HARRY DEUTSCH
FICTION AND FABRICATION
(Received 1 May, 1984) Various philosophers have recently set about developing general theories of objects somewhat akin to Meinong's. 1 Each of these theories is based on some version of the following principle of Object Abstraction: (P1) For any class K o f properties, there is an object possessing exactly the properties in K. 2 In light of Russell's famous attack on Meinong and for other reasons as well, (P1) appears in these new object theories in a qualified form. 3 According to Parsons' version, for example, the properties in K must be "nuclear" properties, while others distinguish in various ways between the literal possession of a property and the fictional possession of a property, and it is fictional possession that figures in Object Abstraction. 4 The motivation for these theories seems to be the development of a general account of objects of thought, but fictional objects, in particular, have been the main source of data for the theories, which in turn have been applied principally to the phenomenon of literary fictions, s Closely related to Parsons' object theory is a kind of corollary to Object Abstraction that applies directly to stories: (P2) For any class K o f propositions, there is a story in which exactly the propositions in K are true (Story Abstraction). Of course many of the stories (P2) postulates may not (ever) actually exist in the sense of being written down or otherwise realized; but (P2) asserts that such unrealized and perhaps even unrealizable stories exist as abstract semantical objects. 6 Despite its platonist character, (P2) seems more ontologically neutral than (P1). For it would at least appear that it is one thing to claim that corresponding to the set of properties {goldenness, mountainhood} there is an object, albeit a non-existent one, that possesses just those properties, and quite another to claim that there is a story (abstractly conceived) in which exactly the propositions 'M is a mountain' and 'M is golden' are true. 7 (P2), moreover, in virtue of its generality, and despite its platonist cast, Philosophical Studies 47 (1985) 201-211. 0031-8116/85.10 (~) 1985 by D. ReidelPublishing Company
202
HARRY
DEUTSCH
seems to express something of what it means to say that stories are made up out of whole cloth (or thin air), and that what they represent are pure fabrications or creations rather than discoveries or observations. For to say that a story is 'made up' is at least in part to say that what is true in it is stipulated to be true in it. But the process of stipulation involves a kind of maximally free choice: A n y proposition can be stipulated to be true in some story. The operative principle here - I call it 'The Principle of Poetic License' - forms a part of the content of (P2): For any class K o f propositions, there is a story (abstractly conceived) in which every proposition in K is true. s Combining (P2) with a complementary principle of story-identity, namely, (P3): Stories are the same if the same propositions are true in them, gives us a nicely rounded little theory of stories. Let us call it 'P' (perhaps for 'Parsons'). 9 P is a plausible theory, more plausible, it seems to me, than Parsons' correlative object theory. 1~ It might be objected that P runs counter to the idea that stories are created, but there are convincing replies to this and similar objections, xl I contend, nevertheless, that (P2) is false and that in fact not every class of propositions determines a story in which exactly those propositions are true. Indeed, I maintain that stories are logically closed: The logical consequences of propositions true in a story must also be true in it, though it is important to see that some modification of the notion of logical consequence is essential in this context. Some philosophers have endorsed the thesis that stories are logically closed (Cf. Lewis, 1978, and Wolterstorff, 1976, for example), but it is incompatible with (P2) and Fine (1982) and Parsons (1980) have raised objections to it. I shall defend the thesis - first by giving a direct argument for it based on the fact that (P2) gives rise to a paradox similar to (though differing from) one that Russell discusses in Appendix B of Principles o f Mathematics. I shall then reply to the objections of Fine and Parsons, and I shall try to indicate how and why the notion of logical consequence must be modified in the context of stories. In outline, the direct argument runs as foltows: P is inconsistent. In order to resolve the inconsistency we must somehow modify (P2). In seeking to do so we must (a) avoid placing undue restrictions on the Principle of Poetic License; for such restrictions threaten the very basis of the (true) claim that stories are 'made up'. And we must (b) guarantee that the modified version of (P2) yields a complete theory in the sense that every story that should exist does exist according to the theory.
FICTION AND FABRICATION
203
Two sorts of possible modifications of (P2) come immediately to mind: (1) we can restrict K to sets of propositions; or (2) we can restrict K to those classes that are given by predicative conditions, i.e., roughly, conditions containing no bound variables ranging over stories. (The derivation of a contradiction from (P2) exploits an impredicative condition.) The former solution, however, violates requirement (a), while the latter violates (b). We are thus left in the lurch, and a natural alternative is to require that K be logically closed. Let F be any fixed property of stories such that if S~ and $2 are distinct stories, then the propositions F(SI) ($1 is F) and F(S2) are distinct; and let 'T(p, S)' abbreviate the statement that the proposition p is true in the story S. Consider the class D = {p : 3 S(p = F(S) & ~ T ( p , S))) of propositions. By (P2), there is a story So in which exactly the propositions in D are true. It then develops that both T(F(So), So) and ~T(FiSo), So) a contradiction. 12 By (P2), the class D of propositions yields a story So in which exactly those propositions are true that are of the form F(S) and that are not true in the stories they are about. It is then paradoxical to suppose either that So says of itself that it has the property F, (T(F(So), So), or that it doesn't say this (~T(F(So), So)). One response to this argument is to deny that a property such as F can exist. This seems futile. Let F be the property of being a story. Thus if S1 and $2 are distinct stories, it seems plain that F(S1) and F(S2) are distinct propositions. 13 (P2), therefore, is false. Fine would suggest that (P2) be modified by, in effect, replacing the term 'class' by 'set', and adopting the completing principle of 'story closure': The propositions true in a story form a set. 14
However, in the first place, Fine's proposal would create an unintuitive gap between our ability to formulate and our ability to fabricate. If we can formulate a condition satisfied by every member of even an overly 'large' collection of propositions, we should thereby be able to make up a story in which all of these propositions are true. Is there in fact anything we can 'think up' that we cannot 'make up'? Secondly, the gap (between formulation and fabrication) would be wider than might at first be supposed. Surely there is no consistent collection o f propositions all of which could not simultaneously be true in some story. For otherwise it would be possible that some collection of true propositions could not be 'foretold in fiction' and then indeed would truth be stranger than fiction! The diagonal class D
204
HARRY
DEUTSCH
of propositions, however, might easily be consistent, e.g., if F is the innocuous property of being a story. Is Fine's proposal entails that D, though perhaps consistent, could not be 'story-consistent' (i.e., no story simultaneously 'satisfies' every member of D), and this seems clearly wrong. At the very least, poetic license allows that what might in fact be true, might be true in fiction. Thus, Fine's solution is inadequate) 6 Alternatively, we may try restricting K to predicatively defined classes. (Note the impredicativity of the condition defining D.) As with Fine's proposal, we must complete the solution by requiring that the class o f propositions true in a story be predicatively determined. In a sense, this is the right kind of solution in that in restricting the objects that can be stories, no restriction is placed on the classes of propositions that can be true in stories; that is, the Principle of Poetic License is preserved. To see the problem with this, let K be predicatively defined, and let C(K) be the logical closure of K, that is, the class of propositions entailed by propositions in K, where entailment is construed in a standard modal way, so that p entails q, if there is no possible world in which p is true and q is not. Observe that the condition p E C(K) is not predicative in the required sense, since it (tacitly) contains bound variables ranging over possible worlds, and for present purposes we may regard possible worlds as consistent and complete stories. The predicative solution' thus provides no guarantee that there is any story corresponding to C(K), and hence is inadequate in that it gives rise to an incomplete theory) 7 If we accept the principle of Poetic License (in the name of stories as fabrications), or at least if we accept that any consistent class of propositions is story-consistent (and that seems undeniable), I see no natural alternative except to conclude that stories are, in some sense, logically closed. 18 It follows, perhaps, that much may be true in a story that its author did not explicitly intend or that would not be perceived to be true in it by (to borrow a phrase of Parsons') "a normal attentive reader". It follows also that there can be no genre of "inert literature" such as Fine envisionswherein truth in a story belonging to this genre is completely circumscribed by the explicit dictates of authors. 19 Given the foregoing argument, we may say that the "logic" of fabrication precludes this possibility. 2~ Parsons' statement that "whether or not something is true in a story ought to accord with what a normal, attentive reader understands to be true in the story" is hard to reconcile with the facts. For example, a story may require that its
FICTION AND FABRICATION
205
readers perform difficult deductions or calculations in order to determine what is true in it. Parsons raises two further objections. He argues first that as we read through a story, material that previously formed a part of our understanding of the story is sometimes subsequently omitted. He views this as an obstacle to the formation of principles of inference "to shed light on textual interpretation".21 It is hard to see how this poses a problem for the view that stories are logically closed. The view is that once we have formed the core "account" o f a story, perhaps by the process Parsons describes, we add the logical consequences of propositions in the core account. It makes no difference, then, that propositions which appeared in an incomplete or partial version of the core account do not appear in the final version. Parsons' second objection involves imagining a lengthy novel in which at one point we are told that Molly is a McCoy. Buried elsewhere in remote reaches of the novel are propositions that formally entail that Molly is not a McCoy. Now, that Molly is not a McCoy shouldn't be true in the novel. Parsons concludes that the novel is not logically c l o s e d - even under so discriminating a canon of inference as the Anderson-Belnap system EQ (which suffices for the derivation Of the contradiction). But upon finding that there are propositions true in the novel that entail that Molly is not a McCoy, a natural reaction would be to suppose that a mistake has been made, and that, if possible, the novel should be revised. The implication is that the proposition that Molly is not a McCoy is true in the novel (for otherwise why should it require revision?) and hence that the novel is logically closed - at least to the extent required to derive that proposition. Less readily dismissed are certain objections - not to the thesis that stories are logically closed - but to the claim that stories are closed straightforwardly under ordinary ('classical') logical consequence. There are basically two interconnected objections. First, as Parsons puts it, logical closure "adds all sorts of irrelevant material to the story". Thus every logical truth on any topic would be true in every story; and, dually, anything and everything would be true in any inconsistent story. 22 Secondly, within the possibleworlds semantics appropriate to ordinary entailment, it is far from clear what it means to say that A entails B when A (or B) contains terms that mention fictional objects. It is unclear, for example, what it would mean to say that every possible world in which Hamlet kills Polonius is a world in which
206
HARRY
DEUTSCH
Hamlet kills someone; for it is unclear what it would mean to say of a possible world W that Hamlet kills Polonius in W. Among the myriad possibilities, which is Hamlet and which is Polonius? It seems to me that these objections have considerable force. The first, however, has suffered from the lack of a clearly stated rationale. 23 There are, for example, many stories about what could not be. We do not need to resort to examples such as 'logical fantasies' about round squares and non-selfidentical things. Apart from time trivial inconsistencies, there are all those stories about people turning to stone or into nightingales, and stories about what happens down the rabbit hole, or in Lilliput or Never-Never land. Utter impossibility is a staple of fiction. It does not seem right to say that everything is true in every such story. Yet perhaps David Lewis is right when he remarks that we should not expect to have a well-behaved concept of truth for inconsistent stories, z4 To properly reply to Lewis we need a reason to deny that in an inconsistent story, anything whatsoever might as well be true. The reason, I suggest, is that there is no reason n o t to deny this. What reason could there be? It is not that a proposition and its negation could not be true in the same story; and it is not that there is necessarily some connection in meaning at work in the inference from a pair of contradictory propositions A and not-A to an arbitrary proposition B (as there is, perhaps, in the inference from the conjunction (A and B) to A). That leaves the fact that a proposition and its negation cannot be jointly true. But this is relevant only if truth in fiction is a species of modality subject to an analysis within the framework of possible worlds semantics. Lewis has attempted such an analysis (1978). However, Lewis' proposals are seriously flawed. He tells us, for example, that fictional names such as 'Holmes' and 'Watson' are non-rigid designators that refer in a given possible world W to those individuals (if any) who in W have the properties attributed to Holmes and Watson in the Conan Doyle stories. This ignores the fact that the ordinary contingent properties of things are represented as such in fiction. In the stories, Holmes and Watson meet by chance. They need not have met, and that they need not have met is something that is true in the stories. Yet on Lewis' account, there are no possible worlds in which Holmes and Watson exist and fail to meet. Accordingly, on Lewis' account it is not true in the stories that Holmes and Watson need not have met, but that is contrary to fact. 2s The Parsons-ish theory P and Lewis' possible worlds approach are two extremes of theory, as regards the question of logical closure. What is needed
F I C T I O N AND F A B R I C A T I O N
207
are closure principles that lie s o m e w h e r e b e t w e e n these two extremes, principles w h i c h respect the outrageously fantastic 'possibilities' o f pure fabrication, and w h i c h are sensitive to the fact that truth-in-a-story, unlike t r u t h simpliciter, is 'topic-sensitive'. The propositioas true in a story are propositions a b o u t a circumscribed range o f topics. The same is true o f a formalized t h e o r y , b u t stories are w r i t t e n in natural languages, and in a natural language, a n y o n e can babble a b o u t anything. So we m u s t insure that topics that are 'alien' to a given story are n o t i m p o r t e d into it in virtue o f logical c o n n e c t i o n s o f the ' w r o n g ' sort. It m a y be a necessary t r u t h that rr is transcendental, but even so, it is no part o f the tale o f Hansel and Gretel. As I have tried to show elsewhere, 26 it is possible to satisfy these d e m a n d s by means o f an appropriately novel f o r m o f entailment. It remains to w o r k o u t an a d e q u a t e t h e o r y o f reference and predication in fiction. In part for reasons suggested by the a r g u m e n t I have tried to develop here, I do n o t t h i n k that Parsons' object t h e o r y , or any other e x t a n t object theory, is correct. We c a n n o t o b t a i n objects
at least n o t fictional objects -
by abstraction f r o m arbitrary classes o f properties, but only at best f r o m suitably closed classes o f properties; and even t h e n there remain serious p r o b l e m s o f i d e n t i t y and individuation. 27
NOTES i For references, see Fine (1982 and 1983). 2 In P1, the term 'class' does not mean set, but rather condition. The class of all properties, i.e., the class of properties P satisfying the condition P = P is not a set. (But see note 14 below.) 3 See Parsons (1980) for references and a lucid discussion of Russell's attack. 4 Fine (1983) contains a brief discussion of the differences between these two approaches. s See, for example, Chapters 3 and 7 of Parsons (1980). It is worth mentioning that certain idealizations, e.g., particles and point masses, are sometimes said to be fictions, but these are hardly fabrications and can be interpreted mathematically. People often confuse idealizations with fictions, perhaps because phrases such as 'useful fiction' encourage them to do so (Cf. Parsons, 1983). I gather from the brief remarks on p. 232 (Parsons, 1980) that Parsons does not wish to apply his object theory to such idealizations, but only to genuine fabrications. 6 Alternatively, (P2) can be interpreted as postulating the possible existence of a story corresponding to any class of propositions. (P1) cannot be interpreted similarly since many of the objects (P1) postulates could not possibly exist. 7 The difference is obscured somewhat in the contextual version of object theory developed in Fine (1982). But it can be said that a theory based on (P2) is neither committed to - nor capable of accounting for - de re assertions about non-existent objects.
208
HARRY
DEUTSCH
8 There is an e x t e n d e d discussion in 'Making up stories' of the i m p o r t a n t interrelations between the concepts o f stipulation, fabrication, and creation in fiction. (See Note 10 below.) 9 I do n o t m e a n to suggest that Parsons would or m u s t endorse the theory P. Yet there are certainly echoes o f P in Parsons (1980). Some m a y find (P3) objectionable on the grounds that radically stylistically different texts w o u l d express distinct stories even if the same propositions are true in each. However, we do allow t h a t the same story can be told in verse and in prose; a n d so there is a sense of 'same story' that transcends even such m a r k e d stylistic differences as the difference between verse and prose. L0 Parsons holds that there is an object that quite literally has the properties o f being a m o u n t a i n , being golden, and no other property. (See Fine (1983) for some criticisms of this literalist doctrine.) The m o s t that P is c o m m i t t e d to is the existence of a story in which exactly the proposition that there is a golden m o u n t a i n is true. 11 Briefly, m y view (developed at length in 'Making up stories') is t h a t to create a character is to fabricate ('make up') a story a b o u t it. Similarly to create a story is to m a k e it up " o u t o f whole cloth". This involves a process o f stipulation licensed b y the Principle o f Poetic License. T h u s to say that stories or their characters are created is n o t to m a k e an ontological assertion a b o u t their " c o m i n g into being"; it is, rather, to m a k e an essentially logical assertion (in a broad sense (P2) is a logical principle) to the effect that w h a t is true in a story, or true of a character, is so because it is stipulated to be so, and for no other reason. In these terms, I a t t e m p t to c o u n t e r the anti-platonist, creationist objections to (P2), (P3), and the like, raised by Fine (1982 and 1983) a n d by Levinson (1980). In particular, one w h o adheres to the platonist interpretation o f (P2) is n o t c o m m i t t e d to absurdities such as that C o n a n Doyle h a d s o m e h o w to discover that Sherlock Holmes is a detective. If, corresponding to any class of properties, there is a story in which some (native) object has those properties, then an author w h o writes such a story c a n n o t be said to have discovered that the object has those properties, since no m a t t e r w h a t properties these are, they are the right ones! One w h o merely specifies some condition on natural n u m b e r s c a n n o t be said to have discovered the existence o f a set o f natural n u m b e r s satisfying that condition, since corresponding to any condition on natural n u m b e r s there is a set (possibly e m p t y ) of natural n u m b e r s satisfying t h a t condition. Of course it is often a genuine discovery to find that there is a n o n - e m p t y set o f natural n u m b e r s satisfying some interesting condition. There is no comparable sense in which a u t h o r s discover or 'hit u p o n ' their stories. 12 The p r o o f is as follows. Suppose that T(F(So) , So). T h e n F(So) = F(S') and ~ T ( F ( S o ) , S'), for some story S'. T h u s S O = S', and so ~ T(F(So) , So). If so, T(F(So), So)! Note that I am assuming t h a t appropriate type-theoretical restrictions have already been introduced in the formulation o f P. W i t h o u t such restrictions, a contradiction could be derived by m e a n s o f a simple 'Russell's Paradox' argument. A paradox similar in some respects to the present one was first discovered by R o m a n e Clark (cf. Clark, 1978). Fine's discussion o f such "diagonal difficulties" (Fine, 1982) is very illuminating. As far as I k n o w , no One has heretofore formulated either the principle o f Story Abstraction or the paradoxical a r g u m e n t associated with it. The difficulties Fine discusses have to do with object theory. It seems to m e that as applied to story abstraction rather than object abstraction, the relevant diagonal a r g u m e n t s take a particularly striking and intuitive f o r m . (There are ways other t h a n t h a t gi#en here to derive the paradox.) 13 Parsons would probably claim that F could n o t exist, and I do n o t m e a n to suggest that there is n o t h i n g to debate here. The existence of F is more plausible on a 'structural' view o f propositions t h a n on a possible worlds view. Since the difference between S 1 and $2 m a y be so small as to m a k e the propositions F(SI) and F(S2) virtually indistinguishable, a convincing candidate property F m u s t be either sensitive to small differences between distinct stories or else i n d e p e n d e n t o f any degree o f similarity or dissimilarity between the two. The property o f being a story is the latter sort o f property. T h e
FICTION
AND FABRICATION
209
property of being a story S such that every proposition true in S is true, is a property o f the former kind. 14 I find this proposal s o m e w h a t obscure. The n o t i o n of a set o f properties or propositions is n o t entirely clear. For example, consider the property of being an ordinal n u m ber. Is there a set c o n t a i n i n g this property as an element? After all, the extensional counterpart of this property - the class o f ordinal n u m b e r s - is n o t an e l e m e n t of any set or class. An idea related to Fine's b u t that does n o t involve the n o t i o n o f a set o f properties or propositions would be to replace (172) by the principle t h a t for any condition on propositions, and for any story S, there is a story S' in which exactly those propositions are true that are true in S and satisfy that condition (with appropriate restrictions on the variables free in a formula expressing a condition on propositions). It follows directly from this (via the paradoxical argument) that there is no 'universal story' - a story in which every proposition whatsoever is true; and the proposal is similar in spirit to Fine's t h o u g h t h e two are n o t equivalent. (Cf. Deutsch, 'Making up stories'). Both proposals are subject to the difficulties m e n t i o n e d in the text. is Each proposition in D w o u l d then be true. It is in fact possible to insure that D is consistent. Let us say that a story is consistent if some proposition is n o t true in it. T h e n if F is the property o f being consistent, as so defined, each proposition in D will be true; for if the proposition that S is consistent is n o t true in S, t h e n S is consistent. 16 A further objection to Fine's proposal is that, assuming t h a t the universal class o f propositions is n o t a set, Fine's proposal entails that there is no universal story. But it seems there is; namely, 'This is a story: Every proposition is true. The End.' Regarding this sort o f example Fine c o m m e n t s (Fine 1982, p. 121) that "... from the fact t h a t a universal proposition is true in a story, it does n o t automatically follow that all o f its instances are also true". F r o m the fact t h a t the story says that all m e n are blond, it certainly d o e s n ' t follow that any actual m a n is blond-in-the-story. However, the range of quantifiers such as 'all m e n ' and 'every proposition', as these occur in a story, is determined in part by their intended interpretations. The intended interpretation of 'every proposition', as i t occurs in the foregoing 'story', could well be t h a t it is to m e a n literally every proposition. In fact, as the a u t h o r of the story, I would insist t h a t t h a t is j u s t what I intend the quantifier, as used in m y story, to mean. 17 A way a r o u n d this difficulty would be to restrict K to classes defined in terms o f conditions that involve at m o s t b o u n d e d quantification over stories. This would rule o u t the diagonal class D b u t n o t C(K). D w o u l d then be story-consistent since there are predicative conditions that serve to define the universal class o f propositions. Nonetheless, there would still be no guarantee that there is a consistent story in which every m e m b e r o f D is true, even assuming that D itself is consistent. 18 I do n o t have a p r o o f t h a t the closure solution is adequate, i.e., that a theory based on it would be consistent (other things being equal), t h o u g h I see no reason to d o u b t its adequacy. In fact, consider the abstraction principle obtained from Story A b s t r a c t i o n by restricting the latter to classes o f propositions that are closed under classical entailm e n t . The following a r g u m e n t suggests that this abstraction principle is consistent: T h e family of classically closed classes of propositions has a n o n - e m p t y intersection no e l e m e n t o f which w o u l d find its way into any diagonally p r o d u c e d class defined in terms of this family. F o r the elements of such a diagonally p r o d u c e d class are elements of the c o m p l e m e n t s of the classes for which the diagonal condition is defined. T h u s no logically true proposition would be an element of the diagonally produced class, and hence that class would n o t be logically closed. This is n o t in itself a p r o o f o f consistency t h o u g h I think that a p r o o f incorporating this line of a r g u m e n t is possible. Since I do n o t favor the view that stories are closed u n d e r classical entailment, and since classes o f propositions that are closed in the sense I do favor need n o t overlap, there remains the question of the consistency of the closure solution I advocate. 19 Cf. Fine (1982), p. 117.
210
HARRY
DEUTSCH
20 Is it really plausible to suppose that there can be stories that are n o t closed even u n d e r the simplest f o r m s of inference, e.g., the inference from (A and B) to B? We can imagine stories in w h i c h the characters behave as if this form o f inference fails, or in which it is universally agreed that it fails, b u t this is n o t the same thing. 21 It is unclear w h a t principles Parsons has in mind. Perhaps w h a t he m e a n s is that some of the logical consequences of propositions in the core a c c o u n t will n o t in general be relevant to questions o f interpretation. For example, the truths o f logic - which follow from any proposition in the core a c c o u n t - are n o t in general relevant to the question of w h a t is substantively true in the story, and hence are hardly relevant to the interpretive issues with which critics are concerned. A question then arises as to which principles of logical inference are relevant to w h a t is substantively true in the story. This in itself is a very interesting problem. (An analogous problem arises for theories.) The difficulty Parsons envisions is no obstacle to the project o f isolating the distinctly relevant principles o f inference. 22 If by "irrelevant" Parsons m e a n s 'irrelevant to what is substantively true in a story', and if his position is that the fact o f these irrelevancies provides a reason to deny that stories are logically d o s e d , then he should be prepared to deny, for exactly similar reasons, that theories are logically closed! For a theory will also contain material t h a t is irrelevant to w h a t is substantively true in it. 23 Both Parsons and Wolterstorff (Wolterstorff, 1976) say that we could "live w i t h " having every necessary truth true in every story, b u t neither a t t e m p t s to say w h y any degree of stoicism is called for. What exactly is the problem that we are told we could live with? As I argue in 'Making up stories', once the source o f discomfort is identified, it b e c o m e s less easy to live with. The problem is that truth in fiction is 'topic sensitive' (see p. 9), and it is a problem that c a n n o t be simply ignored. Similarly, Wolterstorff argues that to have everything true in every inconsistent story would trivialize the concept o f t r u t h for inconsistent fiction. But he fails to explain w h y the concept is n o t a trivial one. 2, Cf. Lewis, 1978. 2s The p o i n t is t h a t on Lewis' analysis the sentence 'Sherlock Holmes is n o t a detective' is n o t true in any world. Hence the narrow scope reading of the sentence 'It is possible t h a t Sherlock Holmes is n o t a detective' is also n o t true in any world and so would n o t be true in the C o n a n Doyle stories, if Lewis is right. It is true that the person w h o plays the role o f Sherlock Holmes in a world W m i g h t have some counterpart in a world W' w h o is n o t a detective, b u t this counterpart could n o t be Sherlock Holmes (in W'), and so it is unclear w h e t h e r on this basis we should be entitled to affirm that Sherlock Holmes need n o t have been a detective. Lewis could take the position that, roughly speaking, to say that Holmes need n o t have been a detective is to say that a n y o n e in any appropriate possible world w h o plays the role of Holmes in that world is such that in some (other) possible world that individual is n o t a detective. But w h a t t h e n o f the s t a t e m e n t t h a t Holmes need not have been Holmes? Surely this s t a t e m e n t is not true in the C o n a n Doyle stories; y e t it would seem to come o u t true in t h e m on the foregoing version of Lewis' view. For a n y o n e in any possible world w h o plays the role o f Holmes in that world is such that in some (other) possible world they do not play that role, i.e., in some other world they are n o t Holmes - i n a s m u c h as being Holmes entails being a detective. This m a t t e r is discussed at length in m y paper 'Fiction and modality (in preparation). 26 See D e u t s c h (1983 and 'Making up stories'). 27 Fine (1982 and 1983) contain e x t e n d e d critiques of 'internalist' theories o f nonexistent objects. T h u s some classes of properties will n o t serve to define objects. B u t if I am right, this is so n o t because each m e m b e r o f some class of properties c a n n o t be true o f any object, b u t because the properties in some classes of properties c a n n o t be all that is true of any object.
F I C T I O N AND F A B R I C A T I O N
211
BIBLIOGRAPHY Clark, R.: 1978, 'Not every object of thought has being: A paradox in naive predication theory', No~s 12, Deutsch, H.: 1983, 'Making up stories', typescript. Deutsch, H.: 1983, 'Paraconsistent analytic implication', forthcoming in Journal o f
Philosophical Logic. Fine, K.: 1982, 'The problem of non-existence I', Topoi 1, pp. 97-140. Fine, K.: 1983, 'Critical review of Parsons' Nonexistent Objects', forthcoming in Philo-
sophical Studies. Levinson, J.: 1980, 'What a musical work is', Journal o f Philosophy, 77, pp: 5-28. Lewis, D.: 1978, 'Truth in fiction',American Philosophical Quarterly, 15, pp. 37-46. Parsons, J.D.: 1983, 'Toward a general theory of fiction', Philosophy and Literature 7, pp. 92-94. Parsons, T.: 1980, Nonexistent Objects (Yale University Press, New Haven). Wolterstorff, N.: 1976, 'Worlds of works of art', Journal ofAestheHes andArt Criticism 35, pp. 121-132.
Department of Philosophy, Illinois State University, Normal IL 61 761, U.S.A.