Worlds within worlds? The paradoxes of embedded fiction.
An embedded fiction is a fiction within a fiction. Lewis Carroll's Alice stories, for example, are novels about dreams, Tom Stoppard's The Real Inspector Hound is a play about a play, John Mortimer's Charade is a novel partly about a film, and Ronald Harwood's The Dresser is a play containing some scenes from Shakespeare. Because an embedded fiction is defined as fictional by something that is itself fictional, we are distanced from it, both logically and, sometimes, emotionally. As Kendall Walton puts it:
The couple who 'look at us' from their portrait on the wall of the artist's studio in Velazquez's painting Las Meninas do so less insistently and command less of our attention than the man in the doorway does; the depicted frame separates them from us. Consider a story that ends with the hero waking from a bad dream. The reader, on realizing that it is fictional (in the story) only that it is fictional (in a dream) that monsters were chasing him, not that they really were, heaves a sigh of (fictional) relief.
While watching Michael Frayn's Noises Off, we never lose sight of the fact that what we are watching, for much of the time, is simply a rehearsal of a play (though we may lose sight of the fact that we are watching a play about a rehearsal of a play). On the other hand, we can be more involved in the embedded fiction than in the embedding fiction. Most of the action of The Taming of the Shrew takes place in a play put on for the entertainment of Christopher Sly, a drunk who is duped into thinking he is a lord, but this does not distance us from the central characters. We can, it seems, step into an embedded fiction just as easily as into an unembedded one.
We would expect the relationship between an embedded fiction and the fiction within which it is embedded to be the same as that between fiction and the real world. Just as we cannot literally step into the world of a fiction and intervene in its goings on, so Alice cannot actually step into the dream world of the looking-glass, though fictionally she can. But, notoriously, anything goes in fiction, and there are fictions which break the rules and allow the boundary between embedded and embedding fiction to dissolve. How should we understand these cases? I shall attempt in this paper to define the problems arising from embedded fiction, and to draw some conclusions about our understanding of fictional discourse - in particular, on whether such discourse is illuminatingly represented in terms of possible worlds.
II. THE ISOLATION OF FICTIONAL WORLDS
'How remote are fictional worlds from the real world?' asks Kendall Walton, in the title of a well-known paper. In answering it Walton tries to do justice to two well-entrenched intuitions: one, that we cannot physically intervene in the worlds of fiction; and the other, that authors generate fictional truths. We cannot prevent Bradley Headstone from attempting to murder Eugene Wrayburn, and yet Our Mutual Friend is no more than a human creation. These intuitions pull in different ways, seemingly obliging us to think both that we can, and that we cannot, causally interact with fictional worlds. The theory Walton offers is radically different from that of David Lewis, who offers a 'possible worlds' semantics for discourse about fiction, and yet Walton and Lewis both agree that fictional truths are isolated from actual truths.
Let us look at Lewis's account first. In 'Truth in Fiction' Lewis presents the view that true statements about fiction are true in virtue of the state of affairs in some logically possible world (or worlds). More precisely, it is true that p in the fiction F if and only if p is true in worlds where F is told as known fact. (Lewis later amends this formulation, but this does not matter for our present purposes.) Fictional characters are the possibilia which inhabit these worlds. In other writings Lewis defends a 'modal realism': other possible worlds are as real and concrete as the actual world, but are not spatio-temporally related either to the actual world or to each other. There is no path from one world to another, and consequently there cannot be causal connections between this and other possible worlds. Now we can see how Lewis can reconcile the two well-entrenched intuitions: although authors are causally responsible for the concrete tokens which constitute their fictions, they are not causally responsible for the truth of the propositions which those tokens express. Thus, in deciding to axe a particular character from her novel, the novelist does not thereby alter a fictional world, but simply takes another decision about which world she will describe.
Walton's account is quite different. In 'How Remote are Fictional Worlds from the Real World?' Walton summarizes a position akin to Lewis's, and states that he prefers the following account: statements about fiction are to be taken as elliptical for statements beginning 'In the fiction . . .', or 'It is fictional that . . .', where the operator is analogous to 'It is believed that' or 'It is wished that'. For Walton, a fictional world is not a world in which it is true that p. Rather, the words which occur in a particular novel, play, etc., or the actions in a game of make-believe, are what make it fictional that p. Fictions, and hence their authors, thus causally generate fictional truths. Qua author, I can cause it to be the case that fictionally p, but this entails neither that I cause it to be the case that (truly) p, nor that fictionally I cause it to be the case that p. Only qua character can I fictionally cause it to be the case that p. So authors are still, in a sense, isolated from their characters.
Although this account neatly explains the isolation of fictional worlds from the real world, Walton thinks that it only provides half the story. What is missing is an account of why fictional characters are often the objects of our emotional states, even when we recognize that the characters are fictional. I fear that Bradley Headstone will commit a murder, but I do not fear that fictionally he will commit a murder. To accommodate this, Walton suggests that in asserting some fictional truth, p, we are not (or not always) asserting that it is fictional that p, but rather pretending to assert p. Fictions are games of make-believe, and in such games we may experience fear and anxiety (or at least quasi-fear and quasi-anxiety).
It is useful to distinguish between the activities of discussing a fiction on the one hand and reading, performing or entertaining a fiction on the other. When I make a statement about what happens in a fiction, what I assert falls within the scope of the fiction operator. However, when reading, etc., a fiction, we pretend - on Walton's account - to believe, or assert, a given proposition that is not placed within the fiction operator. Since Lewis's account is concerned only with statements about a fiction, there would appear to be no inconsistency in combining his possible worlds account with Walton's fiction-as-make-believe account. Both Lewis and Walton can agree that a proposition is fictional if it is asserted, or implied, by a fiction. Lewis, however, goes on to analyse fictional truth as a species of factual truth: truth in a possible world. Part of the motivation for this is that Lewis wants to account for the existence of propositions which, though not explicitly stated in a fiction, are nevertheless true in that fiction. We know, for example, that the Archdeacon of Barchester lived nearer to London than Aberdeen and had more than a passing acquaintance with Greek and Hebrew, though was entirely ignorant of the Periodic Table, despite the fact that Trollope did not provide this information in the Barchester chronicles. The challenge taken up by Lewis, and others, is to formulate a procedure for deciding the truth-value of these unstated propositions.
Lewis and Walton agree that fictional worlds are isolated from the real world, but they differ in their interpretations of this isolation. For Lewis it is a consequence of the spatio-temporal isolation between possible worlds, and this goes together with the doctrine that we cannot causally generate fictional truths. For Walton, we can generate fictional truths, but this is not the same thing as intervening in the world of the fiction.
Now let us apply both Lewis's and Walton's treatments to embedded fictions. Suppose a novel to contain a passage like this:
Wearily, I pushed away the half-consumed macaroon and started to think about my central character, Fido. I started to write:
Fido remained untouched by the post-structuralist malaise hovering around the bed-sitter. He had his bone. That was enough.
It is fictional that the author is thinking about Fido, but it is fictional that it is fictional that Fido remained untouched by post-structuralist malaise. In Lewis's terms, there is a possible world in which there is an author expressing, by means of a fiction, propositions which are true in virtue of there being another possible world in which Fido remains untouched by post-structuralist malaise. In Walton's terms, when reading the novel we engage with the first-order game of make-believe in which someone is really eating a macaroon and Fido is only fictionally eating a bone. We can, if we choose, engage in the second-order game in which Fido is really eating a bone, but then the characters of the first-order game disappear (they have to: since Fido is a real being in this second-order game, there can be no such character as the literary creator of Fido). On both accounts, the embedded fiction is just as isolated from the embedding fiction as we are.
III. PATHOLOGICAL FICTIONS
What happens when the barrier between fictional worlds is violated? In The Real Inspector Hound, Birdboot and Moon are two critics watching a performance of what begins as a parody of a Mousetrap-style thriller. As the action proceeds, however, both Birdboot and Moon are drawn into the drama, playing out roles that fuse their lives in the thriller with their lives as critics. Meanwhile, two of the characters in the thriller momentarily step outside to comment critically on the performance. The embedded fiction thus appears to absorb part of its surroundings. Perhaps this is simply a trompe-l'oeil, however. It is true that Birdboot steps onto the stage where the thriller is taking place in order to answer the fictional telephone that is fictionally ringing (which is also a real telephone that is really ringing). It is also true that he addresses the characters of the thriller. Does he thereby, qua Birdboot, become part of the thriller? Arguably not. The illusion that the wall between fiction and embedded fiction has been broken down is easy to create with a play, for the audience can, literally, share a platform with the actors of the play. But this is not at all the same as sharing a platform with the characters of the play. Stepping on stage to prevent Macbeth stabbing Duncan will not save the king. It will, however, wreck the performance.
More problematic than The Real Inspector Hound is an example borrowed from Walton. Flann O'Brien's At Swim-Two-Birds presents us with Dermot Trellis, an author who conceives the idea of writing a story about the consequences of wrong-doing. In the course of writing this story he creates, for the purposes of having her defiled by the base John Furriskey, the ravishingly beautiful Sheila Lamont. So ravishing is Miss Lamont, however, that Trellis ravishes her himself, to the extent of making her pregnant. Here, surely, is a serious case of intervening in fiction, and one we cannot accommodate within Lewis's account. But perhaps we do not need such an outre case to demonstrate the falsity of that account. The following considerations may be sufficient: possible worlds are logically possible worlds, so nothing that involves a contradiction can obtain at a world. But some fictions contain inconsistencies. Therefore the propositions of a fiction cannot be considered to be true in virtue of what obtains in a possible world. (Note that, in contrast to Lewis, Walton's theory does not require of any proposition within the scope of the 'It is fictional that' operator to be consistent.) But in fact merely inconsistent fictions need pose no problem for Lewis's theory. A possible world cannot be completely specified by a finite set of propositions. Now any actual fiction contains only a finite set of propositions, and so specifies, not just one world, but a set of worlds which are similar in certain relevant respects. Let us suppose that the fiction contains both the proposition p and the proposition Not-p. Since the fiction corresponds to a set of worlds, we can simply say that p is true in a sub-set of these worlds, and false in another subset.
In contrast to merely inconsistent fictions, O'Brien's story is seriously problematic because it breaks the rules of embedded fiction: it is a pathological fiction. An author (or fictional author) can appear qua character or qua author in her novel, but not both at the same time. We can easily imagine a novel in which an author writes a novel in which she ravishes another character. The ravishing takes place in the embedded fiction, not the embedding fiction. The author appears qua author in the embedding fiction, qua character in the embedded fiction. But there can be no such resolution of O'Brien's novel. Here the author qua author ravishes the character.
Lewis's fictional worlds are just as isolated from each other as they are from this world. Just as I cannot intervene in another possible world, so I cannot coherently suppose there to be a possible world in which I intervene in another possible world. Hence, on Lewis's account, I cannot construct a coherent fiction in which there are causal links between the embedded fiction and the fiction in which it is embedded. There is a further, formal problem. If any statement about what happens in a fiction occurs, implicitly or explicitly, within the scope of the fictional operator, then the operator should be iterated in reports of what happens in fictions within fictions. What is problematic about At Swim-Two-Birds is that, in reporting what happens in these fictions, there is no consistent way of placing the operators. Is it fictional that Trellis ravishes Sheila, or is it fictional that it is fictional that he ravishes her? The answer must be 'both' - surely a contradiction, both for Lewis and for Walton.
O'Brien's novel generates another paradox. According to Walton, when we engage with a fiction, we participate in a game-of make-believe, in which we imagine what is stated in the fiction really to be happening. If there is an embedded fiction, then our participation in the first-order game requires us to regard the embedded fiction as fictional. How, then, are we to engage with At Swim-Two-Birds? If we imagine Trellis to be real, then Sheila is fictional, and so cannot be ravished by him. If we imagine Sheila to be real, then Trellis does not even exist (since real people are not the creations of fiction-writers, there is no role for Trellis to play), and so cannot do any ravishing. The problem is that we are asked to engage in incompatible games of make-believe. But perhaps this is precisely what makes Walton's account so plausible, for unless something like it were not correct, At Swim-Two-Birds, apparently a counter-example to the account, could not work as it does. Our enjoyment and understanding of this kind of fiction depends upon our grasping the fact that it breaks the rules of 'normal' fiction and so draws our attention to those rules. We may call such fictions 'meta-fictions': part of their theme is fiction itself. Perhaps, then, it is not appropriate to engage in a game of make-believe with them. This is in part what makes games of make-believe disanalogous to games of sport: one who breaks the rules of, say, rugby football, is unlikely to be seen - except perhaps by the academic spectator - as making a witty allusion to the nature of the game.
What this suggests is that there is a discontinuity between what happens before the rules are broken and what happens afterwards. The beginning of At Swim-Two-Birds, before the introduction of Trellis, is strange, but not so strange that we cannot engage with it in a game of make-believe. When we are told that Trellis ravishes Sheila, do we suddenly realize that we are reading a work of meta-fiction and give up our game of make-believe? Do we simply look on it as an amusing exercise which exploits the ambiguities of the relationship between author and character? If so, then of course we would not expect accounts of ordinary fiction to apply. Taking this line would mean that we could retain Lewis's account of truth in fiction for ordinary fictions, by admitting that it is simply a category mistake to invoke the apparatus of possible worlds for meta-fictions.
This suggestion of discontinuity, I submit, is not true to our experience. Or at least, we can find examples of embedded fictions which, although they break the rules, do not force us to give up our game of make-believe. The plight of L. P. Hartley's author, who is pursued in a series of increasingly threatening messages signed 'W. S.' (in the short story of that name) continues to fill us with horror even when, indeed especially when, W.S. turns up in the flesh to reveal himself as one of the characters created by the author, and who has come to exact revenge for being given an irredeemably nasty personality. A knowing snigger hardly seems the appropriate reaction to the story.
Something more, then, must be said about pathological fictions, and how they relate to normal fiction.
IV. SELF-EMBEDDED FICTION AND REFLEXIVE REPRESENTATION
Let us pursue Walton's remark that 'it is fictional that' is to be understood as analogous to 'it is believed that'. The analogy here is quite a close one, as the two are, I suggest, formally very similar. Like 'it is fictional that', 'it is believed that' can be iterated when we want to talk about beliefs about beliefs. In looking at the inferential patterns which characterize the belief operator when it is iterated, we will need to introduce a technical device to represent the fact that beliefs are indexed to particular believers: 'It is [believed.sub.i] that p' is to be read 'It is believed by person i that p'. We can then summarize the logic of the belief operator as follows:
(a) 'It is [believed.sub.i] that it is [believed.sub.k] that p' does not entail "It is [believed.sub.k] that p'
We assume here that i and k are non-identical. This rule captures the fact that we can be mistaken about other people's beliefs. However, setting aside controversial cases of self-deception, we cannot be mistaken about our own beliefs:
(b) 'It is [believed.sub.i] that it is [believed.sub.i] that p' does entail 'It is [believed.sub.i] that p'
So beliefs can be self-embedded: we can have beliefs about our own beliefs, as well as those of others. I suggest that these rules are precisely mirrored by those governing the fiction operator. In these rules 'It is [fictional.sub.i] that' is to be read 'In the fiction i it is the case that':
(c) 'It is fictional that it is [fictional.sub.k] that p' does not entail 'It is [fictional.sub.k] that p'
Part of the action of The Dresser concerns performances of King Lear. Since The Dresser represents King Lear as fictional, King Lear is - at least in this context - an embedded fiction, though unlike most embedded fictions it also has a life of its own as an unembedded fiction. Now let us suppose (what is in fact false) that The Dresser represents Cordelia as surviving her father to be Queen of England. So it is [fictional.sub.The Dresser] that it is [fictional.sub.King Lear] that Cordelia survives Lear. To participate in The Dresser is to make-believe that King Lear has a happy ending. But of course it does not have a happy ending, so it is not [fictional.sub.King Lear] that Cordelia survives Lear. In the case of embedded fictions which do not have a life outside their embedding fictions, in contrast, the embedding fiction cannot misrepresent the embedded fiction. The crucial rule for our present purposes, however, is the counterpart of (b):
(d) 'It is [fictional.sub.i] that it is [fictional.sub.i] that p' does entail 'It is [fictional.sub.i] that p'
Like belief, a fiction can be self-embedded: it can represent its own statements as fictional. In such cases, it involves no contradiction to say, for some state of affairs s in the fiction, both that it is fictional that s and that it is fictional that it is fictional that s. I think this is the correct way to describe At Swim-Two-Birds. When the barrier between the world of Trellis and the world of Sheila dissolves, they become part of the same fiction, but we do not lose sight of the fact that Sheila remains fictional. Since she is fictional, however, so is Trellis. It is [fictional.sub.At Swim-Two-Birds] that it is [fictional.sub.At Swim-Two-Birds] that Trellis ravishes Sheila. Hence it is [fictional.sub.At Swim-Two-Birds] that Trellis ravishes Sheila.
There is a close relationship between a self-embedded fiction and what Walton calls a 'reflexive representation'. Consider a toy telephone. Its (partial) resemblance to a real telephone invites the child playing with it to make-believe that it is a real telephone. It represents itself as a real telephone (though no-one is actually fooled by this). Some novels also exhibit this reflexivity. Jean-Paul Sartre's Nausea begins with an 'Editor's Note' which tells us that what follows was found among Antoine Rocquentin's papers. The novel thus generates the fictional truth that it is a diary. It represents itself as a diary. What I want to suggest is that a self-embedded fiction is a species of reflexive representation: it represents itself as fictional. This may involve some subtlety on the part of the author. Suppose that the main text of a novel was preceded by the words, 'what follows is an entirely fictional account'. We would treat this either as a sincere (though possibly redundant) statement by the author, or as part of a fiction within which the main text is embedded. In neither case would the fiction be one that represented itself as fictional. On the other hand, if a novel were entitled Confessions of a Fictional Character, we might well treat this as a self-embedded fiction because it suggests that the confessions are coloured by the fictional nature of the protagonist, just as we would expect Confessions of a Vicar to be coloured by the clerical nature of the protagonist.
If I am right in treating 'it is fictional that' as structurally identical to 'it is believed that', then we cannot save Lewis's theory by the simple expedient of taking At Swim-Two-Birds and the like as beyond the pale. A fiction that represents itself as fictional is not just a special case which can be ignored by theories of normal fiction. It illustrates a formal property of the fiction operator. To understand 'it is fictional that' is, in part, to understand the rules which govern it, including (d) above. However, there is no consistent possible worlds interpretation of (d). On Lewis's account, 'It is [fictional.sub.i] that i is fictional' is true if and only if, in those worlds where i is told as known fact, i is fictional. But if i is fictional in those worlds, it cannot be told as known fact, for one and the same thing cannot be treated as factual and as fictional at the same time by the same person. Lewis must therefore reject self-embedded fiction as incoherent. The objection to Lewis, then, is that the notion of a self-embedded fiction is formally consistent, and since no consistent possible worlds analysis can be given of such a fiction, such an analysis cannot be correct.
V. PARADOXES: SEMANTIC, SET-THEORETICAL AND EPISTEMIC
I cannot pretend that the rules governing the logic of the fiction operator say everything that needs to be said about O'Brien's bizarre narrative. These cases raise all kinds of issues, but my target so far has been the possible worlds analysis of fiction. Perhaps, however, I have been unfair to that analysis. What if no consistent analysis of fictional truths can be given? After all, some fictions are paradoxical, so any account of them should do justice to their paradoxical nature. Are self-embedded fictions paradoxical?
We can distinguish between different kinds of paradox. One involves the notion of truth: the semantic paradox, of which the Liar is the best known:
This sentence is false
If the above sentence is true then it is what it says of itself, namely false. If it is false, however, then its description of itself as false is correct, and so it is true. Thus it is both true and false. Can we generate a semantic paradox of this kind from a self-embedded fiction? The analogue of the Liar would be this:
This sentence is fictional
This is paradoxical if we take 'fictional' to imply 'not actually true'. If the sentence is indeed fictional, then its description of itself as fictional is correct and so it is actually true, not fictionally so. But if it is true, then it is what it says of itself, namely fictional, and hence not actually true. However, 'fictional' does not formally entail falsehood, even though there is a conversational implication of falsehood, so there is no contradiction and no paradox.
Another kind of paradox is the set-theoretical paradox. Russell's paradox is a well-known example: is the set of all and only those sets that are not members of themselves a member of itself or not? If it is then it contains a member (viz. itself) which contains itself as a member. Since, however, it does not contain any sets that are members of themselves, it does not contain itself as a member. But then it is one of those sets that do not contain themselves as members, and since it is the set of those sets, it does contain itself as a member. Hence it both does, and does not, contain itself as a member.
We can represent an embedded fiction in terms of set theory by treating a fiction as the set of all its constituent propositions. If the fiction embeds another fiction, then there will be a proposition in its set representing all the propositions of some sub-set as fictional. This sub-set constitutes the embedded fiction. It might appear that a self-embedded fiction appears to generate a set-theoretical paradox akin to Russell's, as follows. Does the proposition which represents the propositions of a self-embedded fiction as fictional belong to the set of those propositions or not? If not, then the fiction is not self-embedded, but rather embedded within some other fiction. On the other hand, if the proposition does belong to that set, then it falsities itself, for then the set would contain a proposition which is not fictional, but true. Again, however, the paradox dissolves once one realizes that a proposition can be both fictional and true at the same time. A proposition is fictional if it corresponds to some fiction, true if it corresponds to the facts. There is no contradiction in supposing that a proposition corresponds to both of these at the same time.
Finally, we can define the epistemic paradox - one concerning belief. Moore's paradox is in this category: I can readily concede that there are some truths that I happen not to believe, but I cannot both sincerely assert a proposition and sincerely assert that I do not believe it to be true, 'p, but I do not believe that p' is not an assertion I can intelligibly make, even though the proposition it expresses may be true. Here now is the fictional parallel. Although I can entertain the idea that a given proposition is both fictional and true (I could imagine, for example, that - unbeknownst to Dickens - Our Mutual Friend corresponds to actual fact), what I cannot do is to entertain a proposition as fictional and as true at the same time, for this would involve adopting incompatible perspectives on a proposition. This is why asserting something to be fictional conversationally implies that it is false: the speaker is treating a proposition as fictional rather than as truth.
What I want to suggest is that certain treatments of self-embedded fiction generate an epistemic paradox. On Lewis's account, a proposition is true in a fiction if it is true in those worlds where the fiction is told as known truth. A self-embedded fiction would therefore involve the incoherence of being told as known truth and as fiction. But it is not just Lewis's account which involves an epistemic element. Gregory Currie has also provided an account of truth in fiction which exploits the notion of belief. On his analysis, p is true in fiction F if and only if it is reasonable for the informed reader to infer that the fictional author of F believes that p. The fictional author is supposed to be a reliable narrator who presents the fiction as truth. (One only needs to add that the fictional author is located in another possible world to obtain a variant of Lewis's account.) The fictional author of a self-embedded fiction would therefore present the narrative both as fiction and as truth. Since this is incoherent, Currie too must reject self-embedded fiction.
I have attempted to argue that certain fictions which break the rules of normal fiction, and which therefore appear to be incoherent, are best seen as self-embedded; that is, as fictions which represent themselves as fictional. I have also tried to show that such fictions are not intrinsically inconsistent. They do, however, pose a problem for certain theories of fiction, in that the attempt to apply those theories to self-embedded fiction generates an epistemic paradox: the fiction is supposed to be sincerely entertained as truth and as fiction at the same time. This paradox is generated both by David Lewis's possible worlds account, and by Gregory Currie's fictional author account. The correct theory of truth in fiction, if it is to include self-embedded fictions, must avoid this epistemic paradox.
The problem still remains of how we are supposed to engage with self-embedded fiction. In so far as playing a game of make-believe is pretending that certain fictional propositions are true, are we not required to entertain a self-embedded fiction both as fiction and as truth? Since this would be to require the impossible, it seems that, after all, we cannot engage with such fiction. At Swim-Two-Birds eludes us yet again. Or does it? We can pretend to believe - to make-believe - something that we could not, coherently, actually believe. It has been suggested that we cannot coherently believe ourselves to be nothing more than brains kept alive in a laboratory and stimulated by an (of course) insane scientist in such a way that it seems to us that we occupy bodies and can interact with the world. But what is to prevent us make-believing that we are brains in a vat?
This sketch of an answer hardly amounts to a convincing account of how we can engage with self-embedded fiction. Such an account would take us beyond the scope of this paper, which has tried to show that, whatever fictions within fictions are, they are not worlds within worlds.
1 Kendall L. Walton, Mimesis as Make-Believe: On the Foundations of the Representational Arts (Cambridge, Massachusetts: Harvard U. P., 1990), p. 284. Las Meninas does not, in fact, contain an embedded fiction. The couple in the painting (Philip IV of Spain, and his second wife) do not look at us from a portrait, but from their reflection in a looking-glass. Velazquez himself appears in the main picture, and the reflection tells us whom he is in the process of painting.
2 Walton, 'How Remote are Fictional Worlds from the Real World?', Journal of Aesthetics and Art Criticism 37 (1978), pp. 11-23.
3 David Lewis, 'Truth in Fiction', American Philosophical Quarterly 15 (1978), pp. 37-46.
4 See especially his On the Plurality of Worlds (Oxford: Basil Blackwell, 1986).
5 This part of the theory is developed further in Walton, 'Fearing Fictions', Journal of Philosophy LXXV (1978), pp. 5-27, and in Mimesis as Make-Believe, ch. 7.
6 Or can there? This issue is taken up in section IV.
7 See Lewis's Postscript B to 'Truth in Fiction', in David Lewis, Philosophical Papers Vol. I (Oxford U. P., 1983). This manoeuvre will not deal with other inconsistencies, such as a time-travel story in which the time-traveller changes the past, but then we need only suppose that there are worlds which approach very closely to the fiction.
8 Mimesis as Make-Believe, pp. 117-21.
9 For a discussion of the Liar and Russell's paradox, see Mark Sainsbury, Paradoxes (Cambridge U. P., 1988).
10 I assume the correspondence theory of truth here for convenience, but it is not crucial to the argument.
11 G. E. Moore, Philosophical Papers (London: George Allen and Unwin, 1959), p. 543.
12 Gregory Currie, The Nature of Fiction (Cambridge U. P., 1990), p. 80.
13 One account which does avoid paradox is Alex Byrne's suggestion, in 'Truth in Fiction: The Story Continued', Australasian Journal of Philosophy 71 (1993), pp. 24-35. Byrne offers the following analysis (I have paraphrased somewhat): it is true in fiction F that p if and only if the (ideal) reader could infer that the (re-constructed) author is inviting the reader to make-believe that p. Here no-one is credited with the capacity to entertain F as fiction and as truth simultaneously.
14 By Hilary Putnam, in Reason, Truth and History (Cambridge U. P., 1981), ch. 1.
15 Earlier versions of this paper were presented at seminars at Leeds and Essex. I am very grateful to those present for their comments.
Robin Le Poidevin, Department of Philosophy, University of Leeds, Leeds LS2 9JT, England.