Acta Anal (2014) 29:181–194 DOI 10.1007/s12136-013-0209-1

Sainsbury on Thinking about Fictional Things Anthony Everett

Received: 3 April 2013 / Accepted: 27 August 2013 / Published online: 29 September 2013 # Springer Science+Business Media Dordrecht 2013

Abstract In a number of places Mark Sainsbury has recently developed an attractive irrealist account of fiction and intentionality, on which there are no fictional objects or exotic intentional entities. A central component of his account is an ambitious argument, which aims to establish that the truth of intensional transitives such as “I think about Holmes” and “Alexander feared Zeus” does not require the existence of fictional or intentional objects. It would be good news indeed for the irrealist if Sainsbury’s argument worked. However, I argue that Sainsbury’s argument fails. I conclude by considering how Sainsbury’s irrealist might explain our intuitions about such sentences, drawing upon another component of Sainsbury’s irrealism. Keywords Fictional Objects . Sainsbury . Intensional Transitives . Fictional Irrealism

Mark Sainsbury has recently offered a powerful series of arguments against fictional realism and developed what he calls an irrealist account of fiction and intentionality, an account on which there are no fictional objects or exotic intentional objects such as mythical and imaginary beings.1 Sainsbury develops his irrealist account of fiction in response to a series of familiar arguments offered by realists for accepting fictional objects into our ontology. These arguments begin by noting that there are a wide range of intuitively true statements, which purport to refer to, or quantify over, fictional objects, such as the following: (1) Holmes is a fictional character. (2) Anna Karenina was more intelligent than Emma Bovary. (3) Holmes is smarter than any real detective.

1 See in particular (Sainsbury 2009) and (Sainsbury 2010b) but also (Sainsbury 2010a) and section 6.4 of (Sainsbury 2005). In what follows I will concentrate upon Sainsbury’s irrealist account of fiction, but obviously what I have to say will equally carry over to his irrealist account of intentionality in general.

A. Everett (*) Philosophy Department, University of Bristol, Cotham House Cotham Hill, Bristol BS66JL, UK e-mail: [email protected]

182

(4) (5) (6) (7) (8)

A. Everett

Conan Doyle thought about Holmes. Many people admire Holmes. Holmes was created by Conan Doyle. Holmes is presented in more detail than Watson. Holmes is famous.

The realist then argues that the intuitive truth of these statements cannot be adequately explained unless we accept there are the fictional objects they purport to refer to or quantify over.2 In response to these arguments Sainsbury offers a complex irrealist account of such discourse. In some cases, such as (1), Sainsbury suggests the claim can be paraphrased in a way which avoids commitment to fictional objects.3 In others, such as (2)–(3), he suggests we should take the claim to be made within the scope of a presupposition.4 Thus, when we utter (2), we presuppose that there are such people as Anna and Emma and that they are the way the respective fictions describe them as being. When we utter (3), we presuppose that there is a fictional character, Holmes, who is as smart as the stories describe him as being. Utterances of (2)–(3) are not genuinely true or absolutely true but merely true under the scope of, or relative to, the relevant presuppositions. The realist mistakenly takes them to be genuinely true because she mistakes truth under a presupposition for genuine truth.5 But perhaps the central component of Sainsbury’s account is an important and ambitious argument which aims to show that the truth of fictional intensional transitives such as (4)–(5) do not commit us to fictional objects. Moreover, Sainsbury argues, his argument extends to cover claims such as (6)–(8) as well, since these should ultimately be analyzed as involving intensional verbs. 6 I will call this Sainsbury’s master argument.7 If Sainsbury’s master argument succeeded, the irrealist could happily maintain that much fictional character discourse is genuinely true, and not merely true within the scope of a presupposition, without thereby having to find a non-committal paraphrase or semantics for that discourse. This would be a very welcome result for the irrealist indeed. Unfortunately, I will argue, Sainsbury’s master argument fails. I will conclude by considering where this leaves the irrealist.

2

For classic statements of this sort of argument, see (van Inwagen 2000), (van Inwagen 2001) and (van Inwagen 2003). 3 See (Sainsbury 2009) pp. 101–5. 4 See (Sainsbury 2009) pp. 122–5 and pp. 143–8. 5 Sainsbury’s notion of presupposition has some obvious similarities with Stalnaker’s that Sainsbury brings out in (Sainsbury 2010a). But in his (Sainsbury 2009), rather than invoke a theoretically loaded notion of presupposition, Sainsbury prefers to invoke a broader intuitive and pre-theoretic notion, suggesting that we use a notion of ‘people presupposing something in the sense of taking it for granted for the purposes at hand’ (see (Sainsbury 2009) p. 119). 6 See (Sainsbury 2009) pp. 103–6 and pp. 137–8. 7 This argument is presented in slightly different ways, and plays different dialectical roles, in (Sainsbury 2009) and (Sainsbury 2010b). I follow the latter presentation more closely, but I think that what I say about the argument applies equally well in both cases.

Sainsbury on Thinking about Fictional Things

183

1 The Master Argument Sainsbury takes the ontology of a sentence to be those things that need to exist in order for the sentence to be true, formulating this as follows: Ont: x belongs to the ontology of p iff for every world w, if p is true in w then x exists at w.8 In so far as we have an intuitive notion of the ontology of a sentence, Sainsbury accepts that Ont does not quite capture it. 9 As Sainsbury notes, Ont entails that everything belongs to the ontology of a necessarily false sentence. Indeed, it entails that a necessarily false sentence has impossible things such as the round square and the highest prime in its ontology. It entails that those things which exist necessarily belong to the ontology of every sentence. Moreover Ont entails that two necessarily equivalent sentences share the same ontology. So, presuming that the singleton of Vann McGee exists in just those worlds in which McGee himself exists, it follows that “Vann McGee exists” and “the singleton of Vann McGee exists” have the same ontology. Granted Ont, the ontology of a sentence outstrips what we might intuitively take it to be. In particular, it outstrips those entities that explicitly figure in the truth conditions of that sentence. As we shall see later, this will end up creating problems for Sainsbury’s argument. Sainsbury also suggests the following principle: (EO) If p entails q, then the ontology of q does not exceed that of p,10 where the relevant notion of entailment is necessitation, so that p entails q just in case q is true in every world in which p is true. 11 Given this notion of entailment (EO) obviously follows from Ont. Sainsbury’s strategy is as follows. Let a V-sentence be a sentence involving an intensional transitive verb such as (4)–(5). Let an O-sentence be a sentence formed by applying an intensional operator, among which Sainsbury includes attitude verbs, to a sentential complement. 12 Let’s call a sentence which is not a V-sentence or an Osentence an ordinary sentence. Finally let’s call an ordinary sentence that does not have fictional objects in its ontology a neutral ordinary sentence and an ordinary sentence containing expressions which purport to refer to, or quantify over, fictional entities an ordinary fictional sentence. First Sainsbury argues that O-sentences do not have fictional objects in their ontologies. Then he argues that fictional V-sentences, V-sentences such as (4)–(5) which appear to take fictional entities as their objects, are entailed by O-sentences or by O-sentences taken together with neutral ordinary sentences. It then follows from (EO) that fictional V-sentences such as (4)–(5) do not have fictional objects in their ontologies after all. Since Sainsbury suggests that a range of further sentences such as (6)–(8) can be analyzed in terms of, or are at least entailed by, fictional V-sentences, these sentences will not have fictional objects in their ontologies either. Let’s turn to the details of the argument. 8

See (Sainsbury 2010b) p. 311 and (Sainsbury 2009) p. 140. See (Sainsbury 2010b) p. 311, note 7, and (Sainsbury 2009) p. 140, note 16. 10 See (Sainsbury 2010b) p. 311 and (Sainsbury 2009) p. 140. 11 I take it that the relevant notion of possibility here is metaphysical, rather than, say, logical or physical. 12 See, for example, (Sainsbury 2010b) p. 303. 9

184

A. Everett

Now, for Sainsbury, ordinary fictional sentences such as “Holmes is a detective” certainly do have fictional objects in their ontologies. 13 Such sentences cannot be genuinely true, or true in an absolutely sense, unless the relevant fictional objects exist. Nevertheless, Sainsbury develops a semantic framework in (Sainsbury 2005) (hereafter RWR) which allows the irrealist to assign such sentences truth-evaluable contents without having to invoke referents for the fictional names they contain.14 For our purposes we need not concern ourselves with the details of Sainsbury’s semantic framework here. The point is simply that, while the truth of ordinary fictional sentences may require the existence of fictional objects, their having truthevaluable contents does not. Sainsbury then argues that an O-sentence does not inherit the ontology of the sentence it embeds.15 Thus, for example, while “Holmes is a detective” and “Pegasus can fly” have Holmes and Pegasus in their respective ontologies, the O-sentences “I believe that Holmes is a detective” and “I imagine that Pegasus can fly” do not. All that is required for the truth of “I believe that Holmes is a detective” is that I have a belief with the truth-evaluable content of “Holmes is a detective.” All that is required for the truth of “I imagine that Pegasus can fly” is that I have an imagining with the truth-evaluable content of “Pegasus can fly.” As I just noted, the semantic framework of RWR allows the irrealist to assign truth-evaluable contents to “Holmes is a detective” and “Pegasus can fly” without thereby incurring a commitment to Holmes or Pegasus. It seems, then, that given Sainsbury’s semantic framework, Osentences do not have fictional objects in their ontologies. Two brief qualifications to this argument are in order here. For not all O-sentences can be genuinely or absolutely true without this requiring us to posit fictional objects.16 Firstly, consider factive attitude constructions, such as those involving the attitude verbs “knows,” “remembers,” and “learns.” One can know that P, remember that Q, and learn that R, only if P, Q, and R, are true. So factive O-sentences that embed ordinary fictional Given the semantic framework Sainsbury develops in his (Sainsbury 2005), “Holmes is a detective” cannot be true unless the occurrence of “Holmes” that it contains genuinely refers. Sainsbury holds that, while such sentences are false in an absolute sense (see (Sainsbury 2005) p. 72 and pp. 202–4 and p. 211, (Sainsbury 2009) pp. 26–31 and pp. 38–40, and (Sainsbury 2010b) p. 304), they may nevertheless be true relative to the presupposition that there are fictional characters (see (Sainsbury 2010a) p. 119, p. 125, and p. 147). Of course so-called ‘metafictional’ uses of such sentences, as when we utter “Holmes is a detective” in order to describe what is true in the Conan Doyle stories, strike many people as genuinely true. But for Sainsbury such metafictional utterances of ordinary fictional sentences such as “Holmes is a detective” are false (although the result of embedding “Holmes is a detective” within an “In the fiction” operator will be true), and any intuitions to the contrary result from conflating absolute truth with truth relative to a presupposition (or perhaps conflating absolute truth with ‘fidelity’ to the fiction—see (Sainsbury 2005) p. 203 and (Sainsbury 2009) pp. 26–31). I note there are other ways irrealists might handle such metafictional uses. Brock suggests (Brock 2002) we should understand the relevant sentences as being implicitly prefixed by an “in the fiction” operator, a suggestion Sainsbury tentatively rejects (see (Sainsbury 2009) pp. 124–5). Moreover (Azzouni 2010) develops an irrealist framework in which these sentences can be genuinely true. Thanks to a referee for forcing me to be clearer about these issues. 14 See also the appendix to 2.4 of (Sainsbury 2009) and (Sainsbury 2010b) p. 304. One might, of course, simply reject the framework that Sainsbury offers in (Sainsbury 2005), although I shall not pursue this issue here. For an illuminating critical discussion see (Textor 2010). 15 See for example (Sainsbury 2010b) pp. 303–4. 16 Once again, of course, the irrealist may accept that the relevant sentences are true relative to the presupposition that there are fictional characters. 13

Sainsbury on Thinking about Fictional Things

185

sentences will have fictional objects in their ontologies. Such O-sentences cannot be true without the sentences they embed being true, and the truth of the embedded sentences requires the existence of fictional objects. Secondly, consider O-sentences with a fictional name in subject position, such as “Hamlet believes that grass is green.” Since the truth of this sentence obviously requires the existence of the relevant fictional entity, it too will have fictional objects in its ontology. Sainsbury does not directly address these sorts of cases. But, in fact, his argument loses nothing if we restrict our attention only to O-sentences which (i) are non-factive, and (ii) do not contain expressions purporting to refer to or quantify over fictional objects outside the scope of the operator. Let’s call such sentences O*-sentences. Sainsbury then suggests the following: (V) The V-sentences are entailed by the totality of O-sentences and ordinary sentences.17 As it stands, I don’t think this captures what Sainsbury intends, for it is trivial and will not do the work required. It is trivial because there is no world in which the totality of O-sentences and ordinary sentences can all be true. For this totality will include sentences of the form P and ¬P , as well as necessarily false sentences such as “Cicero is distinct from Tully” and “Mt. Everest believes that mountains are snowy.” Since the members of this totality can never all be true together, the totality will trivially entail everything. More seriously, if it follows from Ont that this totality of sentences will have everything, even impossible objects, in its ontology. So the fact that this totality entails a V-sentence will in no way restrict the ontology of that Vsentence. Still, it seems easy enough to reformulate Sainsbury’s principle. Firstly, since the arguments for fictional realism appeal to true sentences which purport to be about fictional objects, we can safely restrict our principle to V-sentences that can possibly be true; we can ignore necessarily false V-sentences.18 Secondly, in the light of our discussion earlier, we had better formulate our principle in terms of O*-sentences rather than O-sentences. Finally, let us say that a set of sentences are co-possible just in case it is possible for them to all be true in the same world. I suggest the following as a reformulation of Sainsbury’s principle that will better serve his purposes: (VO) Each V-sentence is entailed by a set of co-possible sentences Σ each of whose members is either (i) an O*-sentence or (ii) a neutral ordinary sentence. In what follows I will assume that Sainsbury is willing to accept (VO) in place of the principle he offers. By (EO), the ontology of a V-sentence cannot outstrip the ontology of the sentences which entail it. If (VO) was true, and if we grant that O*sentences and neutral ordinary sentences do not have fictional objects in their ontologies, V-sentences can’t have fictional objects in their ontologies either. If this argument worked, this would be an impressive result. 17

See (Sainsbury 2009) p. 138. In fact Sainsbury expands the class of O-sentences to include all ordinary sentences, all sentences ‘not showing any signs of intensionality,’ and formulates his principle as follows: The V-sentences are entailed by the totality of O-sentences. 18 Given Sainsbury’s notion of a sentence’s ontology, necessarily false V-sentences such as “Mt. Everest is thinking about Holmes” will have everything in their ontologies.

186

A. Everett

2 The First Problem The first worry I want to raise with Sainsbury’s master argument concerns exactly what it establishes. Note that it doesn’t force the fictional realist to abandon her ontology of fictional objects. 19 Suppose the fictional realist accepts the semantic framework of RWR and that she also accepts (EO) and (VO). Now recall that the notion of a sentence’s ontology characterized by Ont is more liberal than our intuitive notion of a sentence’s ontology. It does not merely include those entities which explicitly figure in the truth conditions of the sentence, but rather all those entities, which exist in every possible world where the sentence is true. The semantic framework of RWR assigns fictional O*-sentences truth conditions without our having to posit fictional objects. So this semantics doesn’t force us to accept that fictional O*-sentences have fictional objects in their ontology. But, given the broader notion of a sentence’s ontology in play here, the semantics doesn’t preclude fictional O*-sentences from having such entities in their ontology either. The semantic framework of RWR is simply silent on this issue. This means that, on its own, the master argument cannot serve as an argument against realism. It needs to be supported by a further argument that fictional O*-sentences do not have fictional objects in their ontologies, in the broad sense of ontology employed by Sainsbury; this is an argument Sainsbury does not give. To bring this point out, observe that Sainsbury himself notes that, even granted the framework of RWR, if fictional objects exist necessarily, they will automatically be in the ontology of every sentence, obviously including fictional O*-sentences and V-sentences. So Sainsbury’s argument will certainly not move the realist who takes fictional objects to be necessarily existing entities. Moreover, a realist who takes fictional objects, not to exist necessarily, but rather to be created by the contingent literary activities of authors, might argue as follows. In order for someone to bear an attitude to the content that Holmes is a detective, they need to have a singular concept of Holmes. In order to have this singular concept, they will need to be suitably related to the fictions in which Holmes occurs. In order for these fictions to exist their author must, say, have made pretend acts of reference when writing them using the name “Holmes.” Thus the ontology of “Mary believes that Holmes is a detective” includes pretend acts of reference involving the name “Holmes.” But, the fictional realist might continue, necessarily any such act will bring the Holmes-object into being. 20 Hence the Holmes-object will be part of the ontology of “Mary believes that Holmes is a detective.” In every world where that sentence is true, the Holmes-object will exist. Of course the irrealist will deny that fictional objects exist necessarily or that they are brought into existence as the result of acts of pretend reference. But the point here is that, 19

Although in some places Sainsbury seems to suggest otherwise (see for example (Sainsbury 2009) pp. 138–141) I take it that Sainsbury would ultimately agree with this. His observation that objects that exist necessarily will be in the ontology of every sentence, and hence that if fictional objects exist necessarily, they will be in the ontology of every sentence (see (Sainsbury 2009) p. 141 and (Sainsbury 2010b) pp. 312– 3) suggests he recognizes that, on its own, his argument will not refute realism. 20 Thus, for example, Thomasson suggests: If an author pretensefully uses a name N in the context of writing a fictional work, where the author is not intending to refer back to any real thing with the name N, then there is a fictional character N (Thomasson 2010, p. 136).

Sainsbury on Thinking about Fictional Things

187

in order for Sainsbury’s argument to have any force against the realist, the realist must accept that fictional O*-sentences don’t have fictional objects in their ontologies. However, given the broad notion of ontology in play, nothing we have said so far forces the realist to accept this. The realist can quite happily accept the semantic framework of RWR while still maintaining that fictional O*-sentences have fictional objects in their ontologies. So, on its own, the master argument certainly doesn’t undermine the fictional realist. Sainsbury would need an independent argument that fictional O*-sentences don’t have fictional objects in their ontologies, in the broad sense of ontology in play. If Sainsbury’s argument doesn’t directly undermine realism, could it, nevertheless, help the irrealist by allowing her to block an otherwise powerful argument for fictional realism, the argument from the truth of fictional V-sentences? Well ultimately the master argument is only as strong as our grounds for denying that O*-sentences have fictional objects in their ontologies. If the master argument is to do the required work, our grounds for thinking that O*-sentences lack fictional objects in their ontologies must be stronger than our grounds for thinking that fictional V-sentences have fictional objects in their ontologies. But fictional V-sentences seem, at least prima facie, to have fictional objects in their ontologies. So the irrealist cannot simply assume that O*-sentences lack fictional objects in their ontologies. Rather she will have to offer an argument to this effect. To recap, Sainsbury’s master argument can only be used to attack realism or defend irrealism if supplemented with a further argument that fictional O*-sentences don’t have fictional objects in their ontologies. However, given that the ontology of a sentence are those things that exist in every world in which that sentence is true, it seems the only way to establish that O*-sentences don’t have fictional objects in their ontologies is to establish that there are no fictional objects. At any rate, it is not clear how else we might go about establishing it. The problem here is that, if we can establish there are no fictional objects, we don’t need Sainsbury’s master argument. We have a much more direct argument to the conclusion that fictional V-sentences don’t have fictional objects in their ontologies. The conclusion will follow directly from our argument for the nonexistence of fictional entities. As it stands, then, I don’t see how the master argument can help the irrealist. It doesn’t undermine realism. As far as I can see, it can only be used to attack realism or defend irrealism if the irrealist already has an argument which renders the master argument superfluous. These difficulties arise because Sainsbury employs a coarse-grained notion of a sentence’s ontology. Because of this, even if the truth conditions of fictional O*-sentences do not invoke fictional objects, this doesn’t prevent those sentences from having fictional objects in their ontologies. So it is natural to wonder whether Sainsbury’s argument could be salvaged if we employed some finer-grained notion of a sentence’s ontology. 21 21

More generally, the notion of a sentence’s ontology given by Ont seems the wrong notion for an irrealist to invoke. If we seek to avoid commitment to Fs, the fact that a sentence lacks Fs in its ontology in the sense given by Ont is only of limited relevance. Suppose that we can introduce an actuality operator which rigidly picks out the actual world, so that “Actually P” is true with respect to a world w just in case, in the actual world, P. Then “Actually there are cats” does not include cats in its ontology, in the sense given by Ont, since the sentence is true with respect to every world, but presumably cats do not exist in every world. Nevertheless, the irrealist about cats can hardly accept the truth of “Actually there are cats” while maintaining their irrealism about cats on the grounds that, in the sense of “ontology” given by Ont, “Actually there are cats” doesn’t have cats in its ontology.

188

A. Everett

Obviously we would also have to employ a corresponding finer-grained notion of entailment on which (EO) and (VO) held. There are various notions that one might try here, but, fortunately, we don’t need to consider their details. For I will argue below that there are intuitively true V-sentences whose truth would commit us to the existence of fictional objects. Consequently, it doesn’t really matter what notion of a sentence’s ontology or of entailment is employed in Sainsbury’s master argument; his argument will still fail to achieve its goals.

3 The Second Problem Suppose that Mary is thinking about Holmes but is unaware that he is a fictional character, mistakenly supposing that he is a real person. Despite Mary’s confusion, the following reports would seem to be intuitively correct22: (9) Mary is thinking about a fictional character, (10) Mary is thinking about the most famous fictional detective. In the circumstances we are considering, however, the descriptive conditions fictional character and famous fictional detective are certainly not part of the content of Mary’s thought. Mary does not deploy the corresponding concepts when she thinks about Holmes. She might vehemently deny she is thinking about anything fictional. Indeed, she may lack the concepts fictional character and famous fictional detective all together. Because of this (9)–(10) will be false on their intensional readings, readings on which their truth is sensitive to how Mary is thinking about her subject matter.23 Rather it seems we must understand the readings of (9)–(10) on which they are intuitively true as ones on which they make extensional de re reports, reports in which the descriptions take wide scope relative to the verb: (11) A fictional character is such that Mary is thinking about them, (12) The most famous fictional detective is such that Mary is thinking about them.

Note that the claim for the moment is not that “Mary is thinking about Holmes” entails (9)–(10) but rather simply that (9)–(10) are true in the circumstances we are envisaging. Note that “Mary is thinking about Holmes” certainly doesn’t appear to entail (10), even on Sainsbury’s coarse-grained notion of entailment, since intuitively there will be worlds in which Holmes is not the most famous fictional detective (say those in which Miss Marple is the most famous fictional detective). While, on Sainsbury’s coarsegrained notion of entailment, “Mary is thinking about Holmes” would plausibly entail (9), on a fine-grained notion of entailment it would entail (9) only modulo the further assumption that Holmes is a fictional character. 23 On this reading the truth of (9), for example, would require that Mary be thinking of Holmes as a fictional character, or at least that she recognize that Holmes is a fictional character. Since this is not the case in the scenario I outlined, (9) will not be true on an intensional reading. The burden is very much on Sainsbury here to explain how (9)–(10) could have true non-extensional readings in the scenario described. I will explore the different ways in which Sainsbury might deny that (9)–(10) have true de re readings later on in this section, but argue that they all threaten his master argument. In contrast note that (9) would be true on an intensional reading if Mary is thinking that Kennedy was inspired by a fictional character. Likewise, (10) would be true on an intensional reading if Mary is thinking that the most famous fictional detective could have got the goods on Nixon. Thanks to a referee for forcing me to be clearer about these issues. 22

Sainsbury on Thinking about Fictional Things

189

However, the truth of (11)–(12) seem to commit us to fictional objects. Likewise, suppose that Mary reads War and Peace and mistakenly takes it to be a work of fact rather than fiction. Suppose she thinks about the many people she has read about in War and Peace. Then the following report seems correct: (13) Mary is thinking about many fictional characters. Since, however, Mary is unaware that she is thinking about fictional characters, (13) is false on its intensional reading. Rather, it seems, the relevant reading of (13) must be a relational reading, on which the quantifier takes wide scope: (14) Many fictional characters are such that Mary is thinking about them. It is very hard to see how (14) could be genuinely true unless there were fictional characters. Here is a particularly pointed case of the same sort of phenomenon. Suppose that Mary does not know of Sally’s existence, but that Mary and Sally both happen to be reading the same Sherlock Holmes story and are both wondering whether Holmes will catch the crook. Then it seems Mary is thinking about Holmes and that Sally is also thinking about Holmes. So the following are true: (15) Mary is thinking about Holmes. (16) Sally is thinking about Holmes. These jointly entail: (17) Mary is thinking about Holmes and Sally is thinking about Holmes. But note that, on any reasonable notion of entailment, (17) would seem to entail24: (18) Mary is thinking about the thing Sally is thinking about. In the situation we are considering (18) seems intuitively true. But since Mary does not even know of Sally’s existence, the descriptive condition the thing that Sally is thinking about is certainly not part of the content of Mary’s thought. Indeed, since Mary lacks a concept of Sally, she does not have the conceptual resources to formulate even this description. Consequently it seems we must understand (18) as making a de re report; we must take the description “the thing that Sally is thinking about” to take wide scope relative to the attitude verb, so that we render (18) as: (19) The thing that Sally is thinking about is such that Mary is thinking about it.

24

Strictly speaking for this entailment to hold a further background assumption is required to ensure that Mary and Sally are both thinking about ‘the same Holmes,’ as it were. That is to say, for the entailment to hold we need to exclude cases where, for example, (15) is true because Mary is thinking about Holmes (the fictional character) and (16) is true because Sally is thinking about Holmes (the glacier in Antarctica). Either of the following, both of which I take to be true in the scenario we are considering, will do the trick: (i) Mary’s thought and Sally’s thought are prompted in exactly the same way by their reading the same Holmes story and thinking about its central character, (ii) under the presupposition that the Holmes-stories are fact rather than fiction, Mary’s thought and Sally’s thought both count as having Sherlock Holmes as their object. Thanks to a referee for forcing me to be clearer about this point.

190

A. Everett

But then the truth of (18) in the situation we are considering seems to require the existence of the think Sally is thinking about, namely Holmes.25 The irrealist has little room for maneuver here. One response to these examples would be to try to argue that (9), (10), (14) and (18) are genuinely true but that their truth does not commit us to the existence of fictional objects after all. But to be convincing this would require providing a semantics on which (9), (10), (14) and (18) count as true in the situations we considered above, even though there are no fictional objects. It is far from clear what such a semantics would look like. Moreover, of course, this semantics would have to permit sentences (such as (9), (10), (14) and (18)) involving expressions purporting to denote or quantify over fictional objects to be true, even where the expressions are not within the scope of any intensional operator.26 One suspects that any semantics that could do this would permit any intuitively true fictional V-sentence to be genuinely true without requiring the existence of fictional objects. But if so, then the need for Sainsbury’s master argument itself would be undercut, the irrealist could achieve the same result more directly by invoking a semantics they needed anyway to make the master argument work. The other response to these examples would be to deny simply that (9), 10), (14) and (18) are genuinely true. But this raises two serious problems. Firstly, it significantly limits the dialectical interest of the master argument. The fictional realist motivates her position by noting that a range of fictional V-sentences are intuitively true and arguing this commits us to fictional objects. The aim of the master argument was to show that the irrealist can accept the truth of these fictional V-sentences 25 One might well worry whether (18) does, in fact, have a true de re reading and presumably Sainsbury will deny that it does. But (18) does intuitively seem to have a true reading. This creates problems for Sainsbury’s master argument since this reading cannot be an intensional reading, as Mary does not know of Sally’s existence and so the descriptive condition the thing that Sally is thinking about cannot be part of the content of Mary’s thought. Because of this Sainsbury must either deny that the intuitively true reading of (18) is in fact genuinely true or he must deny that the intuitively true reading of (18) is a de re reading. I will argue later in this section that both options raise problems for Sainsbury’s master argument. Note in particular that, in the latter case, Sainsbury will need to explain what the relevant reading of (18) is. More precisely he will need to provide a plausible semantics for (18) on which it has a true non-intensional reading that is not a de re reading. Sainsbury himself does not offer such a semantics and it is very hard to see what such a semantics might look like. Nevertheless, if Sainsbury maintains that (18) has a reading on which it is genuinely true in the scenario considered, the burden lies with him to provide such a semantics. For otherwise the realist can simply argue, from there being a true non-intensional reading of (18), to there being a true de re reading of (18), and, hence, to there being fictional objects. The referee also rightly notes the similarity between (18) and other cases of so-called intentional identity such as the Hob-Nob sentence discussed by (Geach 1967):

(*) Hob thinks a witch has blighted Bob’s mare, and Nob wonders whether she (the same witch) killed Cob’s sow, suggesting such sentences don’t have true de re readings but may have alternative true readings. These cases are puzzling and I have little illuminating to say about them here. But once again I would note that, in so far as (*) seems to have a genuinely true reading that is not de dicto, we seem under pressure to accept there genuinely is some object which both Hob and Nob have in mind (see (Salmon 2005) for a view along these lines and for criticisms of various alternative approaches). Thus the onus is very much on those who maintain that (*) could be genuinely true without there being some object which both Hob and Nob have in mind to provide a semantics on which this could be so. 26 Because of this Sainsbury cannot simply maintain, for example, that all V-sentences are disguised O*-sentences, although in any case he expresses skepticism about this view (see (Sainsbury 2009) pp. 135–6). Thanks to a referee for forcing me to be clear about this.

Sainsbury on Thinking about Fictional Things

191

without thereby incurring a commitment to fictional objects. If the irrealist has to deny that the intuitively true (9), (10), (14) and (18) are genuinely true to avoid a commitment to fictional objects, then the master argument has failed to achieve its goal and the irrealist is left in basically the same dialectical situation as before. The second problem with the response is this. If we deny that (9), (10), (14) and (18) are genuinely true then we must also deny that any sentences which entail them are true. Now intuitively (15) and (16) certainly seem to entail jointly (18).27 But then, if (18) is not true in the scenario we envisaged, neither is the conjunction of (15) and (16). Indeed, since it is hard to see how one of (15) and (16) could be true in the scenario we considered while the other is false, in practice we will have to deny the truth of both (15) and (16). Moreover, of course, we must also deny that any O*-sentences which entail (15) and (16) are genuinely true. Note that the sort of problem generated by (15) or (16) can in principle be generated by any pair of intuitively true V-sentences containing the same fictional name in object position. Suppose Alice and Beth are unaware of each other’s existence, and that Alice admires Holmes while Beth pities Holmes. It seems that, in this situation, “Alice admires the thing Beth pities” is true. But as with (18), the truth of this sentence in this situation seems to require the existence of the thing Beth pities, it seems to require the existence of Holmes. If the irrealist denies that “Alice admires the thing Beth pities” is true it seems she must also deny the truth of “Alice admires Holmes” and “Beth pities Holmes” which entail it. Similar problems are generated by (9)–(10).28 Consequently, the irrealist who adopts this response ends up having to deny that a very significant range of intuitively true fictional V-sentences are genuinely true. The response we are considering therefore threatens to undermine entirely the goal of the master argument. The only way the irrealist might deny that (9), (10), (14) and (18) are true while avoiding this result would be to deny that the relevant entailments hold. Thus, she must maintain that (15) and (16) are true while denying that they entail (18), and so on. However, given that the relevant entailments intuitively do seem to hold, the irrealist can’t simply adopt a revisionary position, insist that they don’t, and just leave it at that. Rather she owes us an account of why they don’t. She must explain, for example, how (15) and (16) can be true while (18) is not. To do this the irrealist needs to provide a semantics for fictional V-sentences such as (15) and (16) on which their truth did not require the existence of fictional objects. Perhaps such a semantics can be given. But if it can, the need for the master argument is once more undercut. The irrealist can achieve the same result much more directly by invoking a semantics she needs anyway if her argument is to work.

27 28

That is to say, given a further background assumption of the sort discussed in footnote 24.

One might well think that it is necessarily true that Holmes is a fictional character. If so, then on a possible worlds notion of entailment (15) entails (9), there is no world in which (15) is true but (9) is false. On a finer-grained notion of entailment (15) may not, on its own, entail (9). But plausibly “Holmes is a fictional character” and (15) will jointly entail (9). Likewise plausibly “Holmes is the most famous fictional detective” and (15) will jointly entail (10).

192

A. Everett

4 Where Does This Leave the Irrealist? How should Sainsbury’s irrealist respond to this situation? First let’s consider what the irrealist should say about fictional O-sentences. Turning first to fictional O*-sentences, it is hard to deny that at least some fictional O*-sentences are genuinely true. After all, when we imaginatively engage with the Conan Doyle stories it is surely genuinely true that we imagine Holmes is a detective. It is surely genuinely true that those who mistake the Conan Doyle stories for fact genuinely believe Holmes is a detective. Moreover, given the framework of RWR, there seems nothing to prevent fictional O*-sentences from being true. For the ordinary fictional sentences they embed will have truth evaluable contents and there seems nothing to prevent people from having attitudes which have those contents. So the irrealist had better accept that at least some fictional O*-sentences are genuinely true. As we have seen however, this is not a problem for the irrealist since the framework of RWR allows O*-sentences to be true whether or not there are fictional objects. On the other hand the irrealist had better deny that intuitively true fictional O-sentences involving factive verbs are genuinely true. But here the irrealist might invoke another component of Sainsbury’s irrealism, his suggestion that certain intuitively true sentences are in fact only true under the scope of a presupposition. The relevant O-sentence would not, therefore, be genuinely true but merely true under the presupposition that there are fictional objects which are the way the fictions describe them as being. Matters are more complex when it comes to fictional V-sentences. Firstly, some have argued for propositionalist accounts of at least some intensional transitives, on which the verb is to be understood as taking a sentential complement in the same way as an O-sentence. 29 Propositionalist accounts of certain intensional transitives, in particular verbs of requirement and desire such as “wants” and “needs,” are quite plausible and supported by a range of linguistic data. Sainsbury seems willing to accept these analyses for at least some intensional transitives.30 In so far as the irrealist accepts such analyses, she had better accept that the relevant fictional V-sentences are genuinely true. On the other hand, as Sainsbury himself notes, propositionalist accounts of other intensional transitives, in particular verbs of evaluation such as “love” and “fear,” do not look very promising.31 So whether or not the irrealist takes intuitively true fictional V-sentences involving such verbs to be genuinely true will obviously depend upon the semantics for those she ultimately adopts. Finally recall that, for reasons already rehearsed, the irrealist had better deny that (9), (10), (14) and (18) are genuinely true. So at the very least it looks as if the irrealist will have to count some intuitively true fictional V-sentences as genuinely true while counting others as false. One might worry that this will force the irrealist to offer an unsatisfyingly heterogeneous and non-uniform explanation of our intuitions about the truth of these sentences. Fortunately it doesn’t. For the irrealist can argue as follows. When we use fictional V-sentences, our aim is to characterize the content of various attitudes in the same way that we characterize the content of attitudes that are directed towards real things. We therefore talk and think as if the attitudes we are characterizing were directed 29

See (den Dikken et al. 1996), as well as (Larson 2002) and (Parsons 2010). See (Sainsbury 2009) pp. 132–35 and (Sainsbury 2010b) pp. 313–14. 31 See, for example, chapter 4 of (Forbes 2006), as well as (Montague 2007), and (Sainsbury 2009) pp. 135–36. 30

Sainsbury on Thinking about Fictional Things

193

towards genuine objects. In Sainsbury’s parlance we presuppose there are fictional objects; we take their existence for granted for the limited purposes of our attitudecharacterizations.32 On this view we produce and consume all fictional V-sentences within the scope of the presupposition that there are fictional objects. 33 In all such cases, we judge intuitively true V-sentences to be true because they are true under the scope of this presupposition. Perhaps in some of these cases the correct semantics for an intuitively true V-sentence doesn’t require the existence of fictional objects, say because the correct semantics for the verb is propositionalist. If so, the irrealist can accept that the V-sentence is not merely true under the presupposition but also genuinely true. In other cases, where the truth of the V-sentence would require the existence of fictional objects, she can maintain that the relevant V-sentences are not genuinely true but merely true under the scope of the presupposition. But in both sorts of cases we can provide a uniform explanation of our intuitions by noting that the Vsentences are true under the scope of the presupposition and maintaining that it is this that drives our intuitions. Thus I think Sainsbury’s irrealist has the resources to respond to the realist’s argument. The irrealist can avoid a commitment to fictional objects. But she has to admit that at least some intuitively true fictional V-sentences are not genuinely true. Sainsbury’s master argument ultimately fails.

References Azzouni, J. (2010). Talking about Nothing. Oxford: Oxford University Press. Brock, S. (2002) Fictionalism about fictional characters. Nous, (36), 1–21. den Dikken, M., Larson, R., & Ludlow, P. (1996). Intensional “transitive” verbs and concealed complement clauses. Rivisita di linguistica, 8, 29–46. Forbes, G. (2006). Attitude Problems. Oxford: Oxford University Press. Geach, P. (1967). Intentional identity. The Journal of Philosophy. (64), 627–32. Larson, R. (2002). The grammar of intensionality. In G. Preyer & G. Peter (Eds.), Logical Form and Natural Language (pp. 228–262). Oxford: Oxford University Press. Montague, M. (2007). Against propositionalism. Nous, 41(3), 503–518. Parsons, T. (2010). Meaning sensitivity and grammatical structure. In M. L. D. Chiar, K. Doets, D. Mundici, & J. van Benthem (Eds.), Structures and Norms in Science (pp. 369–383). Dordrecht: Kluwer. Sainsbury, M. (2005). Reference Without Referents. Oxford: Oxford University Press. Sainsbury, M. (2009). Fiction and Fictionalism. London: Routledge. Sainsbury, M. (2010a). Fiction and acceptance-relative truth, belief, and assertion. In F. Lihoreau (Ed.), Truth in Fiction (pp. 137–152). Frankfurt: Ontos Verlag. Sainsbury, M. (2010b). Intentionality without exotica. In R. Jeshion (Ed.), New Essays on Singular Thought (pp. 300–318). Oxford: Oxford University Press.

32

Sainsbury may well be sympathetic to such a move. In his (Sainsbury 2009) and (Sainsbury 2010a) he suggests treating some intuitively true V-sentences, such as “The Greeks worshipped Zeus” and “John thought about Pegasus” as being merely true within the scope of a presupposition, see in particular (Sainsbury 2009) pp. 143–6 and (Sainsbury 2010a) pp. 157–9, although he is reluctant to extend this treatment to all V-sentences. In his earlier (Sainsbury 2005) Sainsbury notes that, on the semantic framework of RWR, fictional V-sentences appear to come out as false and he tentatively considers some strategies for explaining our intuitions to the contrary, see pp. 207–8. 33 This will be equally true for the realist and the irrealist; it is simply that the realist will believe what they presuppose while the irrealist will not.

194

A. Everett

Salmon, N. (2005). Mythical objects. In his Metaphysics, Mathematics, and Meaning (pp. 91–107). Oxford: Oxford University Press. Textor, M. (2010). Proper names and practices: On reference without referents. Philosophy and Phenomenological Research, 1(81), 105–118. Thomasson, A. (2010). Fiction, existence and indeterminacy. In J. Woods (Ed.), Fictions and Models: New Essays (pp. 109–148). Munich: Philosophia Verlag. van Inwagen, P. (2000). Quantification and fictional discourse. In A. Everett & T. Hofweber (Eds.), Empty Names, Fiction, and the Puzzles of Non-Existence (pp. 235–4). Stanford: CSLI Press. van Inwagen, P. (2001). Creatures of fiction. In his Ontology, Identity, and Modality (pp. 37–56). Cambridge: Cambridge University Press. van Inwagen, P. (2003). Existence and ontological commitment. In M. Loux & D. Zimmerman (Eds.), Oxford Handbook of Metaphysics (pp. 131–157). Oxford: Oxford University Press.