DANIEL A. KRASNER
SEMANTICS AND FICTION
0. OVERVIEW Philosophers of language have traditionally had difficulty accommodating fiction within their theories of language, and these difficulties have traditionally had consequences beyond the immediate problems. The facts that fictions seem to make true things that aren’t true, and seem to refer to things that don’t exist have posed problems for accounts of truth and existence, and related topics. Once we get past the naive claim that fictions are just packs of lies about things that just don’t exist, we find that philosophers have, for the most part presented us with a number of unappetizing choices. Some, e.g., Saul Kripke and Kendall Walton, have claimed that fictional discourse isn’t fully meaningful, or doesn’t have its normal meaning, and have attempted to dispense with the problems that way (Kripke 1973, Lecture 1; Walton 1990). Others, e.g., Terence Parsons, Alan McMichael, Ed Zalta, and Richard Routley, have tried to find ways in which predicates can be true of things that don’t exist (Parsons 1980; McMichael and Zalta 1980; Routley 1979), or, like Saul Kripke and Peter Van Inwagen, have otherwise sought to complicate their ontologies (Kripke 1973, Lecture 3; Van Inwagen 1977). All of these approaches tend to involve significant epicycles in otherwise fairly satisfactory philosophical theories, and tend to preclude the formation of unified accounts of language or being. I hope to show that an account of fictional discourse closer to that of nonfictional discourse than any of the above is possible, and even plausible, one which would make the semantics of fiction depend on what’s true in possible worlds similar to the actual one, just as the semantics of nonfictional discourse depends on the actual world itself. There are many similarities between fictional and nonfictional discourse that can best be explained that way. But first, a cautionary note. Gregory Currie draws a distinction between what he calls the fictive use of fictional names and the metafictive use of fictional names (Currie 1988). The former is a use of a fictional name within a fiction which contains the Erkenntnis 57: 259–275, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
260
DANIEL A. KRASNER
character it is a name for, and the latter is a use of the name in conversation about the fictional character outside of the fiction. I will expand this distinction into one between fictive discourse (discourse that goes to compose a fiction), and metafictive discourse (discourse about a fiction). I want to make clear, that, in this paper, I shall be analyzing metafictive discourse, and shall have little or nothing to say about fictive discourse. Now to the similarities I mentioned. Fiction can be mistaken for nonfiction, and vice-versa. When Thomas More’s Utopia came out several people mistook it for an actual traveler’s account. This was doubtless a mistake, but it doesn’t seem as serious as it would have to be if the semantics of fiction and nonfiction were very different. In that case it would be a sort of a category error, but it doesn’t seem that grossly wrong. Much fiction writing seems to obey constraints that also govern nonfiction writing. In each case, the author goes to lengths to make his writings seem plausible. Fiction and nonfiction authors alike often strive for an air of realism, and try to see that their works fit with known facts. And considerations of emotional tone apply in each case; one doesn’t include vulgar jokes in a description of a funeral, whether it’s real or fictional. All this makes sense if the author is talking about similar things, but is difficult to explain otherwise. Finally, the way we evaluate fictions is likely to be like the way we evaluate nonfiction. We criticize each for portraying impossible events, or events impossible given plausible constraints. We assume a background of known fact to each. And emotional reactions to fictions are likely to be similar to, albeit less intense than, reactions to like nonfictions. All this suggests a similar semantics, involving possible worlds similar to the actual one. I submit that the proposed semantics for fiction that would best explain these similarities, by making fiction about possible worlds similar to the actual, is a counterfactual semantics, and I intend to discuss such in this paper. I shall begin, in Section 1, by giving David Lewis’ versions of counterfactual semantics, the only well-worked out versions extant, and explain the advantages of them (Lewis 1983). I shall then point out the problems they have in some detail, in Section 2. Then I shall give my own version, and explain how it would work and how it can overcome the objections to Lewis’ version, in Section 3. By shifting the focus from the narrator to the audience, we will be able, I think, to fashion an account true to the spirit and basic insights of Lewis’ accounts, and yet proof to the objections that can be made against them, and more generally applicable, and flexible.
SEMANTICS AND FICTION
261
1. LEWIS ’ COUNTERFACTUAL SEMANTICS
Lewis has a two candidates for the counterfactual analysis of metafictive statements. First he claims that any metafictive statement ϕ is elliptical for ‘In the fiction, ϕ’, then he proposes the following analysis (Lewis 1983, p. 270): Analysis 1: A sentence of the form “In the fiction f , ϕ” is non-vacuously true iff some world where f is told as known fact and ϕ is true differs less from our actual world, on balance, than does any world where f is told as known fact and ϕ is not true. It is vacuously true iff there are no possible worlds where f is told as known fact.
These are basically the Lewisian truth conditions of a counterfactual in which the antecedent is: ‘the story is told as known fact’ and the consequent is the metafictive statement (Lewis 1973, pp. 13–19). You will notice that the counterfactual, instead of merely stipulating that the story is true, stipulates that it is told as known fact, i.e., that it is true, that it is told, somehow, and that the teller knows that it is true (Lewis 1983, p. 266). The stipulation that the stories be told, that there be a narrator, can be motivated by certain characteristics of fiction. For one thing, many fictions seem on the face of them to purport to have narrators, particularly ones in the first person (Lewis 1983, p. 266). For another thing, fictions make use of implicature, and implicature, as is well known, involves inference to the intentions of the utterer. Yet the actual utterers actual intentions do not seem to be relevant, since, from his point of view, virtually every sentence of the fiction flouts conversational maxims, inasmuch as it’s expected to be false, and there’s nothing to distinguish the ones that call for implicature in the fiction (Lewis 1983, p. 266). A third argument for a fictional narrator come from an observation by Saul Kripke. He points out that the real world could turn out to echo a fiction (Kripke 1980, pp. 157–158). Suppose, for instance, that, unbeknownst to Joseph Conrad, all the events of his The Secret Agent took place in real life, with people exactly like those described in the story, down to the names, doing exactly what the story said, in every particular. In that case, if the antecedent of our counterfactual doesn’t mention a knowing narrator, and only specifies that the plot of the story take place, or some such, the actual world would be the closest world in which the antecedent is true, and the fiction would describe actual events. But it does not seem that in the real world the name ‘Verloc’, as used by Joseph Conrad in the story, refers to the actual Verloc; Conrad knew nothing of that man; it is sheer coincidence that they share a name and match the same description. Yet, in the story, the name ‘Verloc’ clearly does refer to the Verloc described in the story. Thus there’s something true in the fictional world that’s not true in
262
DANIEL A. KRASNER
the actual one, so the actual world can’t be the fictional world. There must be further differences. Lewis suggests a fictional narrator, who unlike the actual Conrad, knew the truth of what he was telling (Lewis 1983, p. 265). It might be objected that someone might write the story of the actual Verloc word for word the same as Conrad’s story of the fictional. Lewis replies that this would not make the actual world a fictional one, or Conrad’s Verloc the actual one, because the two texts are not the same stories, since they arise from different acts of storytelling; so, since the fictional world is one in which the same story is told as true, we still have some difference with the actual world. Being the same story is not merely a matter of lexical identity, according to Lewis, but of the acts of generation (Lewis 1983, p. 267). That is Lewis’ first serious proposal for a metafictive semantics. He finds some motivation to seek another, however which I shall now explain. Let us take Charlotte Brontë’s novel Jane Eyre as an example. In the novel, use is made of the assumed science of phrenology as a shorthand way to delineate character, although none of the tenets of phrenology are ever explicitly said to be true. Phrenology was taken seriously at the time, but we now know that phrenology is rubbish, that none of it’s conclusions can be taken seriously. Shall we therefore conclude that the fact that the various characters’ heads correspond to their characters in ways predicted by phrenology is merely a strange coincidence? We would be obliged to do so, if we accepted Lewis’ first stab at fictional semantics, since worlds in which phrenology is rubbish would be closer than worlds in which it isn’t, but it seems at least a little odd. Lewis therefore proposes the following analysis (Lewis 1983, p. 273): Analysis 2: A sentence of the form “In the fiction f , ϕ” is non-vacuously true iff, whenever w is one of the collective belief worlds of the community of origin of f , then some world where f is told as known fact and ϕ is true differs less from the world w, on balance, than does any world where f is told as known fact and ϕ is not true. It is vacuously true iff there are no possible worlds where f is told as known fact,
where collective belief worlds are worlds at which all the things commonly and overtly believed in a given community are true. Thus, Jane Eyre is evaluated at worlds where phrenology works, and the passages regain their points. Lewis doesn’t choose between these two semantics, but leaves the reader to decide for himself which is preferable (Lewis 1983, p. 271).
SEMANTICS AND FICTION
263
2. PROBLEMS WITH LEWIS ’ ANALYSES
2.0. Overview These are Lewis’ proposals for fictional semantics. I would like to discuss a number of ways in which I think they fall short, and are capable of improvement. In Section 2.1 I shall consider concerns that arise with regard to trying to specify the fictional worlds by means of the act of storytelling. In Section 2.2 I shall consider problems involving the implicit stipulation that fictions have a reliable narrator apiece. In Section 2.3 I shall consider problems that raise questions as to how exactly fictions must coincide with the facts of their counterfactual worlds. The objections I shall moot are chosen not so much for their original merit, as many of them have occurred in the literature before, but because I feel they highlight fundamental failings in Lewis’ analyses, an understanding of which helps to point the way to a more satisfactory account. 2.1. Problems IS and DM: Transworld Identity of Stories The first problem I call IS for individuating stories. It is crucial to Lewis’ semantics that the same story be told as fact in one world and fiction in another. Identity of stories, according to Lewis, is partly a matter of wordfor-word identity of texts, and partly a matter of identity, or similarity, among acts of storytelling (Lewis 1983, p. 267). Here the account gets obscure. What is this act? And how can two of them be identical, or even similar, when one is the recounting of fact, and the other the spinning of a fiction? Presumably the factual account would consist of assertions, with a sprinkling of other normal speech acts, while the fictional would be something entirely different. It would thus seem that acts of storytelling can’t be reduced to any of the speech-acts we’re familiar with. What are they? All this does not amount to a solid argument against Lewis’ position. I was not aiming for a solid argument against Lewis’ position, however. I was aiming to show that that position is in need of further elaboration, explication and refinement. Lewis doesn’t specify what he means by ‘story’ or ‘act of storytelling’, merely supposing that some suitable specification can be had. I think that there’s enough room for doubt that we should make an attempt to see how such a specification should go, to see that the difficulties don’t overwhelm us. One salient difficulty has to do with whether stories are supposed to be partly or wholly semantic entities, or whether acts of storytelling are individuated in terms of semantics. Lewis seems undecided on this; at one point he says that for identity, the stories “match word for word, and the
264
DANIEL A. KRASNER
words have the same meaning” (Lewis 1983, p. 267), then, in the next paragraph, he speaks of fictional names having different semantics in worlds in which the story is told as fiction from that in worlds in which the story is told as fact (Lewis 1983, 267). If a story is a semantic entity, then the content of the story has to be the same in the fictional world as in the actual, because it’s the story that is stipulated as being the same in the two worlds. But which world determines the meaning of the sentences that make up the story? (I call this problem DM for determining the meaning.) In the actual world, not all of the words have meanings, particularly not the fictional names, but also sometimes other words, as in The Clockwork Orange, in which people talk in a futuristic slang unlike anything real. If these words are to have their meanings, it seems they’ll have to get them from the fictional world(s). But if we say that the meanings of these words in the actual world are derived from their meanings in their fictional worlds, not only do we strangely neglect everything we know about how fictions are actually written, which is not by intuiting merely possible narratives and writing them down like a medium, but by forming them out of a language we already know, but we also run into a sort of a circle. The words actually mean what they mean in the relevant possible worlds, but a possible world is relevant just in case it contains the same story, and the story is the same only if in it the words have the same meaning. These truths are supposed to be stipulative, but it’s a stipulation of the meaning of words in terms of their meanings. None of this shows that Lewis is wrong. What it does show, I believe, is that we must either understand stories, as the word is used by Lewis, as syntactic entities, or else abandon them for such. We must eschew semantics in the antecedent of our counterfactual, and instead make use of something like the text. 2.2. Problems NN and UN: Stipulation of Sincere Narrator Challenges can be made to the stipulation that the story must have a unique, reliable narrator. Sometimes things are described in fiction that it seems no one person would be in a position to relate to an audience (Byrne 1993, pp. 29–30; I call this NN for no narrator.) Similar problems arise with regard to statements that discuss characters or events from more than one fiction, such as “Holmes was smarter than Poirot”. Currie calls such uses of fictional names transfictive (Currie 1988), and I shall speak of transfictive discourse. It seems unintuitive to regard the Holmes and Poirot stories as having had a common narrator. Although Lewis, in his original paper, expressly denies having any semantics for transfictive utterances (Lewis 1983, p. 263), he does give hints
SEMANTICS AND FICTION
265
of one in an appendix he subsequently added (Lewis 1983, p. 276), hints that could be expanded to deal with some of the other cases mentioned in the previous paragraph. The idea is (using the aforementioned example for exposition) that the speaker pretends he came across (presumably) nonfictional versions of both the Holmes stories and the Poirot stories, and the sentence is true if it’s true in the closest worlds in which such a pretense is true, if there are any. Lewis evidently intended this to be a supplement to his previous proposals, so that truth would come, somehow, both from the pretense that the author is a reliable narrator, and this one. It’s not clear how this would work, but it’s seems clear that the pretense of reading is doing most of the work in this case, and if we can get a better semantics for the above mentioned utterances by changing the focus from the narrator to the audience, as Lewis himself implies, that would seem a reason for trying to change it for metafictive utterances, if we can, for the sake of a unified theory. Another, yet more serious, objection is that sometimes it seems clear that the narrator is not telling the truth. Lewis is aware of this problem and discusses it in a long footnote (Lewis 1983, p. 266, fn. 6). Sometimes the storyteller purports to be uttering a mixture of truth and lies about matters whereof he has knowledge, or ravings giving a distorted reflection of the events, or the like. . . . In these exceptional cases also, the thing to do is to consider those worlds where the act of storytelling really is whatever it purports to be – ravings, reliable translation of a reliable source, or whatever – here at our world.
The suggestion is that, instead of having one analysis for fiction, we have many, one for normal fiction (one of those we’ve been considering), one for fiction narrated by liars, one for fiction narrated by madmen, and so on. The analyses would have similar form, in that, in each case, we would evaluate the fiction as if it were “whatever it purports to be . . . here at our world”. Not only does this suggestion seem to rule out a uniform treatment of fictions, it also clearly determines the wrong fictional worlds. We are supposed to take the worlds where the fiction is what it purports to be at our world, but at our world it generally purports to be fiction, and few would mistake it for anything else. The sort of purporting that seems to be relevant is what the fictions purport to be in the fictional world(s), giving the ‘purports’ operator wide scope, and the ‘in the fiction’ operator narrow scope. We cannot use this formulation, however, for we would be stipulating what the fictional worlds were by means of a sentence mentioning the fictional worlds; our stipulation would be circular. We must find a way of picking out the fictional worlds that doesn’t presuppose that we already have them picked out; we can’t use the narrator’s relation to the truth to do that (the argument is similar to one made by Currie 1990, p. 70).
266
DANIEL A. KRASNER
The stipulation that the stories are told by a unique reliable narrator marks a difference between Lewis’ account of fictions and our ordinary way of approaching nonfiction. In the latter case, we are rarely explicitly guaranteed a reliable interlocutor beforehand, and yet we often, though not always, manage to trust them anyway. This, and the considerations adduced above, NN, and UN, suggest that the stipulation is neither necessary nor sufficient for a good understanding of fiction. We need some more flexible formulation that gives us a single reliable narrator when we intuitively should have one, but allows for other kinds of narration when they’re called for. 2.3. Problems IW and MI: Problems of Truth There are also problems related to what Lewis has to say about the nature and source of fictional truth. It might seem that the previous few objections dealt with this by dealing with the reliability of the narrator, but I think it’s possible to find cases in which, even though we are intuitively inclined to find a unique reliable narrator as Lewis requires, there still seems to be something wrong about his account of fictional truth. I shall consider a few cases next. Fictions sometimes seem to be indeterminate in ways which Lewis’ accounts wouldn’t allow. In Emily Brontë’s Wuthering Heights, for example, it is deliberately left open whether Heathcliff and Kathy’s ghosts wander the moors after their deaths, for important dramatic reasons. Yet, the matter would have to be settled under either of Lewis’ analyses. In fact, there are no ghosts, so there would be none in Wuthering Heights on the first analysis. At the time the book was written, literate Englishmen did not believe in ghosts, so Heathcliff and Kathy’s fate would be the same on the second analysis. (This example was given me in private communication by James Krasner, a professor of Victorian literature.) Lewis can allow for a sort of indeterminacy when some of the closest worlds would be some way and some another, but that doesn’t seem to be the case here; here, all the closest worlds would be without ghosts. I call this problem IW for indeterminate worlds. Then, there is the problem that things sometimes seem to be true in a story that are metaphysically impossible (problem MI). These couldn’t be true in any possible world, but Lewis’ counterfactual semantics require the use of possible worlds. He proposes to deal with inconsistent fictions by decomposing them into consistent fragments and basing his account of truth in the whole on truth in the parts. One way he suggests would be to call a sentence true if it’s true in all the parts, the other, to call it true if it’s true in any, and not allow implication from mixed premises from
SEMANTICS AND FICTION
267
different parts (Lewis 1983, p. 277). There are stories this doesn’t work for, however. In Ray Bradbury’s “A Sound of Thunder”, a time travel expedition changes the present, so that something that happens, doesn’t happen. The moral of the story is that a small change can make a big difference, but, unless we can reason from the claim that the event happened and the claim that it didn’t, both, we can’t draw that moral, and either of Lewis’ proposals would block that reasoning. Both ways, besides, would oblige us to have separate semantics for possible and impossible fictions, which is uneconomical. The above arguments exploit tension in the requirement that fictions be judged as being literally true in some possible world(s) (as opposed to the requirement that they be spoken reliably, which I considered in the previous section.) They suggest that we need some further distance from the subject matter of the story than is provided by an infallible fictional narrator. They also suggest that we might do better to allow incompleteness and contradiction. As Gregory Currie has pointed out, incompleteness and contradiction are characteristic of belief systems (Currie 1990, p. 74). Thus, we might wonder whether defining fictional truth partly in terms of belief might be possible within a broadly counterfactual semantics. I intend to consider this, and the other lessons of Section 2, in the next section. 3. A NEW COUNTERFACTUAL SEMANTICS FOR FICTION
3.0. Analysis and Overview All the above problems argue for a different semantics than the ones proposed by Lewis. Let me review what sort of differences we would want to see. We would want to see the relevant possible worlds determined by some syntactic aspects of the story. We want more flexibility in the requirement there be a truthful narrator, with some presumption that there is such, perhaps, but one that can be overridden by the evidence of the stories, and this can possibly be gotten by shifting the focus from the narrator to the audience. We want some distance between what is true in the story and what is true in a possible world, possibly of a sort that can be provided by a doxastic account. I shall now propose an account that integrates such changes into a roughly Lewisian counterfactual framework. Assume that metafictive remarks have an implicit in-the-fiction operator, and analyze it as follows. ‘In the fiction f , ϕ’ is true iff ‘If a likely audience member came across the lexically defined utterance of f , or a text of which it would be a good translation, under circumstances in which he had some justification for believing it wasn’t fiction and no justification for the belief that it was, it would be reasonable for him to believe ϕ’ is true.
268
DANIEL A. KRASNER
By “a likely audience member”, I mean only one who is likely to come across it under the conditions specified. I am using ‘lexically defined utterance’, as it is commonly used, to mean a token of the sequence of uninterpreted sentences that makes up the fiction. I shall sometimes call this the ’text’, but one should not assume by that that I mean something written, a spoken token might well be more appropriate in some cases. This account requires a good deal of development. In Section 3.1, I shall consider the reason for using the lexically defined utterance as I do; in Section 3.2, the importance of the unsullied justification for believing it’s not fiction; in Section 3.3, why I make fictional truth a matter of justified belief rather than correspondence to fact in other possible worlds.
3.1. The Lexically Defined Utterance I shall now discuss the clause that makes the text or something of which it would be a good translation the constant feature between possible worlds. I allowed for translations because there are some stories, e.g., science fiction stories set among alien species, that it seems could not have had an English translation, although that’s how we read them. This provision may be otiose. It’s possible that there’s a sense of text according to which translations of a text are the same text; after all, when we read a book in translation we ordinarily say that we are reading the book, not that we are reading another book that means the same thing. But I did not want to have to argue for such a difficult and peripheral point, so I simply stipulated that translation was allowed. Some might worry that adding the phrase “or a text of which it would be a good translation” would introduce unacceptable ambiguities into the interpretation of texts. After all, there could, theoretically, be two languages which produced tokens syntactically identical, but semantically distinct, so that a given work might have two radically different translations. I agree that this is a logical possibility, but I submit that, in the closest world in which a typical work of fiction had a translation, it would be translated in the most obvious and natural way, so as to respect the actual meanings of words in the texts. Also, someone once suggested that, by allowing for translations, I am reintroducing semantical entities into the antecedent of the counterfactual, leading to circularity again. I deny the circularity. My antecedent does not require that we find a text with any specific semantic properties which we aren’t given; it merely requires that the semantics of the hypothetical text, whatever they may be, correspond to the semantics of the actual text, whatever they are.
SEMANTICS AND FICTION
269
One might wonder whether, from a sequence of uninterpreted sentences, one can get a meaningful story. Tyler Burge has recently argued that one has an a priori justification for regarding things that seem like utterances as having been produced by rational beings, and being in fact meaningful (Burge 1993, p. 471). Whether or not this is true, it seems to be the case that worlds in which we would be justified in believing them meaningful are closer than worlds in which we would not. Surely, if I came across something that seemed exactly like a newspaper article as far as subject matter and style went, I would be justified in believing that it did not come about through chance, and that it was a newspaper article, and meant what it seemed to mean. Thus, we get not only a meaningful story, but a fictional narrator, without stipulation, which gives us much of what Lewis stipulated, but with greater flexibility. But the main reasons for making this change have to do with problems IS (that we don’t know how to individuate acts of storytelling) and DM (that it seems impossible to determine the meanings of terms used in a specifically fictional sense, or precisely what worlds are relevant.) If one determines the fictional world by the lexically defined utterance, and not by its meaning, one doesn’t have these problems. We get the fictional worlds from the utterance, and get the meanings from the fictional worlds. We are saved from the danger of a circle. 3.2. Unsullied Justification for Believing Non-Fictional Now for the advantages of the unsullied-justification-for-believing-it-notfiction clause. Previously, I raised the possibility that there might be some defeasible reason to believe that a narration should be regarded as reliable, without explicitly stipulating it. I shall now provide one with the help of some current work in epistemology. Tyler Burge and others hold that one is a priori (albeit defeasibly) justified in believing the testimony of others (Burge 1993, p. 465; e.g., Coady 1992). I shall not argue for this here beyond saying that it is the preponderant view of the justification of testimony nowadays. Thus, if we have justification for believing of fictional assertions that they aren’t fiction, and are consequently testimony, we have a defeasible justification for believing that they are true, without the inconveniences of stipulating it. We have stipulated that we have such justification, and we have, I think, as much reason to believe in a reliable narrator as we need. We might well have reason to believe in a reliable narrator even without Burge’s theory of the justification of testimony; for, even if there is no a priori justification for it, we do generally accept testimony as reliable, and it is evidently right that we do so, and it would seem that in most counterfactual situations, we still would. Just ask yourself
270
DANIEL A. KRASNER
whether you’d be inclined to accept something that seemed like testimony, if you found it; I think the answer might well be yes. But I thought the common account of such justification was worth mentioning. If anyone finds both Burge’s theory, and my counterfactual suggestion implausible, I can simply amend my account to stipulate that the counterfactual audience member have as much justification for believing testimony in his world, as we have in ours. What about those cases in which a single reliable narrator seems awkward, problem NN for instance? We can infer to as many narrators as we like, and of whatever kind we like. In the case of transfictive utterances, we can postulate a virtual fiction, so to speak, composed of the separately narrated fictions referred to in the transfictive utterance, and speak of what’s true in that. Because I do not specify that a fiction was told in any way (only how it was read) we can regard these cases in these more natural ways. We do not have to regard the fiction as having been narrated by someone knowledgeable at all, if it’s easier not to; we can instead regard all or parts of it as a reconstruction or prediction of events that have not been observed, much as a forensic scientist might do, or, if the narrator seems to have access to knowledge no human could, like the undisclosed thoughts of a dying man, we can postulate a narrator with superhuman or even supernatural means of observation. We can also handle unreliable narrators (problem UN) easily. The assumption that the narrator is truthful is defeasible; if, based on internal evidence, it’s most natural to believe that he’s lying or deluded, the nearest worlds will be worlds in which he is. Some people might think that this formulation allows for unreliable narrators too easily. They might say that, in the case of a far-fetched fantasy or science fiction tale, or a story in which the narrator seems too knowledgeable, it might be better to regard the story as an elaborate lie or a madman’s ravings, and the worlds as substantially like ours. This is where the claim that we have no reason to believe the story fictional comes in. There are two basic sorts of reasons for believing that a story is delusional or ridden with lies. One is that it seems to be clearly false, based on evidence external to the text, the other is that it is written in the style of a liar or a madman. In these cases, the first sort of reason, in the absence of the second, would incline us to believe, not that the story is delusional or deceptive, but that it is fictional. After all, that is exactly what we do believe about such stories under similar circumstances. When presented with farfetched science fiction or fantasy tales that differ greatly from reality, and are written in a style that seems sane and trustworthy, we conclude that they are fiction, rather than lies or ravings. Likewise with stories in which peoples’ most private, incommunicable thoughts are revealed. In the actual
SEMANTICS AND FICTION
271
world we have evidence that such stories are fictional from the fact that they are advertised as fictional, in addition to the nature of the content, but does anyone want to maintain that he couldn’t tell that science fiction was fiction if he weren’t told? Or that he couldn’t guess at the fictionality of stories in which the narrator seems all but omniscient? But, if one can tell, that means that, in the absence of internal evidence that the story is written by a madman or a liar, the nearest satisfactory worlds would be ones in which the evidence supports the claim that the story is substantially true, because otherwise we would have evidence that it was fictional. So we should have unreliable narrators only when internal evidence, evidence from the story itself, warrants it, and that is what we want. There’s one last objection to consider. That is that sometimes fictions say they are fictions fictively. Kurt Vonnegut, for instance, in Breakfast of Champions appears as a character who also happens to be the author of the novel being read. This would seem, at first glance, to provide the hypothetical audience with a reason to regard the text as fiction that couldn’t be evaded. Cases like this are complex, but I believe may be handled in one of two ways. Either the fiction is a quote of someone telling a story, in which case what’s fictionally true is that someone’s telling a fiction, and the hypothetical audience member has no reason to believe fictional the account of its being told; or, if it’s impossible to extricate the claims of fictionality from the action of the story that way, as it is with Breakfast of Champions, the author is claiming the story is fiction only in a specialized, quasi-meinongian sense, a sense which doesn’t preclude interactions between the fictional and real worlds. I only require that the hypothetical audience member have no reason to believe the story fiction understood possibilistically, and the narrator’s claims provide no reason to do that. 3.3. Reasonable Belief I define fictional truth in terms of reasonable belief, instead of correspondence to fact. A benefit of this emphasis on reasonable belief that I’d like to point out is that it provides us with a way around problem MI (impossible fictions). Impossibilities can never occur, but it can be reasonable to believe that they do. I have studied Gödel’s incompleteness theorems in some depth, and know them to be true, but if someone who seemed knowledgeable enough and seemed to be speaking seriously assured me strongly enough that someone had disproved them, I might well come to believe them false, and I think I would be reasonable in doing so. Likewise, I disbelieve in the possibility of time travel, but if I found a text like Bradbury’s, and believed, with justification, that it was intended seriously, then I might come reasonably to believe that it had happened. (Believing
272
DANIEL A. KRASNER
with justification is the problem here. The mere fact that the subject matter is so far-fetched might make it unreasonable to believe in the author’s seriousness. To avoid this, the counterfactual antecedent would take us to a world where it is not far-fetched to believe in time travel, or another impossibility. There can be worlds wherein it’s reasonable to believe such. It’s just that the thing one has reason to believe can’t be the case.) Some might think this does not go far enough. A story might make true something of the form ’P and not P ’, which, one might think, it could never be reasonable to believe. I have doubts about both parts of this problem. The only clear-cut case of such a story I know is one by Graham Priest (Priest 1997, pp. 573–579), designed to show such are possible, and I am reluctant to draw conclusions purely from a philosophical thought experiment. Maybe there’s a reason real writers don’t write such. But if your intuitions differ, my account still has resources. What’s reasonable to believe depends on the epistemic situation of the believer. Someone who could believe a story with a blatant contradiction wasn’t fiction would likely have low standards for accepting things at face value. I am not ruling out an audience of children or half-wits or Hegelians, and it might well be reasonable for an audience of one of these types to believe things which better logicians would balk at. This formulation de-emphasizes literal content to some degree, and to a like degree emphasizes implicature. If the author writes in a way that suggests things that don’t quite jibe with reality, we nonetheless must consider his work from the point of view of one for whom it would be reasonable to believe that it weren’t fiction, and we can do this by assuming that the author made a mistake, that the fictional world is different from ours, or that the person who would read it is in an epistemic situation that would make it reasonable for him to believe things that we don’t. It might be indeterminate which is the case, some of the closest worlds might be one way and some the other. In the case of fictions that seem to depend on common beliefs that are at variance with the facts, we seem capable of arriving at the correct conclusions. We are obliged to regard the narrative from the viewpoint of an auditor for whom the presumption that it was not fictional were undefeated, and, in the case of fictions that depend heavily on common or once-common fallacies, that might be done most easily by assuming he shares the belief in these fallacies, since a reader who didn’t share the belief in the fallacies might be tempted to consider the stories fictions. (We could also assume that the fallacies are true.) Thus, we get the truth of phrenology in Jane Eyre. As for Wuthering Heights, although a likely audience member would presumably be a Victorian Englishman who didn’t believe in ghosts, a compelling tale of eye witnesses of such
SEMANTICS AND FICTION
273
might reasonably bring him to have some doubts about that belief, which would be enough to get us the indeterminacy we need for that book. Indeterminacy could also arise from differences of opinion among likely audience members – if they would disagree on whether P was true, it would be indeterminate. But there would be broad agreement on many things that would therefore be true in the story, even things having nothing to do with the story. For example, it’s true in the Sherlock Holmes stories that the Boxer Rebellion took place. I think this is intuitive; Sherlock Holmes didn’t live in a world that ended at the boundaries of the places described in the accounts; he lived in the world that we know. The use of doxastic and epistemic notions in analyses of fictional truth is not new with me; Currie did it first (Currie 1990, p. 80), and Alex Byrne and John F. Phillips have appealed to them in very similar ways (Byrne 1993, p. 33; Phillips 1999, p. 287). The presentation of these accounts so that they could be understood involves the explication of various technical terms and concepts, and space unfortunately does not permit me to do so here, but I can make some general comments on how my account improves upon them. The first two use make-believe to do much the same work that counterfactuals do in my account, getting us from the real to the fictional world. I think counterfactuals are better understood than make-believe, and so a better choice. Phillips’ analysis seems circular, in that he defines what’s true in a fiction in terms of what the author would want to be true in a fiction. All of them make use of idealizations that are introduced specifically for the purpose, and strike me as ill-defined. And all of them make reference to an author/narrator as well as a reader, though you can get the benefits of a doxastic account just from the latter, and though it sometimes leads them to implausible views about the author, as when Currie claims that the fictional author is both reliable and a member of the milieu which produced the book (Currie 1990, p. 79), which would make the narrator of Tolstoy’s War and Peace a late 19th century Russian who was intimately acquainted with the thoughts of some people from the early 19th century. I should like to consider one last objection before closing. It is that my account cannot deal with stories which claim that no one exists who can plausibly be the narrator. Let me begin by pointing out that it would be, at best, very difficult to write such a story on my account. Notice that I do not rule out any possible intelligent being as the narrator, and cases in which, say, the earth falls into the sun, can be regarded as having been narrated by extraterrestrials, or immaterial intelligences, or God. One would have to be very drastic really to rule out all possible narrators. Even if the story says there was no intelligent life of any kind, that ‘no’ can often be regarded as a restricted quantifier (maybe the narrator doesn’t think he counts.) Some
274
DANIEL A. KRASNER
proposed cases might also be regarded as jokes; a narrator tries to make fictional audience believe that he’s telling a true story, then reveals it to be a lie by an obvious falsehood at the end. If it were established, nonetheless, that there could be stories which plausibly had no narrator, which, I say again, is questionable, I would, I think, be forced to say one of two things: either that my account was good only for central cases of fictions, and might fail if a sufficiently odd example were adduced, or that these stories weren’t really fictions, and so didn’t fall within the purview of my account. I think the difference between these positions would be mainly terminological. In either case I’m saying that my account doesn’t cover them, and denying the intuitions that any account that purports to explain fiction should explain them. My personal preference would be for the latter; I’m inclined to think that if one comes up with an account that does a good job in the vast majority of cases, particularly the central ones (and I hope I have,) one is entitled to be a little arbitrary in peripheral cases in revising prephilosophical concepts. Whether you accept this or not, however, what I have to show, in either case, is that stories in worlds without intelligent beings, if they exist, are sufficiently different and peripheral to justify separate treatment. I think this is intuitive. A story without a narrator would be radically deficient in characters, since they are usually intelligent beings, and in plot, since it’s difficult to imagine a plot without intelligent characters, and authorial persona, since much of that would suggest that the narration was intelligently produced. It would be very defective fiction, if it were fiction at all. I don’t think anyone has ever written such a story except with the pure intention of producing an ostensible example of fiction without intelligent life in it. Even if you think they succeeded, it is open to us to say such cases are simply too exceptional to be considered counterexamples to an account aiming to deal with paradigm cases of fictions.
REFERENCES
Burge, T.: 1993, ‘Content Preservation’, The Philosophical Review 102, 457–488. Byrne, A.: 1993, ‘Truth in Fiction: The Story Continued’, Australasian Journal of Philosophy 71, 24–35. Coady, C.: 1992, Testimony: A Philosophical Study, Clarendon Press, Oxford. Currie, G.: 1988, ‘Fictional Names’, Australian Journal of Philosophy 66, 471–488. Currie, G.: 1990, The Nature of Fiction, Cambridge University Press, Cambridge. Kripke, S.: 1973, ‘Reference and Existence, Lecture 1’, John Locke Lectures. Kripke, S.: 1973, ‘Reference and Existence, Lecture 3’, John Locke Lectures. Kripke, S.: 1980, Naming and Necessity, Harvard University Press, Cambridge. Lewis, D.: 1973, Counterfactuals, Harvard University Press, Cambridge.
SEMANTICS AND FICTION
275
Lewis, D.: 1983, ‘Truth in Fiction’, in Philosophical Papers, Vol. I, Oxford University Press, New York, pp. 261–280. McMichael, A. and E. Zalta: 1980, ‘An Alternative Theory of Nonexistent Objects’, Journal of Philosophical Logic 9, 297–313. Parsons, T.: 1980, Nonexistent Objects, Yale University Press, New Haven. Phillips, J.: 1999, ‘Truth and Inference in Fiction’, Philosophical Studies 94, 273–293. Priest, G.: 1997, ‘Sylvan’s Box: a Short Story and Ten Morals’, Notre Dame Journal of Formal Logic 38, 573–582. Routley, R.: 1979, Exploring Meinong’s Jungle and Beyond, Australian National University Press, Canberra. Van Inwagen, P.: 1977, ‘Creatures of Fiction’, American Philosophical Quarterly 14, 299– 308. Walton, K.: 1990, Mimesis as Make-Believe: On the Foundations of the Representational Arts, Harvard University Press, Cambridge. 3270 Motor Avenue, #18 Los Angeles CA 90034 U.S.A.