JOHN F. PHILLIPS

TRUTH AND INFERENCE IN FICTION (Received in revised form 4 March 1997)

I

Anyone who has read Catch-22 is probably aware of the fact that Yossarian wants to live forever or die in the attempt. Moreover, anyone who has read the novel would probably agree that the statement ‘Yossarian wants to live forever or die in the attempt’ is true if understood as ‘In the novel Catch-22, Yossarian wants to live forever or die in the attempt’. If someone disagreed with this statement, we would simply point to the relevant passage in the novel where Yossarian’s comical ambition is explicitly mentioned and that would, seemingly, be the end of the matter. There is, however, a philosophical puzzle which becomes apparent when our discourse about fiction is subjected to closer scrutiny; for there is a certain class of statements about fiction which seem to be true, despite the fact that there are no passages in the text in which any of the statements in question are clearly expressed. For example, consider the statements ‘Yossarian and Dunbar are friends’ or ‘Milo Minderbinder is obsessed with making money’. While neither statement is, so far as I know, explicitly set forth in the text, those of us who have read the novel would, I take it, agree that both statements are true. The puzzle, then, is, if we agree that these statements and others like them are true, in spite of the fact that they are not explicitly stated in the text, what is it that makes them true? Perhaps the most straightforward and intuitive response to this puzzle is to claim that ‘Yossarian and Dunbar are friends’ is true because it is reasonable to infer, upon the basis of what is explicitly stated in the novel, that (in the novel) Yossarian and Dunbar are friends. This proposal has a great deal of merit, and it is a first step toward the solution of the more general problem of specifying truth-conditions for statements about fiction.1

Philosophical Studies 94: 273–293, 1999. c 1999 Kluwer Academic Publishers. Printed in the Netherlands.

274

JOHN F. PHILLIPS

In this paper, I shall consider two well-known accounts of how it is that truth-values are assigned to statements about fiction, the accounts of David Lewis and Gregory Currie. I shall argue that neither of these accounts is satisfactory and I shall propose, in their stead, my own account. After presenting my proposal, I shall endeavor to show that my theory correctly handles all of the problem cases for the theories of Lewis and Currie and, hence, is preferable to those theories. II

In order to set the stage for a discussion of the views of Lewis and Currie, it will be useful to first consider some alternative views. A story or a piece of fiction, I shall assume, is essentially a collection of propositions, where a proposition is the semantic content or what is expressed by a declarative sentence. Propositions are distinct from sentences, since two distinct sentences may express the same proposition. This is demonstrated by the sentences: (1) (2)

Rumpole is a barrister. Rumpole is a counselor-at-law.2

Thus propositions are not sequences of words, but rather what is expressed by certain sequences of words.3 Given this conception of fiction, the problem of determining the truth-conditions for statements about a story is simply the problem of determining which propositions are members of the class of propositions which constitute that story. In other words, if a story is a collection of propositions, then ‘Odysseus is a crafty fellow’ is true in the Odyssey just in case the proposition expressed by ‘Odysseus is a crafty fellow’ is contained in the collection of propositions of which the Odyssey is comprised. Evidently, what is required if we are to have truth-conditions for a given story is what I shall term a principle of story composition, which is just a principle for determining precisely which propositions are contained in a given story. It follows from the claim that a story is essentially a class of propositions that it is possible for two different authors to write the same story. For example, in Borges’ tale of Pierre Menard, the story Don Quixote penned by Menard is identical to the Don Quixote of

TRUTH AND INFERENCE IN FICTION

275

Cervantes if and only if these stories consist of exactly the same propositions. The question of whether these stories are identical will be resolved by appealing to the principle of story composition, which will determine whether the stories consist of the very same propositions. Thus the principle of story composition not only determines what is true in a story, but it also plays a crucial role in what we might call ‘story individuation’. Two of the most obvious and intuitive candidates for the principle of story composition are drawn from the simple observation that stories are, in some sense, created by their authors. The first candidate for the principle of story composition is arrived at by taking the story to be a product of its author. A story is constructed according to the author’s plan, but once the construction is finished, the story stands alone and apart from its author. The author’s intentions play a role in how the story is constructed, but once this work is completed it is the product of this labor, and not the author’s intentions, which determines what propositions are in the story. Thus, the first candidate for the principle of story composition is the principle that all and only those propositions expressed by the text of the author or in the utterances of the storyteller are contained in a given story. This principle gives rise to the view that what is true in a story is what is explicitly mentioned in the text; nothing more and nothing less. I shall label this view the Literal View.4 The Literal View, as we have already seen, appears to be an inadequate theory. The Literal View fails to account for the truth of statements such as ‘Iago is duplicitous’, ‘Tom Sawyer is a human being’, ‘King Lear had a consort’, ‘Holmes has a blood type’, and countless others like them. Intuitively, there are many propositions that are part of a story which are not expressed by any sentences in the text. A better theory, then, would be one which could account for the truth of statements of this kind. This leads us to a second candidate for the principle of story composition. A proponent of the second candidate might begin by noting that a story is conceived of or imagined by its author. When Jack London imagined the adventures of an unimaginative man travelling through the Yukon in mid-winter, a man too foolish to realize he is freezing to death, what he imagined was a story. London entitled the story To Build a Fire and wrote and published it.

276

JOHN F. PHILLIPS

Looking at matters in this way, it might seem reasonable to suppose that whether a proposition is part of To Build a Fire depends upon whether London conceived of it as being in the story. Thus what is true in the story is just what the author intended to be true in the story, and the author’s intentions are formed with reference to what he has imagined. This view has a certain intuitive plausibility, but as it stands it is in need of revision. One problem arises when we consider a story which, due to some oversight by its author, contains a contradiction. The stock example of this is the shifting location of Watson’s war wound in the Holmes stories, but there are many others. In Alice in Wonderland, there is a table on which there both is and is not a milk-jug, and Anna Karenina both begins and does not begin on a Monday.5 The relevance of examples of this kind is that occasionally authors, such as Conan Doyle, Carroll, and Tolstoy, make mistakes. But it seems wrong to saddle authors with mistakes of this kind. A reasonable solution is to build a mechanism into the above account to handle these situations. One way to do this is to modify the view so that what is true in a story is what the author would intend to be true in the story under what might be called ideal conditions. Conditions are ideal if the author is cooperative and clear-headed, and able to specify each proposition which is a part of the story he has imagined.6 I shall call this revised view the Intentional or Authorial View. As we have seen, the Literal View is an inadequate theory, since it counts propositions like ‘King Lear had a consort’ as false, when such propositions are intuitively true. The Authorial View, likewise, is flawed. Suppose Coleridge imagined a comic caper involving a bungled attempt to steal the Crown Jewels. After thinking up this tale, Coleridge puts pen to paper and writes Rime of the Ancient Mariner. Now if anything is true in Rime of the Ancient Mariner, it is that an ancient mariner accosts a man who is on his way to a wedding and relates to him a tragic tale. But what if Coleridge, were we to have asked him, would have vehemently denied this? If Coleridge, in full possession of his faculties, were to honestly claim that his poem has nothing to do with an aged seafarer relating a tragic tale, but is, instead, about a failed jewel heist, it follows, on the Authorial View, that in Rime of the Ancient Mariner it is true that

TRUTH AND INFERENCE IN FICTION

277

thieves botch their attempt at making off with the Crown Jewels, but it is false that an ancient mariner detains a man en route to a wedding and tells him a tragic tale. This, of course, is absurd. Rime of the Ancient Mariner, whatever Coleridge may have intended or imagined, is not about an unsuccessful burglary. Similarly, even if Conan Doyle intended to disclose, in a final story which was never written, that Holmes was a Martian masquerading as a human, it is nonetheless true in the Holmes stories that Holmes is a human being.7 These examples, and the many others that might be concocted, show that, in a wide range of cases, the Authorial view fails to accord with our intuitions regarding what is true in a work of fiction. On the Authorial View, much that is intuitively false or indeterminate may turn out to be true. Contrast this with the Literal View on which many propositions that are intuitively true turn out to be indeterminate. This contrast suggests that the Authorial View’s failure is closely connected with the intuition that motivates the Literal View, the intuition being that a story, in some sense, stands on its own. The reason we think that Holmes is human regardless of Conan Doyle’s intentions is that we think there is a sense in which a story, whether written down or spoken aloud, is divorced from the intentions of its author. The upshot of this discussion, it seems to me, is that the correct account of what the truth-conditions for statements about fiction are must strike a balance between the intuition, on the one hand, that since an author creates his story, the author’s intentions ought to play some role in determining what is true in his story, and, on the other hand, the intuition that there is a sense in which the story is a product of the author, and not a creature of the author’s imagination. We might fuse these separate intuitions into a single precept: a story is the product of its author’s imagination. With this admittedly vague principle in hand, I turn now to the theories of Lewis and Currie.

III

David Lewis has proposed two distinct theories, each of which provides necessary and sufficient conditions for the truth of statements about fiction.8 A number of criticisms have been levelled by

278

JOHN F. PHILLIPS

Gregory Currie at each of these theories.9 As it turns out, most of the difficulties which face Lewis’ first proposal are surmounted by his second proposal. Thus, Lewis’ second analysis appears more viable, and, in a wide variety of cases, this proposal yields intuitive results. I think, however, that we have good reasons, over and above the objections raised by Currie, for also rejecting Lewis’ second proposal. For Lewis, a belief is overt in a community just in case nearly everyone in the community has this belief, nearly everyone in the community believes that nearly everyone in the community has this belief, and so forth. The possible worlds where the overt beliefs of a community are true are what Lewis calls the collective belief worlds for that community. Lewis’ second proposal is captured in the following definition: (L1): A sentence of the form ‘In the fiction F, ’ is true if and only if, for any collective belief world w of the community of origin of F, there is some world where F is told as known fact and  is true which is more similar to w than any world where F is told as known fact and  is false.10 If we accept Lewis’ possible worlds analysis of the relevant subjunctive conditional, his proposal may be formulated more simply as: (L2): A sentence of the form ‘In the fiction F, ’ is true if and only if  would be true according to the overt beliefs of the community in which F originated, if F were told as known fact. Thus, on Lewis’ second analysis, the truth-conditions for statements about fiction are supplied by appeal to a counterfactual which incorporates the overt beliefs of the community of origin of the story, and what is true in the story on the Literal View. One problem with this account is that Lewis’ second proposal, like his first, results in many extraneous facts being imported into the story. There are many questions one might ask of a story which have no determinate answer because the story simply does not address those questions. In other words, there are many propositions which are irrelevant from the standpoint of any given fiction; the fiction fails to comment on or involve these propositions in any way, and,

TRUTH AND INFERENCE IN FICTION

279

hence, these propositions are not part of the fiction. One feature of Lewis’ analysis, however, is that any widely held belief in the community of origin for a particular story is true in that story. So, for example, if it was widely believed in Victorian England that there are eight planets in the solar system or that the Queen is a bit of a prude, then, so long as nothing to the contrary is stated explicitly in the stories, it is true in the Holmes stories of Conan Doyle, the Alice stories of Carroll, and any other fiction of the period, that there are eight planets in the solar system and Victoria is a bit of a prude. As another example, suppose that nearly everyone in Elizabethan England believed in God. On Lewis second proposal, it is true in Julius Caesar that God exists. This does not seem right. While it is true that the proposition ‘God exists’ is not explicitly denied in Julius Caesar, it’s hard to see how that fact, combined with the widespread belief among Shakespeare’s contemporaries that God exists, ought to make it true in Julius Caesar that God exists, especially since this proposition is irrelevant to the story. Perhaps Lewis could provide convincing arguments which show that, for every proposition, either that proposition or its negation must be a part of any given story. This would defuse the above objection to Lewis’ second proposal. This seems unlikely, but there are, in any case, two other important objections to Lewis’ view. First, there is the problem generated by statements like ‘Al and Max are mob hit men’ or ‘Isabella takes after Mr. Woodhouse more than Emma does’. Statements of this kind pose a problem because, although these statements are true, they are neither explicitly made nor are they entailed by any statements that are explicitly made in their respective stories. The reason it is true in The Killers that Al and Max are mob hit men is that all the available evidence suggests that this is true. Al and Max admit they have come to town to kill a man they do not know; their dress and speech pattern, the fact that their intended victim was a boxer, and the discussion of the intended victim double-crossing somebody in Chicago combined with a lack of any evidence to the contrary combine to make it clear that Hemingway intended it to be part of the story that Al and Max are mob hit men. This example is especially compelling since, to claim that it is false, merely probable, or just not part of the story

280

JOHN F. PHILLIPS

that Al and Max are mob hit men, is simply to misunderstand the story. Examples of this kind suggest that, if there is a preponderance of textual evidence that points toward a proposition’s being true in a given fiction, we take the proposition to be true in the fiction. Thus a strong inductive argument, in the case of fiction, can completely secure the truth of its conclusion.11 In the real world, a strong inductive argument indicates that its conclusion is probably true; in fiction, a strong inductive argument establishes its conclusion as true simpliciter. The problem for Lewis, then, is that his second proposal appears to be unable to handle cases involving statements of this kind. If Emma were told as known fact, would it be true that Isabella takes after her father more so than Emma? Perhaps, but perhaps not. Lewis’ second proposal provides us with two potential means of obtaining the correct answer. Unfortunately, both Emma and overt belief in Austen’s England are of no avail. The former is inadequate since it is neither stated in nor entailed by the story that Isabella is more like her father than Emma; the latter is likewise deficient, since there were no overt beliefs in Austen’s England which insure that it would be true that Isabella bears a greater resemblance to Mr. Woodhouse than does Emma, if Emma were told as known fact.12 Appeal to possible worlds appears to be of no advantage here either. Indeed, on Lewis’ possible worlds account of the relevant counterfactual in (L1), it turns out to be false that Isabella is more like her father than Emma. For it is not the case that, for any collective belief world w, there is a world where Emma is told as known fact and it is true that Isabella takes after Mr. Woodhouse more than Emma which is closer to w than any world where Emma is told as known fact and Isabella does not bear the greater resemblance to her father. In order to see this, one only has to consider the collective belief world w which is closest to the actual world. For any world w where Emma is told as known fact and where Isabella takes after her father more than Emma, there is a very similar world w where Emma is also told as known fact, but where Emma bears a greater resemblance to her father. Given that the only differences between w and w involve propositions concerned with the traits of Emma, Isabella, and Mr. Woodhouse, a comparison of w and

TRUTH AND INFERENCE IN FICTION

281

w to w suggests that w and w are equally close to w. This is because w simply provides no means of discriminating between w and w .13 Inability to deal with cases of this kind, it seems to me, is a substantial problem for Lewis’ view. Inconsistent fictions present another important problem for Lewis’ second proposal. For an inconsistent fiction F, there is no possible world in which all the propositions of F are true. But then, according to (L2), everything is true in an inconsistent fiction. This is plainly wrong, and Lewis has taken steps to remedy the situation.14 His suggestion is that, for an inconsistent fiction F,  is true in F just in case  is true in some logically consistent fragment of F. I find this suggestion to be fairly agreeable; in nearly every case, it seems to provide intuitive and correct results. This suggestion is, however not without fault. One of its shortcomings is that it is ad hoc.15 A better account would be a unified theory which supplied truth-conditions for every fiction, rather than singling out inconsistent fictions for special treatment. A second problem is that Lewis’ suggestion may give the wrong results in cases in which the contradiction in the story is inadvertent. In Alice in Wonderland, there is and there is not a milk-jug on the table at the mad tea party. As any attentive reader will note, this contradiction in the story is clearly an oversight on Carroll’s part; he obviously intended for there to be a milk-jug on the table. However, on Lewis’ suggestion, it is true in the story that there is a milk-jug on the table and it is true in the story that there is not a milk-jug on the table. Intuitively, it is true in the story that there is a milk-jug on the table and any suggestion to the contrary ought to be ignored. In light of the various difficulties which face Lewis’ proposals, it would seem that alternative theories are worth seeking. Perhaps the most important and viable alternative theory is the theory advocated by Gregory Currie. Both the successes and the failures of Currie’s theory serve to motivate my own account. Thus, it will be useful to assess the strengths and weaknesses of Currie’s theory before considering my own.

282

JOHN F. PHILLIPS

IV

The basic idea underlying Currie’s account is that truth in fiction can be likened to belief in the actual world.16 Currie’s proposal, in essence, is that what is true in a story is what the author or storyteller believes. The author that Currie has in mind, however, is not the actual author, but rather a fictional author, a character in the story who relates the events of the story to us as known fact. The fictional author, like the actual author, is a member of the community in which the story told by the author originates. On Currie’s view, if certain beliefs prevail in this community, then, barring any evidence in the text to the contrary, it is reasonable to suppose that the beliefs of the fictional author conform to the prevailing beliefs of the community. Currie’s proposal can thus be stated as follows: (C):

A sentence of the form ‘In the fiction F, ’ is true if and only if it is reasonable for an informed reader of the text to infer that the fictional author of F believes that .

An informed reader, for Currie, is a reader who knows the community of origin of a story and is familiar with what was commonly believed in that community. Thus, Currie’s proposal is in the same vein as Lewis’ second analysis; what is true in a given fiction is a product of what is true on the Literal View and the beliefs of the community of origin of the fiction. Currie’s proposal is, at least initially, fairly plausible, and it has several advantages over Lewis’ proposal. Most of the difficulties discussed earlier in connection with Lewis’ view are neatly avoided by Currie’s proposal. The case involving Isabella and Emma, for instance, is easily handled by Currie’s theory. The abundance of textual evidence in support of Isabella’s being more like her father than Emma makes it reasonable to infer that the fictional author believes that Isabella takes after Mr. Woodhouse more so than Emma. Hence, it is true in Emma that Isabella bears a greater resemblance to her father than does Emma, in conformity with our intuition. Inconsistent fictions do not present a problem either. No special provisions are necessary, and Currie’s theory yields the right results for stories that contain unintentional contradictions. For example, in Alice in Wonderland, the textual evidence indicates that the fictional

TRUTH AND INFERENCE IN FICTION

283

author commits an error when he states that there is nothing but tea on the table. Therefore, it is reasonable to infer that the fictional author believes there is a milk-jug on the table at the mad tea party, and it is thus true in the story that there is a milk-jug on the table and it is false that there is not a milk-jug on the table. These cases represent only a small sampling of the vast array of cases in which Currie’s proposal conforms to our intuitions. There are, however, several problems with Currie’s proposal. These problems are of two sorts: there are, on the one hand, problems stemming from Currie’s appeal to a fictional author, and there are, on the other hand, problems which arise because of Currie’s appeal to the beliefs of this fictional author. One difficulty of the latter sort is a problem noticed by Currie himself.17 Imagine a novel in which reference is made to a weapon whose single edge is both sharp and dull. Suppose that in some later passage in the novel, the hero, wielding this weapon, accidentally bludgeons his long lost brother, but does not kill him since the weapon is so dull. With the same stroke that injures his brother, the hero dispatches the villain thanks only to the sharpness of his weapon. Here we have a story in which a single object both has and lacks a certain property, and the object’s possession of these contradictory properties is crucial to the story. More importantly, this contradiction is not inadvertent. Thus, this is a story which contains a proposition of the form P&P, and hence it is true (in the story) that P&P. On Currie’s view, it is true in this story that P&P just in case it is reasonable to infer that the fictional author of this story believes P&P. The problem, however, is that it seems as though it is virtually never reasonable to infer that a fictional author believes a proposition of the form P&P. If that is so, then, there will be stories which clearly include propositions of the form P&P, but for which, on Currie’s proposal, P&P is not true in the story. It certainly seems wrong to rule out the possibility of statements of this form being true in a story tout court. Such an exclusion violates what has been called the Principle of Poetic License, the principle that, by performing the right sort of act, an author can make any proposition true in his story.18 The Principle of Poetic License is very intuitive; clearly someone could write a rudimentary story incorporating the weapon with contradictory properties as outlined

284

JOHN F. PHILLIPS

above and, so long as the writer had a certain minimal competence in English, it would be true in this story that the hero wields a weapon which is both keen-edged and blunt. On Currie’s view, creating such a story turns out to be a fairly daunting task. A guiding principle in our dealings with other language-users is what might be called a presumption of rationality. We assume a person with whom we are communicating is rational until such time as that person proves otherwise. Thus, if we encountered someone who stated that he believes a proposition of the form P&P, our reaction would be to suppose that he does not really believe the proposition in question, but rather is playing a joke on us, speaking non-literally, or merely confused about what he is saying. Clearly, it would take serious effort on the speaker’s part, or extended observation on our part, to convince us that the speaker really believes a proposition of the form P&P.19 So, in most cases involving intentional contradictions, Currie’s theory fails to yield the right results. A second problem for Currie’s view is a difficulty previously raised for Lewis’ proposal. Currie, like Lewis, thinks a belief that is prevalent in the community of origin of a fiction, in the absence of any explicit denial of this belief in the text, must be true in the fiction. So Currie, following in the footsteps of Lewis, imports irrelevant propositions into fiction. For instance, Currie would have it that it is false in the Holmes stories that Pluto exists undiscovered.20 But the existence or non-existence of Pluto is simply not relevant to the Holmes stories; it is neither true nor false in the stories that Pluto exists undiscovered. A curious feature of Currie’s view is that, if the Holmes stories were written today, it would be true in the stories that Pluto exists undiscovered, since the belief that Pluto exists is common in modern England. Thus, in The Seven Percent Solution, it is true that Pluto exists undiscovered, whereas in the Holmes stories of Conan Doyle it is false that Pluto exists undiscovered. There is something decidedly odd, however, about a theory which requires the author of a modern Holmes story to mention explicitly that there are only eight planets in the solar system in order to make his story consistent with the Holmes stories of Conan Doyle. Several problems arise in connection with Currie’s notion of a fictional author. The fundamental problem with the fictional author

TRUTH AND INFERENCE IN FICTION

285

is that he has all the marks of a deus ex machina. In other words, a natural question to ask is, why must a fictional author be introduced into the story? After all, we have a real author. Why must we posit a fictional one as well? One possible response to this question is to point out that many stories clearly possess a fictional author; Lolita has Humbert Humbert, Moby Dick has Ishmael, The Adventures of Huckleberry Finn has Huck, and so forth. Each of these novels are ostensibly written by characters in the novels themselves, and one might label these characters the ‘fictional authors’ of their respective stories. This approach turns out to be fruitless. Consider the case of Huck Finn. On their journey down the Mississippi, Huck and Jim become entangled with two men who accompany them on the raft for part of the voyage. One of these men claims that he is, in fact, the rightful King of France, an assertion that Huck plainly believes. Thus, if Huck is the fictional author of the story in Currie’s sense of ‘fictional author’, then it is true in the novel that Huck and Jim spend quite a bit of time in the company of the rightful King of France. This, however, is undoubtedly wrong. The man claiming to be the rightful King of France is clearly an old fraud; Huck is simply too naive to recognize the old man for what he actually is. It is the immediate realization that the old man is a pathetically obvious imposter that lends the whole episode its humor. Anyone who failed to pick up on this could hardly be said to have understood the novel. Currie’s proposal avoids this difficulty because, on his view, the fictional author of the story is not Huck, but rather a nameless character who must not only have knowledge of all of Huck’s thoughts and deeds, but also must, given Huck’s youthful naivete, have a great deal of knowledge that is not possessed by Huck himself. Currie’s notion of a fictional author is, in my view, rather obscure. In the story To Build a Fire, there are, on the face of it, only two characters, a man and a dog. The story is not told from the point of view of the dog, and the man dies toward the end of the story. The fictional author, then, must be some mysterious third character in the story, a character who has full access to the thoughts of the man and the dog as well as complete knowledge of the events of the story. The fictional author thus appears to be in the same epistemic position with respect to the story as the real author. Indeed, the only discernible

286

JOHN F. PHILLIPS

difference between the author and his fictional counterpart is that the author is a real person whereas the fictional author is a character in the story. Given the fact that we already have a real author, one might suspect that positing a fictional author is unnecessary. There is virtue in economy, and it would be better, it seems to me, if we had a theory which could mimic the successes of Currie’s theory without positing the existence of the fictional author in the story. Another problem with Currie’s notion of a fictional author is that there could be stories that end with the destruction of everything in the universe or a story that ends with the words ‘and so none were left to tell the tale’. Stories of this sort seem to foist some rather bizarre beliefs upon the fictional author. In the first case, he must believe that he is dead, while in the second he apparently must believe that he does not exist. Moreover, it seems to be a consequence of Currie’s view that it is impossible to actually write a story in which no one is left to tell the tale, or in which all life in the universe is destroyed. For, in every story, there must be a fictional author who relates to us the events of the story. But if the fictional author tells us the tale, then there is someone left to tell the tale and there is something in the universe which has not been destroyed. Currie seems to have two options here. He could give special treatment to stories of this kind by claiming that in these cases the fictional author ought to be understood as making predictions about the future, rather than telling us of events that have already occurred.21 Alternatively, he could simply ignore cases of this kind while acknowledging that his theory does not apply to them. Neither of these options is particularly attractive; the first is ad hoc, while the second excludes cases any theory ought to be able to handle. These problems indicate that, although Currie’s theory may be a step in the right direction, an intuitively satisfactory theory continues to elude us. In the next section, I shall present my own proposal and show how it resolves all of the difficult cases which confront the theories of both Lewis and Currie.

V

The most problematic features of Currie’s theory are his positing a fictional author and his appeal to the beliefs of the fictional author

TRUTH AND INFERENCE IN FICTION

287

in order to determine what is true in a given fiction. A moral drawn from our earlier discussion was that what is true in a story is neither solely what the author intended to be true in the story, nor only what is literally true in the story. I began this paper by describing a puzzle, the puzzle being how it is that truth-conditions are to be specified for statements about fiction which are not explicitly set forth in the story. The obvious, albeit rather uninformative, solution to this puzzle is to claim that statements of this kind are true just in case it is reasonable to infer that they are true, on the basis of what is literally true in the story. The idea is that what is true in fiction is what is literally true in the story plus what can be reasonably inferred from the propositions that make up the story on the Literal View. This intuitive picture, along with the aforementioned problems with Currie’s view and the moral of our discussion of the Literal View and the Authorial View, motivates the following proposal: A sentence of the form ‘In the fiction F, ’ is true if and only if, it is reasonable for an informed reader to infer from the text that, under ideal conditions, the author of F would agree that  is a part of F. Note that what I mean by ‘informed reader’ is a reader who is aware of the community of origin of the text, who is familiar with the content and structure of the text, and who is a competent speaker of the language in which the text is written. Recall that conditions are ideal if the author is cooperative and clear-headed, and able to specify each proposition which is a part of the story he has imagined. There are several noteworthy features of my account which are worth discussing before considering how it handles the aforementioned difficulties facing the theories of Lewis and Currie. First, it should be noted that communal belief plays a role in my theory only insofar as it is the backdrop required for understanding the text. Evidence in the text combined with the beliefs of the community of origin might make it reasonable to infer that the author would decree some proposition to be part of the fiction, under ideal conditions, but communal belief alone never suffices to establish this. A second point worth mentioning is that my account does not settle all bets – there is still room for disagreement over what is true in a story. In particular, it seems that there could be disagreement over

288

JOHN F. PHILLIPS

whether there is enough textual evidence to warrant some inference regarding what propositions an author would agree were part of the story, under ideal conditions. It seems to me that reasonable people can disagree over whether, say, it is true in The Three Musketeers that d’Artagnan is one of the finest swordsmen in France. There is some textual evidence in its favor, yet someone might claim that d’Artagnan’s success against various opponents some of whom, we are told, are among France’s finest, may not be enough evidence to reasonably infer that Dumas intended it to be part of the fiction that d’Artagnan is one of the finest swordsmen in France. For my part, I am inclined to think the evidence is sufficient, but others might reasonably disagree. A theory which provided clear-cut answers to every question, however, would gloss over what is, on the face of it, an important characteristic of fiction; namely, that there are at least some borderline cases in which it is unclear whether a proposition is part of a given story. Let us now consider how my proposal handles each of the problems which afflict the theories of Lewis and Currie. An objection raised against both Lewis and Currie was that their theories counted certain propositions as true in stories, propositions which are properly construed as being indeterminate, that is, neither the proposition nor its negation is part of the story. My proposal provides the right results in these cases. For example, in the Holmes stories, it is neither true nor false that Pluto exists undiscovered, since it is not reasonable to infer, based upon the textual evidence, that Conan Doyle intended this proposition or its negation to be a part of the stories.22 Similarly, since the text provides no evidence to indicate that Conan Doyle intended it to be part of the story that Queen Victoria is or is not a prude, it is neither true nor false in the Holmes stories that the Queen is a prude. Cases like those drawn from The Killers and Emma, are also dealt with in appropriate fashion. Given the abundance of textual evidence, it is clearly reasonable to infer that Austen intended the proposition ‘Isabella takes after Mr. Woodhouse more than Emma’ to be part of the story. Therefore, it is true in Emma that Isabella is more like her father than Emma is. Similarly, it is reasonable to infer that Hemingway intended that the proposition ‘Al and Max are

TRUTH AND INFERENCE IN FICTION

289

mob hit men’ be part of The Killers, and thus, in agreement with our intuition, it is true in the story that the killers are mob hit men. Inconsistent fictions are also handled satisfactorily by my proposal. Carroll’s oversight in Alice in Wonderland is easily spotted by an alert reader. It is quite plausible to infer from the text that Carroll intended it to be part of the story that there is a milk-jug on the table at the mad tea party. Moreover, it is clear that he did not intend it to be part of the story that there is no milk-jug on the table. Thus, it is true in the story that there is a milk-jug on the table, but it is false in the story that there is no milk-jug on the table. The problems raised involving stories which end with the destruction of everything in the universe, or with ‘And so no one was left to tell the tale’ are resolved by banishing the fictional author. Furthermore, my proposal is consistent with the Principle of Poetic License. If the text contains an irremediable contradiction, say a single sentence which asserts both P and P where the truth of each proposition is in some way crucial to the story, then it seems reasonable to infer that the author intended P&P to be part of the story. There is no reason why an author cannot intend any proposition, even one of the form P&P, to be part of a story and thus, at least in principle, any proposition can be true in a story, provided the author performs the right sort of action. Another difficult case, one presented by Lewis as a problem for both of his analyses, is also handled by my proposal. Lewis’ example involves Brecht’s Threepenny Opera in which each of the principal characters is treacherous and continuously engaged in betraying his fellows. A minor character in the story is the street-singer who, in fact, never actually betrays anyone. Lewis claims that there is, nonetheless, good reason to believe that the street-singer is treacherous; everyone in the story turns out to be treacherous when put to the test, and there is no reason to think that the street-singer is an exception to this rule. Given Lewis’ reasoning, it is reasonable to infer that Brecht intended it to be true in the story that the streetsinger is treacherous, since the textual evidence suggests that every character in the story is treacherous. Thus, on my view, it is true in the Threepenny Opera that the street-singer is a treacherous fellow. The proposal I have made, we have now seen, is able to overcome the difficulties presented for the views of Lewis and Currie. My

290

JOHN F. PHILLIPS

proposal also preserves many, if not all, of the virtues of the theories of Lewis and Currie, and it provides intuitive results in a broad spectrum of cases. I am unaware of any particular cases in which intuition and my proposal diverge. Nevertheless, a few comments on two potential objections to my proposal would seem to be in order. One objection which might be raised at this point is that, unlike the analysis offered by Lewis, which essentially involves the reduction of the intensional notion “In the fiction F : : : ” to extensional notions by treating “In the fiction F : : : ” as a restricted quantifier over possible worlds, both Currie’s proposal and my own involve trading in one intensional notion for another such notion which is, at best, no clearer or better understood than the notion “In the fiction F : : : ” itself. In other words, it might seem reasonable to claim that our understanding of when “In fiction F, ” is true is greater than or equal to our understanding of what it would be reasonable for an informed reader to infer the fictional author of F believes or what it would be reasonable for an informed reader to infer the author intended. This claim, in my view, is mistaken. But even if this claim is granted, I do not believe the objection under consideration is decisive. The problem here is that the advantages conferred by Lewis’ extensional reduction are offset by its drawbacks. Foremost among these is the cost incurred by introducing the apparatus consisting of sets of possible worlds and their similarity relations. The shortcomings of this analysis are that it seems to bear very little relation to what readers of fiction actually do when they determine what is true in a given fiction and it does not serve to explain the notion of some proposition  being true in a fiction F in terms of a more familiar and well understood notion. On the contrary, my understanding of “In fiction F, ” is true is surely as good, if not better, than my understanding of when, for any collective belief world w of the community of origin of a fiction F, there is some world where F is told as known fact and  is true which is more similar to w than any world where F is told as known fact and  is false. Thus, it seems that, with respect to the objection at hand, both Currie’s proposal and my own are, at worst, on equal footing with Lewis’ approach. As I have argued above, there are many problem cases for Lewis’ theory all of which are handled easily by my account. Therefore, it seems

TRUTH AND INFERENCE IN FICTION

291

that, in comparison with the possible worlds approach of Lewis, my proposal has the upper hand. A second objection might be raised in connection with an important aspect of my proposal. One notable feature of my account is that, in contrast with the more permissive theories of Lewis and Currie, my theory is much more restrictive; that is, many propositions which, according to both Lewis and Currie, would be true in a particular story, turn out to be indeterminate on my proposal. There may be cases, then, in which it seems intuitive that some proposition is part of a story, but where, on my account, the proposition is not part of the story. In my view, however, the prospects for such cases appear rather dim. If a proposition is not explicitly set forth in the text and there is insufficient textual evidence for its being intended by the author to be part of the story, it is hard to see what basis there is for claiming that this proposition is part of the story. Even if there are such cases, it still seems that my proposal has the advantage over the proposals of both Lewis and Currie, as I have endeavored to show.23

NOTES 1

Throughout this paper, the locution ‘statements about fiction’ is intended to specify a particular class of statements about fiction, statements of the form ‘In the fiction F, ’ or statements which ought to be understood as having that form. There are a variety of statements about fiction and fictional characters that cannot be handled by any of the theories to be examined in this paper. The following sentences, for example, require separate treatment: ‘Richard III is my favorite Shakespearean villain’, ‘Cyrano is a better swordsman than d’Artagnan’, ‘King Arthur would be a better ruler than the current Prime Minister’, ‘Dorian Gray is a fictional character’, ‘Heart of Darkness is a postcolonial parable’, ‘Sauron is the personification of evil’. 2 It is perhaps worth noting that the success of this example depends upon our holding fixed the linguistic context within which these sentences are embedded. 3 A cluster of subtle and thorny problems are being passed over in silence. For example, some might claim that sentences such as ‘Jack Crabb got the drop on Wild Bill Hickock’ fail to express a proposition. This seems to have been the view of Gareth Evans in (1982). I take the position that these sentences do express propositions (i.e. they are intelligible) although I remain neutral as to whether these are propositions in the full-blooded sense, or whether they are propositions in an anemic sense (i.e. they are neither true nor false) due to the presence of non-referring terms (e.g. ‘Jack Crabb’ in the above example). My neutrality on this question and my assumption that stories are collections of propositions has no substantive effect upon the analysis carried out in this paper. Neither of these positions has any bearing on the conclusions I shall reach, and I thus refrain from

292

JOHN F. PHILLIPS

investigating the issues underlying these stances. A detailed treatment of some of the issues that arise here, and a discussion of Evans’ view, is given in Currie (1988). 4 Determining what propositions are explicitly expressed in a text or series of utterances requires an understanding of what the sentences, whether they be uttered or written down, mean. What a sentence means, in turn, is determined by the syntactic and semantic rules of the language in which the sentence is written or spoken. I refrain here from endorsing any particular theory of language, but it is important to recognize that even what is literally true in a story may be a matter of some controversy. For simplicity, I have assumed throughout that there is unanimity on what is literally or explicitly true in a story. 5 The first example is noted by Martin Gardner in Carroll (1990). The second is reported by John Heintz in Heintz (1979). 6 The notion of ideal conditions might be more workable if we choose a particular occasion for the imaginary “interview” with the author. So we might say that what is true in the story is what the author would have intended to be true in the story immediately upon its completion, if conditions had then been ideal. 7 This example is taken from Currie (1986). 8 See Lewis (1978). 9 Currie’s critique is contained in Currie (1986) and Currie (1990). I should also note that Lewis himself (in Lewis op. cit.) discusses several problem cases for each of his analyses, one of which will be discussed in section V of this paper. 10 I shall follow the practice of presenting all of the theories in this paper as theories about what is true in stories that are written down rather than spoken aloud. It is a simple matter, however, to extract a theory for verbal stories (told on some occasion) from any of the theories considered below. 11 Exactly how strong the argument must be I cannot say; nevertheless, there are many cases, such as the example from Emma, in which we do reason in this way. 12 Unless, unbeknownst to me, rather bizarre beliefs were widespread among folk in Austen’s England. For instance, if it was nearly a universal belief that young matchmakers in relatively well to do families always bear the greatest resemblance to their respective fathers, then, by sheer chance, Lewis’ account could handle this case. 13 Presumably, in order for w to be closer to w than w there must be some proposition true in both w and w which is false in w . Because Emma, Isabella, and their father do not exist in w, there can be no such proposition since, by hypothesis, w is just like w with the exception of certain propositions concerning Emma, Isabella, and Mr. Woodhouse, none of which are true in w. 14 See Lewis (1983). 15 Both of the difficulties I discuss here were set forth in Currie (1986). 16 Currie’s view is presented and discussed in detail in Currie (1990). 17 See Currie (1990). 18 The ‘Principle of Poetic License’ is the appellation employed in Deutsch (1985). 19 There could be a story, then, for which we would have enough evidence to reasonably infer that the fictional author believes a proposition of the form P& P, but such stories present another difficulty for Currie’s proposal; namely, if the fictional author really believes such a proposition, it is no longer clear what sort of evidence is to count as making it reasonable to infer that the fictional author



TRUTH AND INFERENCE IN FICTION

293

believes some proposition. For example, suppose such a story also contained an inadvertent contradiction. If the fictional author believes a proposition of the form P& P, he might have all sorts of outlandish beliefs, and our ordinary presupposition of rationality must be abandoned. In such a case, it may well be more reasonable to infer that the fictional author has additional contradictory beliefs rather than inferring that he has merely “misspoken”. 20 See Currie (1990). 21 This, in fact, is one suggestion Currie makes when he (briefly) considers cases of this kind in Currie (1986). He is not satisfied with this solution and it appears as though he would be willing to admit that a theory which is able to deal with these cases would, at least in one respect, be superior to his own. My theory easily handles these cases, as will become apparent in the next section. 22 For sake of economy, I shall use ‘intended’ instead of ‘would have agreed was part of the fiction, under ideal conditions’. 23 I should like to thank Sheryl Ross for many helpful discussions. Thanks also to Mike Byrd and Ellery Eells for their comments on an earlier draft of this paper.



REFERENCES Carroll, L. (1990): in M. Gardner (ed.), More Annotated Alice, New York: Random House. Currie, G. (1986): ‘Fictional Truth’, Philosophical Studies 50, 195–212. Currie, G. (1988): ‘Fictional Names’, Australasian Journal of Philosophy 66, 471–488. Currie, G. (1990): The Nature of Fiction, New York: Cambridge University Press. Deutsch, H. (1985): ‘Fiction and Fabrication’, Philosophical Studies 47, 201–211. Evans, G. (1982): Varieties of Reference, New York: Oxford University Press. Heintz, J. (1979): ‘Reference and Inference in Fiction’, Poetics 8, 85–99. Lewis, D. (1978): ‘Truth in Fiction’, American Philosophical Quarterly 15, 37–46. Lewis, D. (1983): ‘Postscript to “Truth in Fiction” ’, Philosophical Papers Volume 1, New York: Oxford University Press.

Department of Philosophy University of Wisconsin Madison, WI 53706 U.S.A.