This is the penultimate draft of a paper that is published in Nicholas Griffin and Dale Jacquette (eds.), Russell vs. Meinong: The Legacy of "On Denoting", Routledge, 2009. Please quote only from the final draft.

Who is Afraid of Imaginary Objects? Gabriele Contessa Carleton University [email protected] People often use expressions such as ‘Sherlock Holmes’ and ‘Pegasus’ that appear to refer to imaginary objects. In this paper, I consider the main attempts to account for apparent reference to imaginary objects available in the literature and argue that all fall short of being fully satisfactory. In particular, I consider the problems of two main options to maintain that imaginary objects are real and reference to them is genuine reference: possibilist and abstractist account. According to the former, imaginary objects are possible concrete objects. According to the latter, imaginary objects are actual abstract objects. I will then propose an account, the dualist account, which, I argue, combines the respective advantages of both accounts without sharing any of their respective disadvantages. According to this account, imaginary objects are not fully reducible to either abstract objects or possible objects: they are abstract artefacts that, in some contexts, stand for possible objects.

‘There are more things in the heaven and earth Horatio, then are dreamt of, in your philosophy’ William Shakespeare, Hamlet, Act I, Scene V I. THE PROBLEM WITH IMAGINARY OBJECTS People often use expressions such as ‘Sherlock Holmes’ and ‘Pegasus’, which appear to refer to imaginary objects, and make assertions about them such as: (1) Sherlock Holmes smokes a pipe

and (2) Sherlock Holmes was created by Arthur Conan Doyle in the late Nineteenth century.

In what follows, I shall call internal those sentences that, like (1), talk of imaginary objects as if they were concrete actual objects and external those sentences that, like (2), talk of imaginary objects as fictional characters, mythological creatures and so on (whatever these entities might be). I shall call object-fictional those internal sentences that occur in a work of fiction and meta-fictional those sentences, whether internal or external, that occur outside of a work of fiction and are about that work of fiction. Thus, (1) is an object-fictional sentence if it occurs in one of Arthur Conan Doyle but it is a meta-fictional sentence if it occurs in the context of the discussion of Conan Doyle’s works. Prima facie, speakers’ reference to imaginary objects would seem to be incompatible with a widely accepted philosophical principle. One of the first formualtions of this principle in western philosophy can be found in Plato’s Sophist (237B–E). In this paper, however, I will refer to a much more recent formulation of the principle, which has been put forward by John Searle (1969) who calls it the Axiom of Existence. In Searle’s formulation, the Axiom of Existence states: ‘Whatever is referred to must exist’ (Searle 1969, p.77). The problem is that, if we take ‘existence’ to designate the mode of being that is proper of actual concrete objects such as flesh-and-bone people, animals and material objects, then imaginary objects, like Sherlock Holmes and Pegasus, do not exist and therefore, according to the Axiom of Existence, it is not possible to refer to them.

Those who accept the Axiom of Existence have two options. They can either maintain that reference to imaginary objects is only apparent or that imaginary objects exist after all. In what follows, I will call eliminativist any account according to which imaginary objects do not exist and reference to imaginary objects can be explained away as merely apparent. To show that apparent reference to imaginary objects is not genuine philosophers have usually relied on one of three main strategies. In Part II, I briefly outline each of them and argue that they all fail to be fully descriptively adequate. In Part III, I consider the option of admitting that imaginary objects exist and that reference to them is genuine reference. I shall call any account that maintains that imaginary objects in some sense exist hospitable. Different hospitable accounts however disagree as to the nature of imaginary objects. According to possibilist accounts, imaginary objects are possible concrete objects, while, according to abstractist accounts, they are actual abstract objects. I will then argue that neither option is fully satisfactory. I will then outline a hospitable account, the dualist account, which, as I shall argue, combines the advantages of the versions of the possibilist account and the abstractist account without sharing their respective disadvantages. According to the dualist account, imaginary objects have an intrinsically dual nature and they cannot be entirely reduced to either abstract objects or imaginary objects. An imaginary object is an abstract object that stands for a possible object. Before starting the examination of the various accounts to imaginary objects, it is important to note that this paper is meant to be a piece of descriptive and not prescriptive metaphysics to use Peter Strawson’s distinction (Strawson 1959). That is, its aim is to make philosophical sense of the practice of talking and thinking about imaginary objects and of the intuitions that underlie this practice. It is not meant to modify or censor this practice and intuitions but to uncover their implicit presuppositions. II. ELIMINATIVIST APPROACHES II.1. The Predicate Account The first eliminativist strategy I will examine, the predicate account, attempts to paraphrase the sentences in which apparently referential expressions occur into synonymous sentences containing no such expressions. Bertrand Russell (1905) and Willard van Orman Quine (1948) are probably the most prominent advocates of the paraphrase program. In its most ambitious version, the paraphrase is supposed to show the “logical form” of the sentence (like in Russell 1905). According to a more modest version, the paraphrase simply proves that reference to imaginary objects could be avoided without changing the meaning of the sentence. The paraphrase would show that, reference to imaginary objects is ultimately dispensable and does not commit us to the existence to these objects. The general idea that underlies the paraphrase program is that (1) is synonymous with (3) There exists one and only one x such that x is Sherlock Holmes and x smokes a pipe.

In the notation of first order logic, (3) has the form ∃x((Px)∧(∀y(Py→x=y))∧Qx), where ‘Px’ is interpreted as ‘x is Sherlock Holmes’ and ‘Qx’ as ‘x smokes a pipe’. Thus, (3) does not contain any referring expression. It only states the existence (and the 2

uniqueness) of an entity that has the property of being Sherlock Holmes and asserting that that entity also has the property of being a pipe smoker. If one concedes that the expression ‘Sherlock Holmes’ as it occurs in (3) is not a name but it is only part of a strange-looking predicate, the first step of the program seems thus successfully completed: (3) does not contain any expression referring to non-existent entities. However, one might wonder whether any genuine metaphysical advantage is gained from this move. Advocates of the predicate account propose to interpret sentences containing expressions that apparently refer to objects that do not exist as sentences that contain empty predicates. If, on the one hand, the advocates of the predicate account embrace an intensional view of predicates as expressions referring to properties, then they exclude non-existent entities from their ontology only at the price of admitting uninstantiated properties in it. This is a price that many would not be willing to accept. First, most philosophers who are realist about properties would not be willing to include non-instantiated properties in their ontology (Loux 2002). Second, an ontology that contains non-instantiated properties is not necessarily more austere than an ontology that contains non-existent objects. If, on the other hand, they embrace a strictly extensional view of predicates, such that two predicates are the same if they have the same extension, then they have to accept that the predicates ‘x is Sherlock Holmes’, ‘x is Dr. Watson,’ ‘x is Pegasus’ and, say, ‘x is the highest prime number’ are, in fact, the same predicate for they both have as their extension the empty set. Both options seem unsatisfactory. The second step, however, is probably even more problematic. If (3) is to be a paraphrase of any internal sentence such as (1), then (1) and (3) must be synonymous and synonymous sentences necessarily have the same truth-value. But it is far from obvious that (1) and (3) have the same truth-value. On Russel’s early account, (3) is plainly false. No actual concrete object satisfies the predicate ‘x is Sherlock Holmes’. However, knowledgeable speakers unconcerned by philosophical issues would agree that (1) is “in some sense” true and its negation false. For example, if a student writes on a true-or-false literature test that (1) is true, her answer is right, if she writes that (1) is false, her answer is wrong.1 Any descriptively adequate account of the semantics of fictional sentences should be able to account for the fact that some internal sentences such as (1) are “in some sense” true and others, such as (4) Sherlock Holmes is an astronaut,

are “in some sense” false. I will call these two phenomena respectively the qualified truth and the qualified falsity of internal sentences. Even if one was ready to deny that internal sentences such as (1) and (4) can be true or false in any sense or even have a truth-value, however, it does not seem possible to

1

A similar problem also affect those who, like Peter Strawson (1950), maintain that sentences that

contain non referring expressions are neither true nor false or those who, like the late Russell (1944), maintain that sentences of these sort are untrue rather than false. Such views cannot account for our intuition that, whatever the truth-values of sentences like ‘Sherlock Holmes smokes a pipe’ and ‘Sherlock Holmes is an astronaut’ may be, they simply do not have the same truth-value.

3

deny that external sentences, such as (2), are unqualifiedly true. That is, they are true according to historical facts and not just true relatively to some work of fiction. One day an historian of literature might claim that (2) is actually false: Sherlock Holmes was created by someone other than Conan Doyle and Conan Doyle only stole her idea. To support this claim, she would need factual evidence. I will call this phenomenon the unqualified truth of external sentences. The predicate account cannot account for the unqualified truth of external sentences. In fact, according to it, external sentences like (2) would appear to be no less false than internal sentences like (1). The predicate account thus is not descriptively adequate (and, to be fair to its advocates, it is not even intended to be so). Speakers who utter sentences like (1) or (2) do not, thereby, implicate that the entities to which they are referring exist as concrete actual objects. In the attempt to avoid any form of ontological commitment to imaginary objects, advocates of the paraphrase program systematically misconstrue the meaning of the sentences in which expressions referring to imaginary objects occur, and systematically misinterpret what the speakers mean to say when they utter those sentences. It is worth noting here that similar problems are also faced by those accounts, like Peter Strawson’s (see Strawson 1950), according to which statements that contain non-referring expressions, such as Error! Reference source not found. and (4), are neither true nor false or those who, like the late Russell (1944), maintain that statements of these sort are untrue rather than simply false. None of these accounts can account for the intuition that, whereas (1) and (4) may both be literally false, (1) is in some non-literal sense true. II.2. The Pretence Account Consider now a second eliminativist account: the pretence account. The pretence account, which in some version or other has been advocated by the likes of Kendall Walton (1973; 1990), and Gareth Evans (1982), is based on the idea that reference to imaginary objects is not genuine but only pretended and that, therefore, it is not in breach of the Axiom of Existence.2 According to the pretence account, the sentences that occur in a work of fiction belong to a special mode of discourse: the fictional discourse, which is to be carefully distinguished from ordinary factual discourse. In writing or uttering an object-fictional sentence in which expressions that appear to refer to imaginary objects occur, the author of a work of fiction pretends that she is referring to actual objects and that she is providing the audience with true information about the objects. The author also intends the audience to participate in the pretence by pretending that the author is referring to something rather than nothing and that what she says about that something is true (cf. Currie 1991, Ch. 1). Advocates of the pretence account usually note that fictional discourse is a form of make-believe similar to children playing with blobs of mud pretending that they are pies (Walton 1973) or to someone shadowboxing pretending that he is fighting imaginary opponents (Evans 1980).

2

An additional footnote in the reprint of (Strawson 1950) seems to suggest that also Strawson

subscribed to such a view of imaginary objects.

4

The pretence is, to a certain extent, also carried over in internal meta-fictional sentences. When a speaker utters an internal sentences such as (1) outside of the context of a work of fiction, both the speaker and its audience pretend that ‘Sherlock Holmes’ refers to a real person. Unlike object-fictional sentences, however, internal meta-fictional sentences are not necessarily pretended to be true. An internal metafictional sentence is pretended to be true only if it is implied by object-fictional sentences in the relevant work of fiction, so that none would pretend a sentence such as (4) to be true. The pretence account faces various problems. Here, I will consider only one, which is that the pretence account does not seem able to account for the unqualified truth of external sentences such as (2). There seems to be a perfectly legitimate sense, although maybe a not entirely literal one, in which Conan Doyle “created” Sherlock Holmes. When uttering (2), a speaker and their audience do not seem to be either pretending that ‘Sherlock Holmes’ refers to any concrete actual object nor they pretend that Arthur Conan Doyle created that object. Rather, the speaker seems to be providing her audience with genuine information about the author of the fictional character that is identified by the name ‘Sherlock Holmes’. Testimony of this is the fact that, even if we are able to pretend that, say, Anna Karenina was created by Conan Doyle, we do not seem able to pretend that Holmes was created by Conan Doyle because we actually believe that Holmes was created by Conan Doyle for it would seem that one cannot, at the same time, one can pretend that p is true only if they do not believe that p is true. II.3. The Mention Account This leads us to consider a third strategy to paraphrase away reference to imaginary objects, which I shall call here the mention account. According to the mention account, when someone is apparently using an expression that refers to an imaginary object they are actually only mentioning that expression.3 According to this account, (1) would be shorthand for (5) Someone wrote a story containing ‘Sherlock Holmes smokes a pipe’ (or containing sentences implying this).4

If (1) was actually only shorthand for (5), then the use of the expression ‘Sherlock Holmes’ in (1) would be ultimately dispensable. The mention account however cannot account for either external sentences or object-fictional sentences. Consider external sentences first. According to the schema suggested above, (2) would be shorthand for (6) Someone wrote a story containing ‘Sherlock Holmes was created by Arthur Conan Doyle in the late Nineteenth century’ (or containing sentences implying this).

3 4

A similar view has been proposed by Gilbert Ryle (1933). Such a formulation of the paraphrase schema is due to Charles Crittenden (1991). Even if

Crittenden is a critic of the mention account, his formulation of the paraphrase schema underlying the mention account seems to be the best available one.

5

Clearly, (2) is not shorthand for (6). An advocate of the mention account could probably claim that, even if that particular schema does not work in external contexts, it is still possible to claim that external sentences are shorthand for as sentences in which the expression ‘Sherlock Holmes’ is only mentioned. If the advocates of the mention account want their claims to be taken seriously, however, they should propose a set of paraphrase schemas that are meant to deal with all external sentence, including ‘Sherlock Holmes is my favourite character in English literature’ or ‘Sherlock Holmes uses abductive reasoning more frequently than Hercules Poirot’. Until then it is not possible to seriously assess their claim that all external sentences are shorthand for sentences in which expressions which refer to imaginary objects are only mentioned. Moreover, the suggested paraphrase schema does not seem to work with objectfictional sentences either. If (1) occurred in one of Conan Doyle’s stories, it cannot be interpreted as shorthand for (5). In fact, in his stories, Conan Doyle is not reporting the content of someone else’s stories to his readers. Advocates of the mention account could probably concede that, when it occurs in the context of object-fictional sentences, the expression ‘Sherlock Holmes’ is used and not mentioned. Nevertheless, they could claim, the expression is not used referentially or, even if it is used referentially, it does not refer to anything. The problem with such a possible reply would be that, in object-fictional sentences, ‘Sherlock Holmes’ seems to be used referentially and seems to refer to a certain character in the novel and not to others. By writing a sentence like (1) in one of his stories, Conan Doyle meant to tell us that the character to which he refers to as ‘Sherlock Holmes’ and not any other of the characters in the novel smokes a pipe. Thus, not only ‘Sherlock Holmes’ seems to be used referentially in object-fictional sentences, but also it seems to be used to refer to a certain object and not others. The practices of writing and reading works of fiction (in which there is more than one character) presupposes that different expressions can be used to refer to different characters. If ‘Sherlock Holmes;’ and ‘Professor Moriarty’ failed to refer to different objects, we would not be able to distinguish between ‘Sherlock Holmes smokes a pipe’ and ‘Professor Moriarty smokes a pipe’. III. HOSPITABLE APPROACHES As they stand, none of the main attempts to show that reference to imaginary objects is not genuine seems to be descriptively adequate and some of the problems that these accounts have encountered could be taken as evidence that reference to imaginary objects is, after all, genuine. Those who adopt a hospitable approach, thus, are willing to grant that whatever is referred to must exist. From this assumption, however, they reach the opposite conclusions: since it seems possible to genuinely refer to imaginary objects, imaginary objects must exist. Since they are not concrete actual objects, they must have a different ontological status. The agreement among advocates of hospitable approaches, however, does not go much further than this. Hospitable accounts can be roughly grouped in two broad families: possibilist and abstractist accounts. Even if both possibilist accounts and abstractist accounts maintain that imaginary objects exist, possibilists and abstractists have rather different ideas as to what they are.

6

III.1. Possibilist Accounts The main idea behind possibilist accounts is that imaginary objects, though not actual, exist and, as such, they can be referred to. Among the various possible variations on the possibilist theme, I will consider only two here, which I shall call the combinatorial account and the possible world account. The combinatorial account is inspired by Alexius Meinong’s theory of objects and Terence Parsons’ neo-Meinongian account of nonexistent objects (Parsons 1980).5 To illustrate the combinatorial account, suppose that one could write a list of all actual objects and next to each of them the set of the properties6 the object instantiates. Once we run out of actual objects on the first column, however, we can still continue the list of sets of properties by writing all sets of properties that are not coinstantiated by any actual object. For example, since no actual mountain is made of gold, the properties goldness and mountainhood are not members of the set of properties of any actual object. According to the combinatorial account, to each set of properties that are not coinstantiated by any actual object corresponds one and only one possible object. So, we can continue our list by adding one possible but not actual object for each set of properties that are not coinstantiated by any actual object. It is among these objects that, according to the combinatorial account, the referent of ‘Sherlock Holmes’ can be found. Since no actual objects instantiates the properties of, say, being a detective and living at 221B Baker Street, according to the combinatorial account, there is a possible object that instantiates those properties. So, ‘Sherlock Holmes’ refers to a non-actual object whose set of properties includes the property of being a detective and living in 221B Baker Street. On this account, (1) is true because being a pipe-smoker is among the properties of that object and (4) is false because being an astronaut is not.

5

Neither Meinong nor Parsons however seem to accept the combinatorial account as outlined here

(or any other of the accounts I consider here for what matters.) Both Meinong and Parsons seem to believe that non-existent objects have no reality whatsoever and yet they can be referred to. If this interperetation is correct, they both reject the Axiom of Existence altogether and their position cannot be identified with any of the positions I consider here. 6

In presenting the combinatorial account, I roughly follow Terence Parsons’ presentation of his

own account. Parsons specifies that the properties must be nuclear where, intuitively, being red or being 6 meters tall are examples of nuclear properties and being existent and being possible are examples of non-nuclear ones. This distinction is of paramount importance for the combinatorial account, but we will not need to clarify it further in this context.

7

The combinatorial account is riddled with problems. I will consider only a few here.7 The first is that of identifying the object to which ‘Sherlock Holmes’ refers within the multitude of merely possible objects. The advocate of the combinational account would probably suggest that it is possible to make a list of all the properties attributed to Sherlock Holmes by his author and then find the object on the list has those properties. This suggestion however is not satisfactory. Suppose that the set H includes all the properties that Conan Doyle attributes to Holmes in his works. Many objects on our list will have H as a subset of the set of their properties. Some of them have a mole on the left shoulders, while others do not. Some are allergic to dust and others are not. Which one of them is Sherlock Holmes? The description of Sherlock Holmes that can be gathered from Conan Doyle’s works does not single out any one of them as the unique referent of ‘Sherlock Holmes.’ The advocate of the combinatorial principle, however, could stipulate that ‘Sherlock Holmes’ refers to the object that has all and only the properties attributed to it by his author.8 So, ‘Sherlock Holmes’ refers to the object on our list that has H as its set of properties. This suggestion, however, generates even more problems. Consider two here. First of all, we would tend to think that, say, Sherlock Holmes has a mother even if Conan Doyle never mentioned that. Second, if we are to explain that Sherlock Holmes is not an astronaut by pointing out that being an astronaut is not included among the properties in H, the fact that having a mole on the left shoulder is not included in H either would seem to force us to conclude that Sherlock Holmes does not have a mole on the left shoulder, which is unwarranted given that it is perfectly compatible with Doyle stories that he has one. Another problem is that possibilist accounts cannot account for the truth of external sentences. (2) does not seem to be true because ‘Sherlock Holmes’ refers to a possible object among whose set of properties, alongside that of being a detective and that of being a pipe-smoker, includes is that of having been created by Conan Doyle. This last property of Sherlock Holmes just does not seem to be on the same level as the other two. If ‘Sherlock Holmes’ referred to the possible but not-actual object that coinstantated all and only the properties attributed to Sherlock Holmes by Conan Doyle, then he would not instantiate the property of being created by Conan Doyle because this property is not a property Sherlock was never attributed to it by Conan Doyle—it is an actual property of Sherlock Holmes. To put it more vividly, if someone fitting Conan Doyle’s description of Sherlock Holmes had existed, he would be a brilliant detective who smokes a pipe and lives in Baker Street, but he would have not been created by Conan Doyle. An even more serious problem is that, as Saul Kripke (1980) has convincingly argued, there is simply no real or counterfactual situation in which some real person or other could have been Sherlock Holmes. It is not sufficient for something to fir the description of Sherlcok Holmes in order for it to be Sherlock Holmes (in fact, it is not even necessary). Even if, unbeknownst to Conan Doyle and the rest of us, there was someone who actually had all the properties attributed to Holmes in Conan Doyle’s

7

Similar objections to the combinatorial account have been put forward by Daniel Hunter (1981). A

different set of problems for the combinatorial account has been presented by Barry Smith (1980). 8

This seems to be the position actually held by Parsons.

8

works, including being called ‘Sherlock Holmes’, ‘Sherlock Holmes’ in (1) would still not refer to that person. Therefore, having all the properties which are attributed to Sherlock Holmes is not sufficient to be the referent of ‘Sherlock Holmes’ in (1). If the combinatorial account is afflicted by many problems, the other version of the possibilist account that I will consider here, the possible world account inspired by David Lewis’s account of truth in fiction (Lewis 1978), does not fare much better— and for similar reasons. According to the possible world account, there is a possible world in which Conan Doyle’s stories are literally true. In fact, since there is more of one way in which Conan Doyle’s stories could be true of a world (as there is much that those stories leave unspecified), Conan Doyle’s stories do not identify one but many possible worlds. Now, in each of these worlds, there is a brilliant detective whom the inhabitants of that world refer to as ‘Sherlock Holmes,’ who smokes a pipe and who lives in Baker Street. According to the possible world account, ‘Sherlock Holmes’ in the actual world refers to the person who is referred to as ‘Sherlock Holmes’ in one of these fictional worlds. The possible world account is afflicted by problems that are very similar to those that characterized the combinatorial account. The first is that there are a number of different worlds that are compatible with Conan Doyle’s stories and in each of those world there is a brilliant detective whose name is ‘Sherlock Holmes’. In some of these worlds, Sherlock Holmes has a mole on his left shoulders in others not. In some of those worlds, unbeknownst to Dr. Watson, he is an astronaut, in others, not. If we were to take seriously the claim that ‘Sherlock Holmes’ actually refers to one of the Sherlock Holmes that inhabit these possible worlds, the advocate of the modal account should be able to determine which one. However, since the description of Sherlock Holmes in Conan Doyle’s stories is incomplete, none of these identifications seems to be more warranted than any other. Moreover, unless one manage to rule out worlds in which Sherlock Holmes, unbeknownst to the narrator of the stories, Dr. Watson, participates to a secret space program, the modal account cannot rule out that (4) is after all true. Another problem is that the possible world accounts cannot account for the truth of external sentences. It does not seem possible to maintain that (2) is true because ‘Sherlock Holmes’ refers to a possible object among whose properties is that of having been created by Conan Doyle. In fact, on most accounts of the nature of possible worlds, it would seem that, if ‘Sherlock Holmes’ in (2) refers to the inhabitant of a possible world, he has not been created by Conan Doyle but exists independently of him. An even more serious problem is the one raised by Kripke. It is not sufficient to have all the properties which are attributed to Sherlock Holmes in order to be the referent of ‘Sherlock Holmes’ in (1). III.2. Abstractist Accounts According to abstractist accounts, insofar as the expression ‘Sherlock Holmes’ refers at all, it refers to an abstract entity: a character in a series of works of fiction. Thus, even if Sherlock Holmes is not an actual person and does not actually smoke a pipe, he (or it?) exists. It is one of the best-known characters of detective fiction, who was created by Arthur Conan Doyle and appeared for the first time in the 1887 novel A Study in Scarlet. Accordingly, we should distinguish two uses of the expression ‘Sherlock Holmes’. The first is the object-fictional use. This is the use that Conan Doyle makes of the name within the context of a work of fiction. In this use, the name, which appears to 9

refer to an extraordinary detective in Victorian London, simply fails to refer to anything. The second use is the meta-fictional use. This is the use that Conan Doyle and his readers make when talking about the leading character in many of Conan Doyle’s works. In this use, the name ‘Sherlock Holmes’ refers to an abstract object (a fictional character) that is part of another abstract artefact (a novel or a short story). The abstract account succeeds where all of the other accounts I have examined so far have failed: it manages to explain reference to imaginary objects in external contexts such as (2). In that sentence, the expression ‘Sherlock Holmes’ refers to an actual abstract entity and attributes to that entity a genuine (abstract) property, that of having been created by Arthur Conan Doyle. However, the standard abstract account does not seem equally successful in dealing with those cases in which expressions appear to referring to imaginary objects occur in an internal context outside of a work of fiction such as (1) and (4). In fact, the expression ‘Sherlock Holmes’ fails to refer to anything and the whole sentence is either false or fails to have any truth-value. Whereas Conan Doyle could not possibly misdescribe Holmes within the context of the stories he wrote, however, there seems to a sense in which someone would be misdescribing Holmes if she uttered (4). Note that she would not be misdescribing him because he is actually an abstract entity and abstract entities cannot be astronauts, but because, as I have already noted, there seems to be a sense in which it seems to be true that Sherlock Holmes is a detective. To avoid this problem various strategies have been proposed. I will consider two here. The first, which has been proposed by van Inwagen (1977), claims that sentences such as (1) should be paraphrased as: (7) There exists place in a work of fiction where the property being a pipe smoker is ascribed to Sherlock Holmes.

In other words, when an expression referring to an imaginary object appears in an internal context, the sentence can be paraphrased as saying that a ternary relation, ascription, holds between an abstract object, a property and a passage in a work of fiction. Van Inwaagen’s account faces two main problems. The first is that there seem to be properties that are not attributed to imaginary objects in any passage of a work of fiction but that nonetheless seem to be, in some sense, true of the object. Suppose that somewhere in Conan Doyle’s stories one can find the sentence: (8) Sherlock Holmes lives in 221b Baker Street,

Suppose also that the following sentence is nowhere to be found in those stories: (9) Sherlock Holmes lives closer to Regent’s Park than to Hyde Park.

Given the topography of Victorian London, (8) implies (9). However, whereas the schema underlying (7) can be used to account for our intuition that (8) is in some sense true, it cannot be used to account for our intuition that (9) is in some sense true as well. In fact, we have assumed that nowhere in Conan Doyle’s stories Holmes is ascribed the property of living closer to Regent’s Park than to Hyde Park. In general, fictional characters seem to have more properties than the ones which are attributed to them in the fictional works of which they are part. Literary critics and readers sometimes engage in passionate discussions about whether Hamlet is clinically depressed or whether or not Holmes is a misogynist, even if these properties are not attributed to them by their authors. 10

These opinions are only partly based on the properties which these characters are explicitly ascribed in the works of fiction of which they are part. They are also based on the readers’ background assumptions about what other properties these characters would have, if they actually had the properties they are ascribed. To the extent to which inferences such as the one from (8) to (9) are legitimate (and I think that, to some extent, they are), the qualified truth of (9) is to be explained as much as that of (8). A second problem is that, according to van Inwagen’s account, the expression ‘Sherlock Holmes’ does not refer to anything when it occurs within a work of fiction. As Nathan Salmon (1998) noted, strictly speaking, nothing can be attributed the property of being a pipe smoker by the occurrence of (1) in the context of Conan Doyle’s stories. Since in the sentences that occur in those stories, the expression ‘Sherlock Holmes’ fails to refer to anything and, a fortiori, it does not refer to the abstract entity Sherlock Holmes. The second version of the abstractist account that I will consider here is the one outlined by Salmon (1998). On this Salmon’s account, ‘Sherlock Holmes’ refers to the same object, an abstract entity, in both object-fictional and meta-fictional sentences. Since obviously abstract entities do not smoke, (1) is false. Nevertheless, Conan Doyle and his readers pretend that it is true by pretending that the abstract object to which we refer as ‘Sherlock Holmes’ smokes a pipe. It is important to note that this version of the abstract account differs from the pretence account in one crucial detail. According to the former, the speaker is genuinely referring to something (an abstract object), while according to the latter the speaker is only pretending to refer to something while referring to nothing. Salmon’s account has the advantage of avoiding some of the problems of the other abstractists accounts by eliminating the asymmetry in the analysis of object-fictional and meta-fictional sentences that characterises other abstractist accounts. According to this version. the expression ‘Sherlock Holmes’ refers to the same abstract object in both its object-fictional and meta-fictional uses. But, whenever the expression occurs in external sentences, the speaker pretends that that object is a person. Against Salmon’s account, Sarah Sawyer (2002) has argued that it is not clear what the pretence involved in it amounts to. Sawyer claims that if this pretence amounts to pretending that the ‘Sherlock Holmes’ refers to a real man rather than an abstract object, then the abstract object to which the expression actually refers becomes explanatorily redundant and the account incurs in (some of) the problems that afflicted the pretence account. If, on the other hand, the pretence amounts to pretending that an abstract entity is a real person, according to Sawyer, it is not clear how or even whether we can do so. Whereas it is clear how one can pretend to be on a sunny beach when they are not, claims Sawyer, it is not clear how one can pretend that the number two being a man who plays croquet (at least, if we consider the number two to be an abstract entity). And, since Sherlock Holmes is an abstract object, the kind of pretence involved in Salmon’s account seems more similar to the second case of pretence than to the first case. Sawyer’s dilemma does not seem particularly serious. If the first interpretation of Salmon’s account was correct, the abstract object would definitely not be explanatorily redundant. In fact, it would be the object to which ‘Sherlock Holmes’ actually refers both in external and internal sentences (even if, when it occurs in internal sentences, we pretend it refers to something else). If the second interpretation was correct (as I tend to believe), it is not clear where the difficulty of pretending that 11

the number two is a man who plays croquet exactly lies. The difficulty noted by Sawyer may arise from the fact that entity referred to as ‘Sherlock Holmes’ cannot be a person in any possible world. Sawyer would be suggesting that it is at least doubtful whether one can pretend that something that is metaphysically impossible is the case. Since pretending that p seems to require pretending that ‘p’ is true, the difficulty noted by Sawyer would probably stem from the fact that it is not clear whether one can pretend that p, if p is metaphysically impossible. In fact, in the case of a logical impossibility, it seems outright impossible to pretend that, say, a hula-hoop is both round and not round. However, the same does not seem to apply to the case of metaphysical impossibility. People seem to be able to pretend that what is metaphysically impossible is the case. For example, even if no dog could have been a horse in any metaphysically possible world, children sometimes pretend that their dog is a horse. However, Salmon’s account is not immune from problems. For example, the account seems to imply that any referring expression that occur in an object-fictional sentence and does not refer to any concrete actual object refers to an abstract object. Thus, whereas ‘London’ and ‘Scotland Yard’ in Conan Doyle’s stories refer, respectively, to the city of London and to its police headquarters, ‘Sherlock Holmes’ brother’ and ‘Sherlock Holmes violin’ refer to two abstract objects. Of the first one, we pretend that it is the brother of the abstract object to whom ‘Sherlock Holmes’ refers (even if abstract objects do not have brothers); of the second, we pretend that it is the violin which belongs to the abstract object to whom ‘Sherlock Holmes’ refers (even if abstract objects do not own violins). Now, suppose that the following sentence occurs in one of Conan Doyle’s stories: (10) Sherlock Holmes shook his head

If ‘his’ in (10) means ‘Sherlock Holmes’s’ as it would seem, then, according to Salmon’s account, ‘his head’ in (10) must refer to an abstract object, for Sherlock Holmes is an abstract entity and, as such, has no head.9 More precisely, ‘his head’ must refer to an abstract object that we pretend to be the head of the abstract object ‘Sherlock Holmes’ refers to. If this was the case, however, Salmon’s account would have some bizarre consequences. For one thing, Salmon would have to admit that Sherlock Holmes and Sherlock Holmes’ head are two distinct abstract objects and that, in writing his stories, Conan Doyle created two distinct abstract artefacts: Sherlock Holmes and Sherlock Holmes’ head. An advocate of Salmon’s account might reply that this analysis misrepresents their view. According to Salmon’s account, we pretend that Sherlock Holmes is a person and, in so doing, we pretend it has a head. Thus, in (10), ‘his head’ refers to the head that we pretend Sherlock Holmes has. In other words, we refer to his “pretended” head. However, either Sherlock Holmes’ “pretended” head exists or does not. If it does not exist, then the advocate of the pretence account could not refer to it without being in breach of the Axiom of Existence. If it is a real object, it is either an abstract object like Sherlock Holmes or it is a different kind of object. If it is an abstract object, then the view gives rise to the bizarre consequences I have mentioned above. If it is

9

A similar problem about creationism in fiction in general has been inspired by Takashi Yagisawa.

As I will argue, I think that creationism in fiction does not necessarily fall prey of this problem.

12

some other kind of real object, the advocates of Salmon’s account still owe us an account of what kind of object the head of a fictional character is. III.3. The Dualist account Both the view that imaginary objects are possible concrete objects and the view that they are actual abstract objects seem to capture some of our intuitions about imaginary objects. However, neither of them seems to be entirely satisfactory. It is interesting to note that the two views seem to some extent complementary—one view seems to be successful where the other fails and vice versa. The view that imaginary objects are actual abstract objects seem to be successful in accounting for our intuitions that (some) external sentences are literally true and that (all) internal sentences are literally false. However, it does not seem to be able to accommodate the intuitions that (some) internal sentences are nevertheless “in some sense” true. The view that imaginary objects are possible concrete objects, on the other hand, seems to be partially successful in accounting for the fact that some internal sentences are “in some sense” true. However, it seems to take those sentences too seriously for, on that view, those internal sentences that are true are literally true not just true “in some sense”. Moreover, on that view, it is not clear how to vindicate the intuition that (some) external sentences are literally true. This complementarity, I suspect, stems from the fact that both accounts fail to recognise the “dual” nature of imaginary objects and attempt to fully reduce them to either abstract or possible objects. In this section, I will argue that it is possible to defend an account that explicitly acknowledges the peculiar dual nature of imaginary objects and, therefore, combines the main advantages of possibilist and abstractist accounts without sharing their respective problems. I will call this account the dualist account. According to the dualist account, an imaginary object is an abstract object that stands for a possible object. ‘Sherlock Holmes’ refers to an abstract object: a fictional character that has real properties such as that of having been created by Arthur Conan Doyle in the late nineteenth century. This actual abstract object, however, stands for one of the many merely possible objects that are compatible with the description of Sherlock Holmes in Conan Doyle’s novels: an exceptionally brilliant detective who, among other things, lives at 221b Baker Street and smokes a pipe. Sherlock Holmes, thus, is not one of the possible objects that have all the properties that are attributed to him in Conan Doyle’s stories but only stands for one of them. According to the dualist account, external sentences such as (2) are literally true because ‘Sherlock Holmes’ refers to an abstract object that was actually created by Conan Doyle at the end of the Nineteenth century. Internal sentences like (1), on the other hand, are all literally false because ‘Sherlock Holmes’ refers to an abstract object and abstract objects do not smoke pipes nor do they have any other concrete property. Nevertheless, the abstract object Sherlock Holmes acts as a stand-in for one of the possible objects who, among other things, are detectives, smoke a pipe, and live at 221B Baker Street. So, even if (1) is literally false, we usually consider it to be in some sense true—true “by proxy” so to speak. (This is not unlike when we talk of an actor as if he was the character he plays in a movie. Although our assertions are literally false of the actor they are in some sense true because they are true for the character the actor plays in the movie.) Analogously, we consider some of them false “by proxy” as well (not only literally false) if they are false of all the possible objects that the fictional character Sherlock Holmes stands for. Finally, we consider them 13

neither true nor false (in the non-literal sense) if they are true of some of the objects Sherlock Holmes stands for and false of the others. Three remarks are in order here. The first two remarks concern ontological economy. Admittedly, the dualist account is ontologically inflationary for it requires that we include both abstract objects and possible objects in our ontology. However, this, in and of itself, is not a reason to reject the dualist account. We are not supposed to accept the dualist account because of it is more ontologically parsimonious than the other accounts. We are supposed to accept it because, unlike the other accounts, it is descriptively adequate—it vindicates a large number of intuitions that seem to underlie the way we talk and think of imaginary objects. Ockam’s razor urges us not to postulate entities unless they are indispensable. So, if there was an account of imaginary objects that was as descriptively adequate as the stand in account but more austere ontologically, I think we should prefer it to the dualist account. In lack of such an account, however, we have to accept the dualist account with all its ontological baggage. The question of ontological economy should be raised only when two equally descriptively adequate account are available. The second remark is that the ontological baggage that comes with the dualist account may be less heavy than it could seem at first. In fact, the dualist account does not commit us to any specific view about abstract and possible objects and most philosophers agree that, since our language seems to be committed to both categories of objects, we need to have some account of talk of abstract and concrete objects. I think that, insofar as one has some descriptively adequate way of accounting for the ordinary talk of abstract and possible objects, they will be able to adopt the dualist account. The third remark concerns the standing-for relation. Here, I only want to note that the relation that holds between the abstract object that is the model and the possible object for which it stands is not some mysterious relation. Most philosophers accept that some objects stand for other objects. For example, the name ‘Julius Cesar’ stands for Julius Cesar, and that this blue area on this map stands for Lake Ontario, and, that when we count five objects on fingers, each finger stands for one of those objects. The relation that holds between the abstract and the possible object is just another instance of this familiar stand-for relation. If we had a satisfactory philosophical account of it in all these other cases, we would have an account that applies to our case. As I have mentioned, I take it that the best argument in favour of the dualist account is that it combines the respective advantages of possibilist and abstractist accounts without sharing their respective disadvantages. Consider the problems of the possibilist account first. According to the dualist account, ‘Sherlock Holmes’ refers to an abstract object that actually exists. It is therefore metaphysically impossible for it to be an actual person in some other possible world. This is not to deny that there are be possible words in which there is a brilliant detective who lives in Baker Streets, etc. and whom the inhabitants of that world refer to as ‘Sherlock Holmes.’ It is only to deny that that possible person is Sherlock Holmes (where ‘Sherlock Holmes’ is used as a rigid designator and refers to the same object in all possible worlds). In that possible world, Sherlock Holmes (where ‘Sherlock Holmes’ is used as a rigid designator and refers to the same object in all possible worlds) simply does not exist—it is a possible abstract object. The dualist account is not even affected by the various problems arising from the identification of a specific imaginary object among the possible ones. In fact, Sherlock Holmes is not a possible object—it is an abstract object that stands for one of the 14

many possible objects that fit Conan Doyle’s description of Sherlock Holmes. Thus, in order to determine what the referent of ‘Sherlock Holmes’ is, we do not need to exactly identify the possible object for which ‘Sherlock Holmes’ stands for insofar as it stands for one or other of the possible objects that have all the properties that Holmes is attributed in Conan Doyle’s stories. Since the description of this object that can be gathered from Conan Doyle’s story is necessarily incomplete, it will be always to a certain extent underdetermined exactly for which possible object Sherlock Holmes stands. But the fact that we cannot determine exactly which possible object Sherlock Holmes stands for does not seem to be a problem insofar as we can identify the abstract object denoted by ‘Sherlock Holmes’ (i.e. one of the fictional character introduced by Arthur in A Study in Scarlet in 1887). Consider now the objections against the two versions of the abstractist account, which I have discussed in section III.2. The first problem with the van Inwagen’s account is that it cannot account for the fact that imaginary objects seem to have properties that are not ascribed to them in the works of fiction of which they are part. The dualist account however, does not seem to have troubles in accounting for this. In fact, if the topography of Victorian London is assumed to be true of the possible world that the fiction describes, then anyone who lives on Baker Street in that world lives closer to Regent’s Park than to Hyde Park. The dualist account also avoids the second problem faced by van Inwagen’s account. The expression ‘Sherlock Holmes’ refers to the same object in both metafictional and object-fictional sentences and in both internal and external contexts, namely an abstract object. However, in internal contexts, this object acts as a stand-in for a possible object. Those sentences in a work of fiction which ascribe concrete properties to a fictional character are thus to be interpreted, within that context, as descriptions of the possible object for which the fictional character stands and not as directly ascribing those properties to an abstract object. In other words, sentences like (1) and (8) are false if interpreted as sentences about the abstract object but are true of the possible object for which the abstract object stands. The dualist account is also immune to the difficulty that faced Salmon’s account. In fact, according to the dualist account, ‘his head’ in (10) does not refer to Sherlock Holmes’s head but to the head of (one of) the possible object(s) for which Sherlock Holmes stands. The dualist account, which I have sketched in this paper still needs to be developed as there are some phenomena (such as the possibility of incoherent fictions) that need to be accounted for. However, the dualist account has already an advantage over its rivals—it succeeds in accounting for a set of phenomena that are not jointly accounted for by any of them—and as such is a very plausible candidate for an account of imaginary objects. References Crittenden, C. (1991). Unreality: The Metaphysics of Fictional Objects, Ithaca, Cornell University Press. Evans, G. (1982). The Varieties of Reference, Oxford, Clarendon Press. Hunter, D. (1981). ‘Reference and Meinongian Objects’, Grazer Philosophischen Studien, 14, 23–36. Loux, M.J. (2002). Metephysics: A Contemporary Introduction, 2nd edition, London, Routledge. 15

Searle, J.R. (1969). Speech Acts: An Essay in the Philosophy of Language, Cambridge, Cambridge University Press. Smith, B. (1980) ‘Ingarden vs. Meinong on the Logic of Fiction’, Philosophy and Phenomenological Research, 41, 93–105. Parsons, T. (1981). Nonexistent Objects, New Haven: Yale University Press. Quine, W.V.O. (1948). ‘On What There Is’, Review of Metaphysics, Reprinted in W.V.O. Quine, From a Logical Point of View, New York, Harper & Row, 1953. Russell, B. (1905). ‘On Denoting’, Mind 14 (New Series), pp. 479–493. ______. (1957). ‘Mr Strawson on Referring’, Mind 66 (New Series), pp.385-389. Ryle, G. (1933). ‘Imaginary Objects’, Proceedings of the Aristotelian Society, (Supplementary Volume) 12, pp. 18–43. Salmon, N. (1998). ‘Nonexistence’, Noûs, 32, pp.277–319. Strawson, P.F. (1950) ‘On Referring’, Mind 59 (New Series), pp.320–344. Strawson, P.F. (1959) Individuals: An Essay in Descriptive Metaphysics, London, Methuen. van Inwagen, Peter (1977). ‘Creatures of Fiction’, American Philosophical Quarterly, 14, 299–308. Walton, K.L. (1973). ‘Pictures and Make Believe’, Philosophical Review, 82, 283–319. ______. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts, Cambridge, MA, Harvard University Press. Yagisawa, Y. (2001). ‘Against Creationism in Fiction’, Philosophical Perspectives, 15, 153– 172.

16