Mythical Mythology: Mythical Objects, Intentional Identity, and Referential Anti-Realism

Abstract

I defend a referential anti-realist solution to the problem of intentional identity. I develop Nathan Salmon's referential realist solution to the problem — according to which mythical objects exist and we can refer to them by using mythical-object names — and consider David Braun's objections to it. I argue that Salmon's solution yields the real identity, rather than the intentional identity, of the objects of multiple subjects’ thoughts. And I develop a referential anti-realist variant of Salmon's view — according to which mythical objects do not exist nor are they otherwise real but we can nevertheless refer to them — which avoids this worry.

Résumé

Je défends une solution référentielle anti-réaliste au problème de l'identité intentionnelle. Je développe la solution réaliste référentielle de Nathan Salmon — selon laquelle les objets mythiques existent et que nous pouvons y faire référence en utilisant des noms d'objets mythiques — et je considère les objections de David Braun à son égard. Je soutiens que la solution de Salmon donne l'identité réelle des objets des pensées de plusieurs sujets, plutôt que leur identité intentionnelle. Je développe enfin une variante référentielle anti-réaliste du point de vue de Salmon qui évite cette inquiétude — selon laquelle les objets mythiques n'existent pas et ne sont pas réels par ailleurs, bien que nous puissions néanmoins nous y référer.

1. Introduction

The central concern of this article is whether referential realism about mythical objects can resolve Peter Geach's problem of intentional identity. Geach's problem arises in cases in which propositional attitudes with a “common focus” are correctly attributed to multiple subjects when the object of those attitudes does not, or at least need not, exist (Geach, 1967). Both Nathan Salmon and David Braun endorse referential realism about mythical objects — the view that mythical objects exist and that we can refer to them by using mythical-object names (Braun, 2005; Salmon, 1998). But, whereas Salmon argues, in addition, that an adequate solution to Geach's problem can be developed within the referential realist framework, Braun denies that Salmon's solution will suffice. In this article, I argue, with Braun, that Salmon's solution is inadequate; but I also argue, contra Braun, that an adequate solution requires that the referential realist framework be rejected in favour of referential anti-realism — the view that that we can refer to mythical objects by using their corresponding names despite their non-existence.

This article consists of five parts. First, Geach's problem of intentional identity is introduced and the criteria for an adequate solution are delineated. Second, Salmon's referential realist solution to Geach's problem is developed and Braun's objections to it are considered. Third, I argue that even if Salmon could provide an adequate response to Braun's objection, his view does not meet the criteria for counting as a solution to Geach's problem. Fourth, I develop a referential anti-realist variant of Salmon's putative solution — supplemented by Ross Cameron's meta-ontology — and argue that it avoids the difficulties faced by Salmon's view. And, fifth, I consider a number of objections to the view developed here.

2. The Problem of Intentional Identity

Consider the following conjunction of attitude reports:

G: Hob thinks a witch has blighted Bob's mare, and Nob wonders whether she (the same witch) killed Cob's sow. (Geach, 1967, p. 628)

G could, as a matter of fact, be true — assuming that Hob and Nob exist — despite the fact that there are no witches. Moreover, as the second conjunct makes clear, there is a sense in which Hob and Nob are thinking about the same thing. After all, the pronoun “she” in,

C2: Nob wonders whether she (the same witch) killed Cob's sow,

prima facie refers to the witch mentioned in the first conjunct. But since the identity between the objects of Hob's and Nob's respective witch-attitudes holds despite the non-existence of witches, this identity is merely intentional. As Geach puts it: “[w]e have intentional identity when a number of people, or one person on different occasions, have attitudes with a common focus, whether or not there actually is something at that focus” (Geach, 1967, p. 627). The challenge is to come up with an analysis of G that is compatible with this intentional identity.

Given that the content of Hob's attitude in the first conjunct,

C1: Hob thinks a witch has blighted Bob's mare,

is expressed using the indefinite description “a witch,” any adequate analysis will presumably have to incorporate a corresponding quantifier. And the first question that needs to be addressed is whether this quantifier has wide or narrow scope with respect to the intentional operator “thinks.” First, consider a de dicto analysis according to which the quantifier has narrow scope and “she” in the second conjunct (C2) is an anaphoric substitute for “the witch who blighted Bob's mare”:

D: Hob thinks a witch has blighted Bob's mare, and Nob wonders whether the witch who blighted Bob's mare killed Cob's sow. (Geach, 1967, p. 630)

But Geach argues that D is inadequate because G could be true “even if Hob had not thought or said anything about Cob's sow nor Nob about Bob's mare” (Geach, 1967, p. 630). One might, of course, wonder what the basis would be for the judgement that Hob and Nob are thinking about the same thing under such circumstances or, more modestly, that their thoughts have a common focus. Following Braun, we might take having a common causal source to be sufficient: Hob and Nob might, for example, have both formed their witch-attitudes as a result of having independently read erroneous reports of a witch engaging in nefarious activities in the same local newspaper (Braun, 2012, p. 153).

Second, consider the following de re analyses according to which the quantifier has wide scope:

R1: There is a witch, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow.

R2: There is someone, x, such that Hob thinks x is a witch and has blighted Bob's mare, and Nob wonders whether x killed Cob's sow.

The trouble with R1 is that it presupposes the existence of witches and, as a result, unlike G cannot be true given the absence of witches. And the trouble with R2 is that it presupposes that Hob and Nob have a particular person in mind whom they suspect of witchcraft. But the truth of G is compatible with their not having identified any actual person as a suspect (Geach, 1967, p. 629). To sum up, an adequate solution the problem of intentional identity requires an analysis of G that satisfies the following criteria: (i) G could be true despite the non-existence of witches; and (ii) Hob and Nob are nevertheless “thinking about the same witch” in a sense that entails neither that they have an existent suspect in mind nor that Hob and Nob are even aware of one another.¹

Given these requirements for a solution to the problem, an analysis in terms of Gottlob Frege's distinction between sense and reference might seem promising.² On the Fregean picture, the sense expressed by a singular referring term plays three roles: it determines the referent of the term; it corresponds to a way of conceiving of the referent; and it serves as the referent of the term in attitude reports and other intensional contexts (Frege, 1997). A basic Fregean de re analysis of G might look something like the following:

F1: There is a sense, α, such that (i) Hob thinks <α is a witch who has blighted Bob's mare> and (ii) Nob thinks <α is a witch> and wonders whether <α killed Cob's sow>,

where the braces indicate that the semantic contents of expressions occurring therein are the senses expressed by them rather than their referents. What is important to note is that, since a sense need not have a referent, F1 entails that Hob's and Nob's respective thoughts have a common focus even if they do not have an existent suspect in mind. As a result, since in addition F1 entails neither the existence of witches nor that Hob and Nob are even aware of one another, the requirements of a solution to Geach's puzzle are satisfied. Salmon argues, however, that this basic Fregean analysis is inadequate because “Hob's and Nob's thoughts need not involve the same manner of specification” (Salmon, 2015, p. 115). The idea is that insofar as Hob and Nob conceive of the (putative) objects of their thoughts differently, there is no single Fregean sense that can serve as the common focus of those thoughts.

One response to this worry might be to endorse a more complex Fregean analysis along the following lines:

F2: There is a sense, α, and a sense, β, such that (i) α and β co-represent for Hob and Nob, (ii) Hob thinks <α is a witch who has blighted Bob's mare> and (iii) Nob thinks <β is a witch> and wonders whether <β killed Cob's sow>. (Salmon, 2015, p. 115)

On this analysis, even though there is no shared Fregean sense that serves as the common focus of Hob's and Nob's thoughts, these thoughts nevertheless have a common focus in virtue of the fact that the relevant constituents of their respective thoughts stand in a co-representation relationship. There are, however, a number of difficulties that arise for F2 as well. First, Salmon worries that it may prove difficult to come up with an account of the co-representation relation “so as to allow that a pair of individual representations α and β may co-represent for two thinkers without representing anything at all for either thinker” (Salmon (2015, p. 115). Second, Fregean accounts of referring expressions run into a multitude of serious problems, including Saul Kripke's well-known objections to descriptivism (Kripke, 1980). And, third, one might argue that F2 is simply too complex — with multiple quantifiers ranging over Fregean senses in addition to the co-representation relation — to be reasonably thought to yield the content of G.

3. Mythical Witches

Salmon develops and defends a solution to the problem of intentional identity that falls out of his more general views on singular reference (Salmon, 1987, 1998, 2015). Salmon endorses a Millian approach to proper names and demonstratives according to which their semantic contents are their referents (Salmon, 1998, p. 278). Moreover, he argues this is true whether or not such expressions occur within the scopes of propositional attitude verbs (Salmon, 1986). A central difficulty for Millian approaches to semantics is the problem of empty names, that is, names that lack referents. After all, insofar as the semantic content of a name is its referent, it follows that empty names lack any semantic content. As a result, sentences containing empty names arguably do not express propositions; and insofar as propositions are the primary bearers of truth, such sentences are arguably incapable of truth or falsity (Braun, 1993, pp. 451–453). Salmon's general approach to this problem is to argue that, appearances to the contrary, most putatively empty names, in fact, have referents. But, while in discourse about fiction and mythology he takes these referents to exist, in discourse about non-actual possibilities and the past, he argues that speakers can successfully refer to things that do not exist (nor are otherwise real) (Salmon, 1998).

For present purposes, I will focus on Salmon's account of mythical discourse. A myth, as Salmon uses the term, is an erroneous theory that at one point was believed to be true (Salmon, 1998, p. 304). Some myths postulate the existence of hypothetical objects. A mythical name is a name of one of these objects, that is, a name of a hypothetical object postulated by an erroneous theory once believed to be true. For example, Urbain Le Verrier postulated the existence of a planet — which he called “Vulcan” — between Mercury and the sun to explain anomalies in Mercury's orbit. Since Le Verrier's theory is erroneous, “Vulcan” is a mythical name. According to Salmon, by means of developing his erroneous theory, Le Verrier created a mythical object. According to the myth, this object is a planet — a concrete celestial body orbiting the sun. But, as a matter of fact, it is an abstract artifact, lacking any spatial location whatsoever, let alone an orbital path. And speakers who use the name “Vulcan,” including Le Verrier himself, refer to this mythical object (Salmon, 1998, p. 305).

Salmon includes witches among the mythical objects. They are the postulates of an erroneous theory designed to explain various phenomena including, perhaps, blighted mares and dead sows and, as such, are abstract artifacts in the same sense that Vulcan is. In light of this, Salmon offers the following de re analysis of G:

R3: There is a mythical witch, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow. (Salmon, 2005, p. 106)

As above, the goal here is a semantic analysis of G the meets the following conditions: (i) G could be true despite the non-existence of witches; and (ii) Hob and Nob are nevertheless “thinking about the same witch” in a sense that entails neither (a) that they have an existent suspect in mind nor (b) that Hob and Nob are even aware of one another. Condition (i) is met because, unlike R1, the truth of R3 does not require the existence of real witches but rather only the existence of mythical witches. Moreover, the sense in which Hob and Nob are thinking about the same witch is that they are thinking about the same mythical object. Condition (iib) is met because there is no reason to suppose that Hob and Nob have to be aware of one another to think about the same mythical witch; after all, they could independently rely on the same source as the basis of their respective witch-thoughts. And condition (iia) is met because there is no particular person that Hob and Nob both believe to be a witch but rather only a particular thing they both believe to be a witch. And, Salmon argues, this more modest claim is required by G: “[R3] does require something not unrelated to this, but no more than is actually required by [G]: that there be something that both Hob and Nob believe to be a witch — something, not someone […]” (Salmon, 2005, p. 106).

Braun has expressed some misgivings regarding Salmon's solution to the problem of intentional identity, however (Braun, 2012). Like Salmon, Braun endorses a Millian approach to the semantic contents of proper names and demonstratives (Braun, 1993, p. 450). And Braun does accept Salmon's metaphysical claim that Le Verrier created a mythical object by means of his mistaken theorizing (Braun, 2005, p. 615). Moreover, he concedes that at least some utterances of mythical names refer to mythical objects, although he denies that all of them do (Braun, 2005, pp. 615–618). But, despite his broad agreement with Salmon's approach to mythical names, Braun develops what he calls the “content objection” to Salmon's analysis of G (Braun, 2012, p. 157). According to this objection, in order for R3 to provide an adequate semantic analysis (of one reading) of G, it has to have the same semantic content as (this reading of) G. But R3 contains the expression “mythical witch” whereas G does not. As a result, insofar as “witch” differs in semantic content from “mythical witch” — and no other expression in G shares the content of the latter — it follows that R3 differs in semantic content from (any reading of) G. Braun concludes on this basis that Salmon's solution to the problem of intentional identity is inadequate (Braun, 2012, p. 157).

Salmon replies to this worry by endorsing the multiple ambiguity hypothesis regarding the term “witch.” On this view, the term has two distinct meanings: a witch₁ is a woman who has the ability to engage in supernatural witchcraft; and a witch₂ is something that is represented as being a witch₁. Taking a witch₃ to be something that is either a witch₁ or a witch₂, Salmon offers the following revised version of R3:

R4: There is a witch₃, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow. (Salmon, 2015, p. 125)

Insofar as the term “witch” is ambiguous between these various senses, Salmon can claim that G has the same semantic content as R4 on at least one of its readings. Braun, however, objects to the multiple ambiguity hypothesis on the grounds that it entails (i) that “There are witches” has true readings — when, for example, “witches” has the sense of witch₂ — and (ii) “Every witch is a witch” has false readings — when, for example, the first occurrence of “witch” has the sense of witch₂ and the second occurrence has the sense of witch₁ (Braun, 2012, p. 160). Salmon accepts these implications of his view, but denies that they count as evidence against it (Salmon, 2015, p. 125).

4. Millian Mythology

Rather than adjudicate this issue between Salmon and Braun, I want to focus on a more fundamental problem that arises for Salmon's referential realist account of intentional identity. For simplicity, I will focus on Salmon's original analysis of G:

R3: There is a mythical witch, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow.

The first thing to note is that not only does R3 commit us to the existence of mythical witches, but it also entails that there exists a particular mythical witch that is the object of Hob's and Nob's respective thoughts. And it is this latter entailment that undermines Salmon's analysis. Following Braun, we might formulate the circumstances under which G is uttered as follows (Braun, 2012, p. 153). A local newspaper contains reports of a person they call “Meg” engaging in supernatural witchcraft in the vicinity. Both Hob and Nob independently read these reports and, on their basis, (i) Hob sincerely says (out of earshot of Nob) “Meg is a witch and she blighted Bob's mare” and (ii) Nob sincerely asks (out of earshot of Hob) “Did Meg kill Cob's sow?” Finally, someone, who overhears both utterances, makes the following attitude report:

G: Hob thinks a witch has blighted Bob's mare, and Nob wonders whether she (the same witch) killed Cob's sow.

And, according to Salmon's analysis, G is true in such circumstances because Hob thinks Meg — an existent mythical witch — blighted Bob's mare, and Nob wonders whether Meg killed Cob's sow.

What is important to note, however, is that what makes this a case of intentional identity is not that G could be true despite the non-existence of witches, but rather that G could be true despite the non-existence of Meg. And it is for this reason that Salmon's analysis is inadequate. Suppose G were uttered in a world rife with witches — women capable of engaging in supernatural witchcraft — but the newspaper reports on the activity of the witch, Meg, are erroneous: there simply is no such witch. As above, G might nevertheless be true in such circumstances, even though its truth requires that Hob and Nob be thinking about the same witch in some substantial sense. But it would do no good to offer R1 as an analysis of G in such circumstances:

R1: There is a witch, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow.

After all, given the stipulation that Meg doesn't exist, none of the witches that do exist in our Meg-less witch world can count as a witch that both Hob and Nob are thinking about. And so R1 would be false in this world, not true. Similarly, insofar as G can be true in a world rife with mythical witches — including this one as Salmon would have it — but in which Meg does not exist, R3 cannot be offered as an analysis of G either. After all, none of the mythical witches that do exist in our Meg-less mythical witch world can count as something that both Hob and Nob are thinking about.

Salmon might reply by simply denying that there are any Meg-less mythical witch worlds. In particular, he might argue that in any worlds in which there are erroneous reports of a person called “Meg” engaging in supernatural witchcraft, those reports bring a mythical witch into existence that is the referent of said name (Salmon, 1998, p. 305). One might worry that this strategy risks overpopulating the world with mythical objects, that it entails that every bump in the night yields a new entity — a thing that goes bump in the night or, perhaps, that is falsely believed to do so. Salmon, however, seems to bite the bullet on this point: “[s]hould we not also admit and recognize such things as fabrications, figments of one's imagination, and flights of fancy as real abstract entities?” (Salmon, 1998, p. 305). One might, of course, reasonably contend that the idea of a world containing some number of mythical witches, along with a newspaper report describing the activities of a witch that does not refer to any of them, is perfectly intelligible. But, even if one concedes Salmon's point and grants that the newspaper report in question inevitably brings a mythical witch into existence, one might nevertheless resist the claim that the report's author or readers — including Hob and Nob — refer to this mythical witch by means of their use of the name “Meg.” Following Braun, for example, we might note that this mythical witch “does not satisfy (or even come close to satisfying) any reference-fixing description that [Hob or Nob] might have had in mind […]” (Braun, 2005, p. 615). Moreover, the causal relations that hold between this mythical witch and Hob's, Nob's, and the report's author's uses of Meg “do not resemble the causal relations that typically hold between objects and utterances of names that refer to them” (Braun, 2005, p. 616).

But a deeper worry about Salmon's analysis of G is that, by taking Hob and Nob to have thoughts about the same existent entity, it entails that G “express[es] (what the speaker took to be) the real, not the intentional, identity of a witch […]” (Geach, 1967, p. 628). As a result, Salmon's analysis does not even really address, let alone solve, the problem Geach identified. But, as noted above, Salmon claims that this implication is required by G: “[R3] does require something not unrelated to this, but no more than is actually required by [G]: that there be something that both Hob and Nob believe to be a witch — something, not someone […]” (Salmon, 2005, p. 106). In my view, Salmon is just wrong on this point. Consider again analysis R2:

R2: There is someone, x, such that Hob thinks x is a witch and has blighted Bob's mare, and Nob wonders whether x killed Cob's sow.

Now Salmon accepts Geach's rationale for rejecting this analysis: “[R2] does not provide a solution. Hob's and Nob's thoughts need not concern any real person. Maggoty Meg is not a real person, and there may be no one whom either Hob or Nob believe to be the wicked strega herself” (Salmon, 2015, p. 116). But Salmon is committed to the thesis that the following analysis avoids this objection:

R2*: There is something, x, such that Hob thinks x is a witch and has blighted Bob's mare, and Nob wonders whether x killed Cob's sow.

But the question is why this should be so. The first thing to note is that R2* follows from R2; after all, if there is someone I believe to be a witch, then there is something I believe to be a witch, at least insofar as the class of persons falls within the class of things. So, presumably, Salmon's view is that persons are the wrong sorts of things to falsely believe to be witches, in order to have an instance of intentional identity, while mythical witches are the right sort of thing (to falsely believe to be witches). But suppose Hob and Nob falsely believed a black cat, or a statue, or a shrubbery to be a witch rather than a person. The question is whether any of these cases would yield an instance of intentional identity, on Salmon's view, or whether only cases involving abstract artifacts, such as mythological objects, would suffice. Arguably, one would have to be more confused to believe an abstract artifact to be a witch than to believe the same of a person; but as the cases enumerated above illustrate, this is a difference of degree rather than kind. And, in any event, it is simply unclear why the extent of one's confusion is relevant to the question. Ultimately, unless and until Salmon can offer a principled rationale for thinking that believing abstract artifacts to be witches is consistent with intentional identity — despite the fact that believing the same of persons is not — his solution will remain unsatisfactory.

5. The Anti-Realist Turn

At this point, there seems to be no promising solutions to Geach's problem of intentional identity. As a result, one might be tempted to follow Braun in denying that the problem is soluble and claiming instead that there is no “reading of sentence (G) that can be true in a world without witches, and yet is true only if Hob and Nob are, in some sense, thinking of the same witch” (Braun, 2012, p. 149). But I argue here that a compelling solution can be found by rejecting Salmon's referential realism about mythical objects in favour of a referential anti-realist account of them. And what is interesting is that Salmon himself endorses referential anti-realism when it comes to past and possible objects, insisting on referential realism only in the case of fictional and mythical objects. This section will consist of three parts: a discussion of Salmon's referential anti-realism about past and possible objects; the application of this view to the mythical case; and a referential anti-realist solution to the problem of intentional identity.

Consider, first, past objects, that is, objects that used to exist but do not currently exist, such as Socrates. On Salmon's view, current uses of past-object names, such as “Socrates,” can be used to refer to such objects despite the fact that those objects do not exist: “There presently exists no one to whom the term ‘Socrates’, as a name for the philosopher who drank the hemlock, refers in English, but there did exist someone to whom the name now refers” (Salmon, 1998, p. 287). Moreover, according to Salmon, current utterances of sentences containing past-object names express propositions that do not exist: “Today the sentence ‘Socrates does not exist’ expresses Soc [the proposition that Socrates does not exist] with respect to the present time. It does not follow that there exists a proposition that this sentence expresses with respect to the present time” (Salmon, 1998, p. 287). And because such utterances express propositions, they are capable of truth or falsity: “The sentence ‘Socrates does not exist’ now expresses Soc [the proposition that Socrates does not exist], and Soc is now true. And that is why the sentence is now true in English (even though Soc does not now exist)” (Salmon, 1998, p. 287).

Salmon also endorses referential anti-realism about (certain) possible objects. Consider, for example, the possible object Noman “who would have developed from the union of S [a particular sperm cell of Salmon's father] and E [a particular ovum of Salmon's mother], if S had fertilized E in the normal manner […]” (Salmon, 1987, p. 50). According to Salmon, Noman does not exist, but speakers can nevertheless refer to him by means of their use of possible-object names such as “Noman”: “Noman is not something, and hence, even though ‘Noman’ refers to him, there is nothing that ‘Noman’ refers to” (Salmon, 1987, p. 94). And, as above, actual utterances of sentences containing possible-object names express propositions that do not exist and are, for that reason, capable of truth or falsity (Salmon, 1987, pp. 93–94). Finally, although it might be tempting to endorse a Meinongian interpretation of Salmon here — according to which past and possible objects are in some sense real despite not (actually currently) existing — Salmon explicitly rejects this view: “By contrast with Meinongians, I am not claiming that there are individuals that do not exist. […] What I am claiming is that there might have been individuals that do not actually₁ exist and that actually₁ have certain properties” (Salmon, 1987, p. 91).³

As noted above, Salmon endorses referential realism in the fictional and mythical cases, despite defending referential anti-realism regarding past and possible objects (Salmon, 1998, pp. 300–304). The central goal here is to argue that referential anti-realism about mythical objects yields a compelling solution to the problem of intentional identity and, in particular, one that avoids the difficulties that arose for Salmon's referential realist solution.⁴ First, to be an anti-realist about witches, as well as other mythical objects, is to deny they exist or are otherwise real — this is, of course, not to deny the existence of real people who are falsely believed to be witches. Meg, for example (from Braun's example discussed above), is in no sense real, according to this view. But second, according to referential anti-realism, speakers who utter the name “Meg” are nevertheless capable of referring to her. Moreover, third, because sentences containing the name “Meg” express propositions — albeit non-existent ones — they remain capable of truth or falsity. Consider the following instantiation of G:

G*: Hob thinks Meg is a witch and has blighted Bob's mare, and Nob wonders whether Meg killed Cob's sow.

What is important to note is that since Meg neither exists nor is otherwise real, on the referential anti-realist picture, but “Meg” can nevertheless be used to refer to her, simple analyses such as,

A1: Meg is a witch and Hob thinks that Meg blighted Bob's mare, and Nob wonders whether Meg killed Cob's sow,

or

A2: Hob thinks that Meg is a witch and has blighted Bob's mare, and Nob wonders whether Meg killed Cob's sow,

satisfy the criteria for an adequate solution to the problem of intentional identity.⁵ A1 and A2 could, after all, be true despite the fact that witches in general, and Meg in particular, do not exist; Hob and Nob are thinking about the same witch, namely Meg, the referent of “Meg”; this does not require that they have an existent suspect in mind, as Meg does not exist; and there is no reason to think that the truth of these statements requires that Hob and Nob even be aware of one another.

Let us return now to the analysis of G:

G: Hob thinks a witch has blighted Bob's mare, and Nob wonders whether she (the same witch) killed Cob's sow.

Given the analyses of G* considered above, the natural strategy would be to analyze G as the existential generalization of either A1,

R1: There is a witch, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow,

or A2,

R2: There is someone, x, such that Hob thinks x is a witch and has blighted Bob's mare, and Nob wonders whether x killed Cob's sow.

The trouble with these analyses, as noted above, is that, because they presuppose either the existence of witches (R1) or that Hob and Nob have a suspect in mind (R2), they fail to meet the criteria for a solution to the problem.

A solution to this difficulty, however, can be found in Cameron's meta-ontology (Cameron, 2013). According to Cameron, the ontological commitments of a statement are determined by the truth-makers for that statement — the conditions that must be satisfied in order for the statement to be true (Cameron, 2013, p. 181). In the ordinary case, the truth-makers for an existential statement — a statement to the effect that entities of certain kind exist — incorporate the entities in question. For example, the truth-maker for the statement,

There are bison in the prairies,

incorporates extant bison. In other cases, however, we may have grounds to accept the truth of an existential statement but reason to resist commitment to the existence of the entities this statement claims to exist (Cameron, 2013, p. 181).⁶ For example, one might take the truth of,

Bilbo Baggins is a hobbit,

to provide adequate grounds for the truth of the existential statement,

There are hobbits,

despite resisting any kind of commitment to existence of hobbits. Cameron's solution to this quandary is to suggest that the truth-makers for such statements need not incorporate the entities claimed to exist. So, for example, the truth-maker for,

There are witches,

need not consist of, or otherwise incorporate, witches. As a result, insofar as one's ontological commitments correspond to the truth-makers of the statements one accepts, one can accept the truth of existential statements about musical works, fictional entities, or mythical entities, including witches, without being saddled with suspect commitments (Cameron, 2013, p. 181).

The final proposal, then, is to accept either R1 or R2 as an analysis of G but to insist that they fall into the category of existential statements whose truth-makers do not incorporate the entities claimed to exist. After all, witches do not exist, and Meg does not either. Instead, what makes R1 and R2 true are facts about Hob's and Nob's respective psychological states and verbal (and other) behaviour, along with the reports, in the local newspaper, of a person they call “Meg” engaging in supernatural witchcraft and, perhaps, witch mythology more generally, among other things. And, as should be clear, both analyses, so understood, satisfy the criteria for a solution to Geach's puzzle of intentional identity. First, both analyses can be true despite the non-existence of witches because their truth-makers do not incorporate witches. Second, both Hob and Nob are thinking about the same witch: the non-existent referent of “Meg.” Third, none of this requires that they have an existent suspect in mind, given that Meg does not exist. And, fourth, there is no reason to think that truth-makers of R1 and R2, so understood, require that Hob and Nob be aware of one another.

6. Objections and Replies

There are, of course, a number of objections one might have to the proposal on offer, some of which will be considered here. First, one might argue that my appeal to Cameron's meta-ontology is misguided because, rather than defending a variant of referential anti-realism about fictional entities, Cameron himself instead defends a deflationary version of referential realism:

[T]he existence of the literary practice is sufficient for the existence of the fictional characters in question: as I would put it, that we appeal to something to do with the literary practice rather than the fictional character itself when specifying the truth maker for claims concerning the existence of, and properties of, the fictional being. (Cameron, 2013, pp. 183)⁷

What is important to note, however, is that I do not mean to claim that Cameron is defending referential anti-realism. Rather, what I claim is that Cameron's basic strategy can be used to defend this view. Moreover, I am presupposing that insofar as B's and not A's are the truth-makers for claims about A's, then the fact that “A's exist” is true in some contexts does not by itself establish realism about A's, even in some deflationary sense. Although I will not develop a general account of what condition — in addition to the truth of “A's exist” — is required for a theory to count as a kind of deflationary realism about A's, I take it for granted that if it is part of the theory that A's are merely imaginary, it does not so count. As will become clear below, the view on offer here counts as a version of mythical anti-realism for exactly this reason.

Second, given that I claim that the statement “There are witches” is true while at the same time denying that witches exist, it might be argued that position on offer here is incoherent.⁸ The first thing to note is that at the level of ontology these claims are compatible: to deny that there are witches is to deny that there are women capable of engaging in supernatural witchcraft, while to claim the statement “There are witches” is true, on my view, is to claim that the truth-makers for the statement — which do not include or consist of women with those capacities — exist. But, at the level of semantics, the view on offer implies that the statement “There are witches” is both true and false. One might attempt to avoid this apparent contradiction by taking there to be two senses of “exist” (or “there are”) or two senses of “witch.” But attempts to distinguish between two senses of “exist” fall prey to well-known difficulties.⁹ And not only does taking “witches” to be ambiguous risk running afoul of Braun's criticisms of the Salmon's “multiple ambiguity hypothesis” discussed above, taking it to mean something other than women capable of engaging in supernatural witchcraft — at least in the cases under consideration — undercuts the motivation to appeal to Cameron's meta-ontology in the first place.

A better strategy is to take the meaning of “There are witches” to be univocal, but to take the true and false instances of it to occur in different contexts of utterance. Consider, by way of analogy, utterances of the same sentence in distinct possible circumstances. If, for example, I utter the sentence “Justin Trudeau is the current Canadian prime minister” in the actual circumstances (in 2024), what I say is true; but if I were to utter the same sentence (at the same time) in possible circumstances in which Trudeau never went into politics, then what I say would instead be false, despite the fact that it means the same thing in both cases. But because the two occurrences of “Justin Trudeau is the current Canadian prime minister” are produced in different possible circumstances, the difference in their truth-values does not yield a contradiction. Similarly, insofar as the true instances of “There are witches” occur in a (suitably) different context than do the false instances, no contradiction will arise in this case either.

Let me elaborate. By a “mythical context,” I mean a context in which speakers and listeners pretend or imagine that mythical objects exist alongside real things.¹⁰ In the case at issue here, speakers imagine (or are under a prescription to do so) that there are witches — women capable of engaging in supernatural witchcraft — in addition to regular folk and other existent things. In a non-mythical context, in contrast, speakers do not pretend or imagine such things (nor are they under a prescription to do so). When “There are witches” is uttered in the mythical context in question, it is true: in that context, there do exist women capable of engaging in supernatural witchcraft. And when it is uttered in a non-mythical context, it is false: there are, in fact, no such women. But because the difference in truth-values occurs in different contexts, no contradiction arises. Since the truth-makers for statements of both mythical and non-mythical discourse are things that really exist, and witches do not really exist, the truth-maker for “There are witches” will have to incorporate such (non-mythical) things as the imaginative activities of the speaker (or the imaginative prescriptions they are under), the rules governing the mythical discourse at issue, witch mythology, and the like, rather than women capable of engaging in supernatural witchcraft. Moreover, given that, on the picture on offer here, witches are merely imaginary, this view counts as a version of mythical anti-realism rather than a deflationary variant of mythical realism.

Third, one might note that Jody Azzouni has also recently defended an anti-realist solution to the problem of intentional identity, and wonder (i) whether and how the view on offer here differs from his and (ii) whether there is any reason to prefer it to its competitor (Azzouni, 2013). At the core of the Azzouni's solution is the thesis that natural language quantifiers are ontologically neutral in the sense that they “are uttered or written regularly by people who don't think that as a result they are ontologically committed to the items talked about” (Azzouni, 2013, p. 344). And, although there are both ontological committal and non-committal uses of quantifiers, this does not entail that “the meaning of ‘there is’, for example, is ambiguous” (Azzouni, 2013, p. 343). Rather, Azzouni asserts that “onticity claims arise in language only contextually” (Azzouni, 2013, p. 344). In light of this, Azzouni argues that both

R1: There is a witch, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow,

and,

R2: There is someone, x, such that Hob thinks x is a witch and has blighted Bob's mare, and Nob wonders whether x killed Cob's sow,

yield adequate readings of G as long as the quantifiers are used non-committally.¹¹ After all, on both readings, G can be true despite the non-existence of witches and Hob and Nob can nevertheless be “thinking about the same witch” in a sense that entails neither that they have an existent suspect in mind nor that they are even aware of one another. And, to the worry that Hob and Nob cannot be thinking and talking about the same witch if there is nothing for them to be thinking and talking about, Azzouni also invokes a non-committal use of “about”: to say that Hob and Nob are thinking about the same thing in this sense does not ontologically commit one to the existence of the thing about which they are thinking.

As should be clear, the view on offer here differs significantly from Azzouni's theory. In particular, while Azzouni invokes two uses of quantifiers — committal and non-committal — my theory take takes them to be univocal but deployed in distinct contexts of utterance. In particular, while Azzouni argues that

R1: There is a witch, x, such that Hob thinks x has blighted Bob's mare, and Nob wonders whether x killed Cob's sow

yields an adequate reading of G only if the quantifier (and the noun-phrase “witch”) is used non-committally, on my view, what is required instead is that the quantifier be used in its univocal sense in a mythical context. Moreover, on Azzouni's view, Hob and Nob are “thinking about the same witch” only in an ontologically non-committal sense of “about,” but on my view Hob and Nob are both thinking about Meg in an ontologically committal sense. Of course, Meg does not exist, in fact, but is imagined to do so only within the mythical context in question; as John R. Searle puts it in the fictional case, “Sherlock Holmes does not exist at all, which is not to deny that he exists-in-fiction” (Searle, 1969, p. 79). But what is important is that, from the point of view of someone who has imaginatively entered into the corresponding “mythical world,” Meg does exist and, as a result, can serve as an object of thought and talk for someone occupying that point of view. Finally, both of these differences count in favour of my view and against Azzouni's alternative. First, even if the appeal to two uses of quantifiers rather than two meanings avoids the difficulties that beset the latter, the non-committal sense of the existential quantifier remains obscure. In particular, it is simply not clear what is being claimed when one uses “There are witches” non-committally. Insofar as such statements are capable of truth — as an adequate reading of G requires — one might wonder what the truth-makers for such claims are; and insofar as they have truth-makers that do not incorporate women capable of supernatural witchcraft, one might wonder to what extent Azzouni's solution at bottom differs from my own. And, second, it is similarly obscure what a non-committal use of “about” amounts to; that is, it is not clear what it means to say “Hob is thinking about Meg” in a non-committal sense of “about” and what would have to be the case for this claim to be true.

7. Conclusion

As we have seen, Salmon's solution to the problem of intentional identity fails because of his realism about mythical objects. Insofar as the sense in which Hob and Nob are “thinking about the same witch” is that they are thinking about the same existent mythical witch, Salmon's analysis counts as a case of the real identity, rather than the intentional identity, of the objects of Hob's and Nob's thoughts. And Salmon's claim that a solution to Geach's problem can be found by taking them to believe some existent thing, rather than some existent person, to be a witch is entirely unpersuasive. But, as argued above, a compelling solution can be found by reinterpreting the kind of analysis Salmon endorses along referential anti-realist lines. After all, insofar as Hob and Nob are thinking about the same witch in virtue of thinking about the same non-existent person, the identity between the objects of their respective thoughts is intentional and not real. Moreover, any suggestion that the analyses on offer presuppose a commitment to the existence of witches can be dispelled by appeal to Cameron's meta-ontology.

There are, of course, a number of questions that might be raised concerning referential anti-realism more generally. In particular, one might wonder what the mechanisms of reference to non-existent things might consist in. Although I will not defend a full theory of reference here, there are a few aspects of the view that are worth emphasizing. First, reference to non-existents is mediated reference in the sense that an utterance of a name refers to a non-existent thing in virtue of standing in the requisite causal-intention relation to an existent thing: just as an utterance of “Noman” refers to the non-existent but possible Noman in virtue of standing in the requisite causal-intentional relation to existent gametes S and E, an utterance of “Meg” refers to the non-existent but mythical Meg in virtue of standing in a similar causal-intentional relationship to the existent report in the local newspaper of a person bearing that name engaging in supernatural witchcraft. But, second, while in the case of reference to existent (and formerly existent) things, the mediator needs to also stand in some kind of causal-intentional relation to the referent, in the case of reference to mythical things, the mediator needs to stand in the requisite imaginative relation to the referent. In particular, the speaker needs to (imagine the referent exists and) imagine of the mediator — in the case at hand, the newspaper report of supernatural witchcraft — that it is suitably related to the (imagined) referent. Third, although a full defence of mythical referential anti-realism is beyond the scope of this article, it is worth noting that one measure of a theory lies in its ability to resolve philosophical puzzles.¹² As a result, the ability of this view to solve the problem of intentional identity counts strongly in its favour.