enrich in their own peculiar (‘‘counterspeaking’’) way the options for epistemic interplay, unpacking, sense-making already familiar to us in principle. Thus, a cursory glance at ‘‘The King of France is not bald’’ will show how the latent perspectival tensions branch out on the two main readings. The narrowscope negator occupies an epistemic and commissive position that differs equally from the echoed-and-opposed (‘‘counterspoken’’) affirmer and from the wide-scope fellow negator vis-à-vis his opposite number. A remarkable measure of this twofold, equivocal bi-perspectivity is that, of the four, all but the third share at least the presupposition of the King’s existence. Such tense alternative (dis)continuities in viewpoint compound where negation meets factivity. Interwoven, their respective second-level representations of discourse produce a third-level, re-re-presentational discourse against discourse: counterspeaking thickens into counterquoting. And the ambiguity of the product—the montage—redoubles accordingly. Among these compounds, in turn, the most recalcitrant-looking is the one where the factivizer apparently denies what he presupposes (analogous to adding ‘‘but there is no such person’’ to the King-of-France shibboleth). Indeed, it recurs in the field as the ultimate argument for cancellation. Thus, ‘‘I don’t know that Bill came’’ would typify inference failure: ‘‘[T]he presupposition that the speaker knows [that Bill came] is precisely what the sentence denies, and such denials override contradictory presuppositions’’ (Levinson : ; see also Kiparsky and Kiparsky : ; Wilson : –; Gazdar : ; Burton-Roberts : , ff.; Beaver : –; contrast Lyons : , ). But what the example proves, and the only typical thing about it, is that its contextlessness rather opens it to multiple sense-makings, none of them presupposition-‘‘overriding,’’ not even amidst ‘‘the speaker’s’’ denial. First check this argument against readings of the factive self-predication that are usually overlooked, beginning with the idiomatic one: () ‘‘I would have won this seat; and you, with your brand of politics, haven’t a hope in hell.’’ ‘‘I don’t know that politics is about winning always,’’ said Jerry, in his ‘explaining things to infants’ voice. (Barnard :) Sternberg • Factives and Perspectives 231 In context, ‘‘don’t know’’ expresses none of its possible literal meanings (hole in knowledge or outright denial) but Jerry’s doubt about the truth of the ensuing proposition. But nor does he as negator cancel this questionable presupposition ‘‘that politics is aboutwinning always.’’ Rather, he imputes it all away—to the adverse interlocutor, whose talk about winning and losing has provoked the echoing idiomatic negation. So the respondent’s counterquote is akin in force to our familiar pattern of indirect within nonfactive indirect quoting, ‘‘You think you know that politics is about winning always,’’ with the difference that here the affirmative quoter’s noncommitment verbally modulates toward invalidation—‘‘and I doubt it.’’ 53 The same ‘‘I don’t know that . . .’’ form of words, though, also lends itself to alternative, nonidiomatic readings, shared with factivity in general. These normalize and multiply resolutions to establish the flexible logic of perspective that they all share with the idiom—among other negatives and other type-specific triggerings—as well as to disprove the cancelationist approach via its own paradigm case. On an entirely neglected literal understanding, for example, the tense may even read as the historic present: () ‘‘I got this box and it’s a pretty big wooden case and it’s for the rabbi. I don’t know that it’s books at first.’’ (Kemelman : ) In the light of the initial ‘‘got,’’ the ensuing present-tense verbs mark a bid for vividness within an ongoing past reference. Once we translate (‘‘backshift’’) the rest of the given utterance into the signified past, there is nothing self-contradictory about the utterer’s presupposing the knowledge that he declares in the predicate not to have: the have and the have-not involve different selves, or different phases of one self in epistemic development.The presupposition straightforwardly holds to contrast two I’s—the experiencing nonknower (‘‘at first’’) and the narrating knower in retrospect—just as it would if the indirect self-report began with ‘‘I didn’t know that. . . .’’ In contrasting the two selves, either negative indirectness (‘‘don’t/didn’t’’) also counterquotes what someone else (e.g., the addressee) might affirm or believe about the earlier self ’s knowledge.Where the corresponding affirmative (‘‘I know/knew that . . .’’ or the addressee’s possible ‘‘You know/knew that . . .’’) would join or equi-focus the indirect quoter and quotee, validating presupposer and validated thinking subject, the shift in viewpoint attached to negation again turns likeinto other-mindedness.Within the rule of negation, however, here the turn polarizes the knowing self ’s commitment with one other’s (the one-time I’s) unfettered ignorance and a second, . Hence the difference even from () or (), which leaves the pressure for invalidation contingent, because context-dependent rather than verbalized in the opening transformer. 232 Poetics Today 22:1 co-fettered other’s (the you’s) untimely predication of knowledge—rather than, as in (), juxtaposing the other’s commitment by the self with the self ’s own questioning attitude toward the presupposition. Along with additional, commoner nonidiomatic readings, this brings us back to the twinned general issue of negation’s scope-ambiguities, resolvable in turn into further commissive and perspectival inequalities. Let me quickly outline them in descending order of width. () ‘‘What size do you want to be?’’ it [the Caterpillar] asked. ‘‘Oh, I’m not particular as to size,’’ Alice hastily replied; ‘‘only one doesn’t like changing so often, you know.’’ ‘‘I don’t know,’’ said the Caterpillar. Alice said nothing: she had never been so much contradicted in all her life before. (Carroll : ) Here the scope of negation is external and the widest possible, approximating that of logic’s negative operator 4 (4p) in the propositional calculus: the curl symbol translates as ‘‘It’s not the case that . . .’’ and falsifies everything that goes after. The Caterpillar therefore never contradicts itself in ‘‘I don’t know,’’ either. For it understandably contradicts instead the entire epistemic state(ment) to which it harks back: both the knowledge predication of itself (‘‘you know’’) and the presupposed knowledge (about ‘‘one’s’’ dislike ‘‘for changing so often’’). Neither is the case in its view as echoquoter. Which also means that the Caterpillar throws the entire responsibility for both falsities, subjective and objective, upon the predicating cum presupposing Alice (and by larger implication, our homocentrism): upon the lower-level quoter who has already bound herself to their truth in venturing to co-involve its own multiform self. Driven into diametric polarity all along, the viewpoints become contradictory indeed, to the unloosing of the counter-viewer. So, in (), might Freddie respond and relate to his wife’s ‘‘You know you love it [pedicare] yourself.’’ Or, he even does respond, silently, if we take the omniscient narrator’s total denial on his behalf (‘‘Freddie knew no such thing’’) as an equivalent interior counterquote (possibly ‘‘I know no such thing,’’ mediated in free indirect style).The widest-scope reading cuts across surface grammatical person and reportive form. In turn, like ()– (), this negative pattern accordingly never warrants what cancelationists would conclude from it: that ‘‘the possibility of denying one’s own presuppositions is a fundamentally important property’’ of the inference type (Levinson : ).The Caterpillar or the narrator/Freddie montage are not denying their ‘‘own’’ (factivized) presupposition along with (factive) assertion but Alice’s or thewife’s; and therefore, again, it is not (propositional) Sternberg • Factives and Perspectives 233 cancelability but (perspectival) shiftability under negation, as elsewhere, that makes ‘‘a fundamentally important’’ hallmark. Relative to either example, the scopemeaningfully narrows in () above, yet only to play still another variation on the thematic rationale. Likewise— with a change of person, as well as tense, and with an extra counterquoted viewpoint—in () ‘‘Contrary to gossip, he did not know that I was pregnant then. I never told him.’’ (Vidal : ) The scope of negation narrows here from the whole complex sentence to the transformer, leaving the inset that-clause duly endorsed. As the followup ‘‘I never told him’’ establishes, the quoting woman was indeed pregnant at the time, only the quoted ‘‘he’’ did not know it.The epistemic self-division on the time axis in ()—knowing I/now versus nonknowing I/then—finds an interpersonal (‘‘I’’ versus ‘‘he’’) parallel across time. On the other hand, rather than operating to negate the entire sequel, as in (), the stressed ‘‘not’’ looks back to the initial ‘‘Contrary to gossip,’’ so as to emphasize the truth of ‘‘his’’ nonknowledge against opposition from still another viewpoint. (Again, the latter target compares with the possibly antagonistic addressee in [], only here it is vocal and named.) For the gossip-mongers would affirm the predication that she denies in response, as well as coendorse her presupposition. There arises a threefold point of view, where each reflector contrasts differently with the other two in an overall epistemic hierarchy. She enjoys full awareness, objective and subjective, ‘‘he’’ has none, ‘‘gossip’’ mixes a true (endorsed, objective, presuppositional) with a false (negated, subjective, predicational) half—always on her view and authority, down to the commissive implications. Further, to pinpoint and classify and explain these scope-ambiguities, their divergent resolutions included, we may appeal anew to the factive verb’s componential analysis. Though the very same negative operator recurs throughout our examples, along with contrastive stress inmost of them to boot, it does not operate on the same built-in component. Or from the inferrer’s side, the tripartite semantic composition of ‘‘know’’ enables us to train the ‘‘not’’ on the component that best fits each discourse. Which here also means negating the sense-component (and the associated proposition) that least fits the discourser—that makes the best candidate for rejection on his or her part amid relegation to another, for counter(re)quoting, in short. Scope, negative operator, presuppositionality, factivity’s sensecomposition: all the relevant coordinates dynamically interact and integrate on this account of the system, and on it alone. In () and (), what is thereby negated (whether doubted or, normally, 234 Poetics Today 22:1 denied outright) is the validator. And if we leave aside the twist introduced by the idiom, the operative rule on this reading even gains in sharpness. Throughout factivity, wide-scope means validator negation, falsifying/ counterquoting everything in the discourse, so that the falsifier/counterquoter shakes off or reverses all the truth bonds involved in it but one: to the association of the negated whole with another mind, a counterquotee who does or would intensely affirm it.The burden of commitment to the rest accordingly falls on some lower, (re)quoted, distanced perspective—in (), the addressee’s—just as it does in the shift to indirect within free indirect quoting, except that the nonvalidation signaled there moves now toward invalidation. Distancing widens into binary (pro/con, yes/no) distancing. With this difference vis-à-vis free indirectness, born of the very meaning of negation, the similarity across the affirmative/negative line is remarkable and systemic. In each case, we then infer the quoter’s self-distancing amid other-binding, and at times under an equally open threat of inconsistency: between what the quoter appears to factivize as positive ‘‘knowledge’’ and then contradicts in ()–(), for example, or within the quoter’s very negated factivizing, by which he disclaims ‘‘knowledge’’ of what has just been and/or will next be expressed in (). At such an extreme, either factive gets threatened not with implicit, contextual dissonance (as in []– []) but with outright (co)textual absurdity. And to avert the threat, we read there a free indirect mediator who affirms in a self-styled knower’s name a complex (half-predicated, half-presupposed) proposition that the mediator himself judges untrue, and we read the Caterpillar here as indignantly rejecting a state(ment) that Alice would co-attribute to it among other (human) knowers. But the similarity persists in more common free indirectness, where the validator gets distanced away, short of overt invalidation.54 . The parallels also meet in free indirect negation to generate a fourth-level, re-re-rerepresentational discourse and a variety of compound distancings. Recall the occurrence of that style in ()’s ‘‘No, said Fox, softly, they had not known that.’’ Or, with a change of rere-presented object from public to inner discourse: () ‘‘I think one of them works in a medicenter.’’ ‘‘Mm, that would explain it,’’ he said. She didn’t know [that one of them worked there]. Or she knew but didn’t know that he knew.’’ (Levin : ) Having gathered that she did know, we read the contradiction as a free indirect insideview of his uninformed and undecided mind: ‘‘She didn’t know, he thought,’’ and so forth. I will leave to you the resultant perspectival and commissive unpacking, except for two geneal remarks. First, themontage is here even thicker than in ordinary negative indirectness. Second, like every corresponding affirmation, every factive negation hovers in principle between the indirect and the free indirect readings, on top of all other ambiguities: recheck even the preSternberg • Factives and Perspectives 235 In (), on the other hand, the quoter rather negates the mentalizer, denying that the quotee had the knowledge presupposed, that ‘‘he’’ shared with her the uncommon epistemic state (hence proposition, commitment, vantage point) expressed by the inset. Having penetrated hismind, she declares it empty of the given re-presentation (concerning her pregnancy): a ‘‘minus insideview,’’ I call it (:  ff.) The negative operator accordingly focuses on the transforming clause, which, ‘‘contrary to gossip,’’ introduces ‘‘him’’ as a nonknower. The paradigmchange fromcancelation/preservation to reshuffleof bond and inference aside, however, does this rationale amount, even lend color, to the traditional formula that ‘‘the speaker’s’’ presupposition ‘‘evaporates’’ under wide-scope (here, validator) negation and ‘‘survives’’ under narrowscope (here, mentalizer) negation? No, not even on the ordinary indirectdiscourse reading of factives, if only because elsewhere it is rather the intensifier that gets ‘‘narrowly’’ negated, to a different, third effect: () He said: ‘‘You don’t know that a woman shot him. I mean you’re not sure, are you?’’ ‘‘No,’’ I said, ‘‘That’s true.’’ (Chandler : ) () ‘‘He’s fixed it. I tell you he’s fixed it.’’ ‘‘Harry, you don’t know that.’’ ‘‘Like hell, I don’t.’’ (Watson : ) From the first to the second negation in either dialogue, the examples are nicely complementary. In (), the opening negator himself proceeds to explain that, by ‘‘You don’t know,’’ he means ‘‘you’re not sure.’’ In (), we have to supply that gloss ourselves, if only by elimination.What is denied in response—so our inference goes—can’t be know’s mentalizer, given that the affirmer has just voiced, twice, a representation (‘‘he’s fixed it’’) of the world as he sees it; nor can it be the affirmer’s commitment to that iterated representation. So, we conclude, it must again be the intensifier, freely challengeable by the hearer. In de-intensifying the predicate, with or without expressive stress, however, the negators do not yet invalidate the knowledge statement at issue.They would just lower its epistemic and commissive status below factivity: ‘‘Whoever knows that—and the presupposition endures relative to the possible knower, boasting the appropriate empirical warrant—you don’t, because you’re only guessing.’’ And the challenged party agrees and disagrees, respectively: his ‘‘No’’ concedes the devious examples, more formally univocal than (), and you will see that our analysis of them as indirect is nevertheless a matter of (high) probability. 236 Poetics Today 22:1 intensifying ‘‘You don’t know’’ in (), while counternegating and so reintensifying it into a virtual ‘‘I do know’’ in (). Without having exhausted the variants, we may safely generalize the principle that runs through them all.55 It even readily extends to and beyond presuppositional negation at large, where matters are usually simpler than in the cross with factivity. Taken as ‘‘sentences,’’ out of context, ()–() would remain multiply ambiguous: nothing then licenses us to infer from the negation the truth, untruth, nontruth of the presupposed complement or any element in and around it. As it is, with the sentences appropriately discoursed, we may reason out the most probable of the negation’s latent equivocal readings. Such disambiguations vary in intricacy, certitude, allotment of postures and responsibilities—according to how the basics interact and best integrate among themselves, as well as with each specific context. That ‘‘how’’ varies in turn with the sheer number of factors to be coordinated into a reasonable network of disharmony. Among the basics, themost universal are the negative operator vis-à-vis its scope.Where presupposition enters, things get involved, and factivity, with its built-in semantic compositeness and alternative nondirect yet always multivocal schemas, drives the tangle to the limit. For coherence, however, we must throughout infer from the given negation some variant of counterspeaking—or in factivity, the higher level of discourse against discourse, counterquoting—and shuffle, adjust, order the unharmonious epistemic viewpoints in play to suit. Discourse against discourse being still discourse about discourse, negation turns out intimately related to factivity, as varieties of quotation with a special bearing on presuppositionality. (The more so, we will find, because the latter elsewhere crosses with quoting other than either negative or factive.) And the unexpected family resemblance even tightens when the commonality rises to high-level unharmonious discourse, such as just exemplified by other-minded free indirectness. Small wonder that the parallels, looser or tighter, freely intersect to extra ambiguating, polyphonic effect, all in the family, so to speak—as is the very extension of the analysis from affirmative to negative. But this also means a sea change in the very status of negation: from the . Among them, conversely, the variable readings traced suffice to expose and explain the mistakes in standard analyses of factive and otherwise presuppositional negation, as well as the helplessness of the projection rule sought by the analysts, whether along ‘‘semantic,’’ ‘‘pragmatic,’’ or mixed lines.With special regard to factivity, contrast, for example, Kiparsky and Kiparsky : , ; Kartunnen : –; Wilson :– passim, –, – ; Gazdar : –, –, –, –, –, –, –, ; Levinson : –, , –, –, , ; Van der Sandt : –, , –, ; Grice : –. Sternberg • Factives and Perspectives 237 distinctive trait or test of presupposition, à la Strawson et al., to another reflex, however significant, of its typal distinctive logic. If entailment and implicature (as well as assertive commitments) vanish under this test, it is for the same reason that they have no equivalent to factivity’s inferencekeeping through perspectival adjustment: because they lack encoded triggers and so vanish under quotation, in short. Either way, the negation test itself grows intelligible at last, once referred to the supertest and its deeper ground. So negation falls into pattern, with a double-edged result. Its importance, nevermind uniqueness among features of presupposition diminishes; while its applicability becomes absolute, and accordingly its contribution to establishing the uniqueness of presupposition among inference types. Two faces, these, of the same thing: the constancy test’s own continuity with the type’s overall rationale and inbuilt resources of advancing from trigger to best fit. As presupposition is an exemplar of inference, so is factivity of presupposition, and its lessons carry over to the rest even more directly. Other triggers present far simpler cases than factivity’s mandatory double (language/ viewpoint) duality: few run to a complex sentence, none to an inbuilt biperspectivity. If anything, factivity can grow even more intricate and ambiguous, whether along its own distinctive lines (notably the lines of ascent from indirect report to indirectness within free indirectness and/or from affirmative report to negative ‘‘counterreport’’) or by absorbing fellow triggers, or both. The latter route to extra presuppositional thickness, we have seen, most often goes through definite reference—the only typemarker comparable to the factive in importance and notoriety, as well as being the staple of representational discourse at large. But all other relative simplexes are equally open to absorption, with various compound effects (or, inversely, joint resistance to ‘‘projection’’ algorithms). Those simpler triggers may indeed absorb each other (e.g., a temporal clause with ‘‘the King of France’’ for a subject) and also gather extrapresuppositional complicating variables like modality or fictionality as well as negation. All other things being equal, however, no such juncture will rival its factive counterpart in inferential thickness: if only because the latter’s two-clause given can always read in principle as a three-clause quoting within quoting, schema within schema, montage upon montage, thought with or without utterance. Nor will any such rival juncture compoundunder negation twoperspectivalmontages. In simplest, independent occurrence, therefore, the rules and lessons of factivity easily apply across 238 Poetics Today 22:1 the board. We need just abstract or extrapolate the type-specific from the factive’s richer trigger-specific differentials: leave aside for generality its irreducible verbal and quotational uniqueness.The shared anchorage in the language system makes the bonds/inferences carried by all presupposition triggers likewise undefeasible, as the common enclosure in discourse makes them all perspectivized and at need reperspectivizable—especially away from the official discourser toward some lower, quoted voice. Here are three assorted independent equivalents—definite noun phrase, temporal clause, nonrestrictive relative clause—under mortal threat of inconsistency: () The dead man beside him raised himself on an elbow. ‘‘He’s not actually drinking the stuff, is he?’’ asked the dead man. (Price : ) () Even after I was happily hanged and Wopsle had closed the book, Pumblechook sat staring at me, and shaking his head, and saying ‘‘Take warning, boy, take warning!’’ (Dickens n.d. []: ) () ‘‘Look, look, Septimus!’’ she cried. For Dr. Holmes had told her to make her husband (who had nothing whatever seriously the matter with him but was a little out of sorts) take an interest in things outside himself. (Woolf  []: ) The alternative to a bizzare, impossible world is again figuring out a hidden lower perspective that will normalize it. The ‘‘dead man’’ plays the role in a military exercise, thus counting as one from a special ontic viewpoint that the narrator realizes (‘‘objectifies’’) for shock effect. Pip ‘‘was hanged,’’ ‘‘happily’’ at that, during a crime story read at him byWopsle and in the eyes of the ‘‘staring’’ Pumblechook, who delusively identifies himwith the murderer. And Septimus, about to kill himself, had nothing whatever the matter with him on the diagnosis of the egregious Dr. Holmes, now echoing in the mind of the wishful wife. Throughout, no dead end—literally—but a presupposition in virtual inverted commas: quoted by the author at a subjective remove, ludic or ironic, distanced by the sense-maker to avert absurdity, optionally framed by both and so beyond the reach of ‘‘projective’’ formula. All simple indeed by now. (Apart from the unending involution of definite reference, contrast, for example, Levinson : , ,  and Van der Sandt : – on the temporal clause.) We can accordingly redefine presupposition among inference types. While entailment is uncancelable and the mixed bag of ‘‘implicature’’ all cancelable, presupposition is uncancelable but also uniquely shiftable, so that it alone enjoys the highest and positive survival value of dynamic adjustability to context or along with context (in my sense). It can always Sternberg • Factives and Perspectives 239 transfer to a viewpoint other than the nominal speaker’s. As we have found, shiftability enables its operative survival evenwhere entailment, though uncanceled, would in effect get suspended or ‘‘killed’’ by contradiction. From the beginning to the end of the inferential dynamics, the commissive force of presupposition never evaporates, though it may, in a sense must, change bearers on the way.The relay goes from the sentence element that triggers the presupposition into the discourse (and our mind) by virtue of its encoding as such within the language system; through the discourser (speaker or, as invariably in factivity, quoter) who formally emits and assumes it within the message; to whichever viewpoint involved, single or joint, that in context looks like the effective, best-fitting candidate for presupposing. In the actual encounter with the discourse, we to and fro among its three universal factors—language, world, and perspective—until they make a reasonable common household, epistemically ordered, inter alia. At the end of the road, therefore, the commitment always falls on someone; the questions always left for the inferrer to determine, along the lines suggested, is upon whom, in whose company, if in anyone’s, how certainly, via which mediacy, whether on social or secret record, and to precisely what extent and effect. Unless, of course, the textor sense-maker leaves, not to say lands the presupposition in trouble as presupposition: discordant and unevaporable, yet effectively unshifted, unassigned, unperspectivized out of the official discourser’s commitment, possibly for a good reason, however offensive to rationalists. But the Law of Reciprocity still holds. With this route come in effect to a dead end, on a certain patterning at least, we break out elsewhere.The way to (best, alternative, or just favored) integration then passes through a mechanism of an altogether different (genetic, ontic, functional, etc.) order and logic, working across, even against the logician’s premise of representational (epistemic, ‘‘propositional’’) consistency.