Fictional Creationism and Negative Existentials

Abstract

In this paper, I defend fictional creationism, the view that fictional objects are abstract artifacts, from the objection that the apparent truth of fictional negative existentials, such as “Sherlock Holmes does not exist,” poses a serious problem for creationism. I develop a sophisticated version of the pragmatic approach by focusing on the inconsistent referential intentions of ordinary speakers: the upshot would be that creationism is no worse—perhaps even in a better position—than anti-realism, even if we restrict our linguistic data to fictional negative existentals.

1. The Problem of Negative Existential for Fictional Creationism

Fictional creationism is the ontological view that fictional objects are abstract artifacts whose existence supervenes on our mental activities involving literary practices. ¹ The main motivation for creationism is that it can provide a straightforward explanation for our intuition about the literal truth of the following so-called external metafictional sentences, sentences that seem to be directly talking about fictional objects:

1. (1) Sherlock Holmes is a fictional character.

2. (2) Sherlock Holmes was created by Conan Doyle.

3. (3) There are some fictional characters inspired by Sherlock Holmes.

Like many creationists, in this paper, I will assume referentialism (or Millianism) about names, the semantic view that the semantic content of a name, if it has any, is simply its referent. Then, according to creationism, the semantic content of the fictional name “Sherlock Holmes” is the fictional character Sherlock Holmes brought into existence by Conan Doyle’s act of creation, and so all the above sentences (1)–(3) semantically express true singular propositions about the fictional character Sherlock Holmes. (1) is true because Sherlock Holmes has the property of being a fictional character. (2) is true because Sherlock Holmes has the property of having been created by Conan Doyle. Creationists’ taking (2) to be literally true aligns with our natural intuition that the existence of Sherlock Holmes is contingent on Conan Doyle’s act of creation. This provides a prima facie reason that favors creationism over other realist positions on fictional objects, such as Meinongianism and Platonism, which should paraphrase (2) to accommodate our intuition about its truth. ² (3) is true because the quantifier in (3) ranges over fictional characters, and some of them (e.g., Dr. Gregory House and Conan Edogawa) are inspired by Sherlock Holmes. These simple and straightforward explanations make creationism theoretically attractive.

Creationists, however, face the problem of negative existential. ³ Consider the following singular fictional negative existentials (“FNEs” for short): ⁴

1. (4) Sherlock Holmes does not exist.

2. (5) There is no Sherlock Holmes.

It seems that creationists should take (4) and (5) to be literally false: Sherlock Holmes does exist. But this conflicts with the apparent truth of (4) and (5). Therefore, the challenge for creationists is to provide a plausible and tenable explanation for the apparent truth of (4) and (5). ⁵

The aim of this paper is to address this challenge by analyzing the nature of FNEs. I will argue that FNEs do not pose any serious problem for creationism: In Section 2, I first argue that there is indeed a certain kind of an FNE that presents linguistic data requiring additional revisionary work for anti-realism but not for creationism, thus favoring creationism over anti-realism. In Section 3, I criticize contextualism about the existence predicate proposed in response to the problem of FNEs. In Sections 4 and 5, I defend my pragmatic approach by distinguishing between cases where speakers intend to refer to a fictional object by using a fictional name in an FNE and cases where they intend to refer to nothing. In Section 6, I diagnose the reason why solving the problem of FNEs is difficult: ordinary speakers’ uses of fictional names in FNEs are often inconsistent and confused.

2. The Literal Truth of Fictional Negative Existentials

In this section, I will first argue that even if we restrict our linguistic data to FNEs, creationism is not in a worse starting position than anti-realism. My aim here is not to provide a conclusive argument against anti-realism, but to provide one reason for considering creationism as theoretically preferable to anti-realism. To be specific, although both creationists and anti-realists need a revisionary, error-theoretic explanation for FNEs such as (4) and (5), when it comes to data related to FNEs in modal contexts, anti-realists require an additional different revisionary approach, while creationists can straightforwardly explain them.

Let us begin our discussion by noting that, according to creationism, fictional objects are contingent beings: although Sherlock Holmes does exist, it might not have existed if Conan Doyle had not created it. Now, consider the following (6) and (7), which contain a modal FNE:

1. (6) Contrary to what actually happened, there might have been no such thing as Sherlock Holmes. Indeed, Conan Doyle might not have written any novels at all. Fortunately, he wrote A Study in Scarlet, and the fictional character Sherlock Holmes came into existence.

2. (7) Contrary to what actually happened, the fictional character Sherlock Holmes might not have existed, for Conan Doyle might not have written any novels at all.

Ordinary speakers will naturally believe that (6) and (7) are literally true, and this intuition of literal truth is straightforwardly explained by creationism, which posits that Sherlock Holmes is a contingent being: Sherlock Holmes does not exist in the closest possible worlds where Conan Doyle did not write any novels.

In contrast, anti-realists cannot take (6) and (7) at face value because, according to anti-realism, the fact that there is no Sherlock Holmes is not contrary to what actually happened. They need to engage in a revisionary work to accommodate our intuition about the truth of (6) and (7). What is worse for anti-realists, however, is that the revisionary work for (6) and (7) becomes an additional one, distinct from their original revisionary explanation for bare FNEs (i.e., FNEs that do not involve any modal operator) such as (4) and (5). The explanations provided by anti-realists for bare FNEs—for example, Walton (1990), Crimmins (1998), and Everett (2013)—are revisionary: they suggest that, when ordinary speakers utter bare FNEs, they are engaging in some sort of make-believe game or pretense, ⁶ even though they seem to sincerely assert bare FNEs. But the same revisionary explanations cannot apply to the FNEs in (6) and (7).

To see why anti-realists require two different kinds of revisionary approaches for bare FNEs and modal FNEs, let us consider Everett’s view as an example. According to Everett (2013, pp. 72–73), utterances of (4) convey the information that actual singular representations that count as referring to Holmes within the scope of the pretense (“actual Holmes-representations” for short) fail to genuinely refer. ⁷ However, this analysis cannot be applied to modal FNEs such as “There might have been no such thing as Sherlock Holmes” in (6) and “Sherlock Holmes might not have existed” in (7). This is because if uttering those sentences conveys the information that it is possible that actual Holmes-representations fail to genuinely refer, (6) and (7) make no sense. According to anti-realism, the fact that actual Holmes-representations fail to genuinely refer is what actually happened, and its possibility is not due to the fact that Conan Doyle might not have written novels. Rather, because Conan Doyle wrote A Study in Scarlet, we came to have Holmes-representations that fail to genuinely refer. Likewise, Everett’s analysis cannot be applied to the sentence “Sherlock Holmes came into existence” in (6). Anti-realists, of course, cannot admit that the utterance of this sentence conveys the information that the actual Holmes-representations genuinely refer.

So, in order to explain the intuitive truth of (6) and (7), anti-realists should acknowledge that the FNEs in (6) and (7) convey substantially different information than bare FNEs like (4) and (5), implying that they require double revisionary works for FNEs. This observation makes anti-realism less theoretically appealing than it may initially seem, especially when considering that creationists do not need any revisionary work at all to account for the intuitive truth of (6) and (7).

Moreover, there is compelling linguistic evidence that the modal FNEs in (6) and (7) are literally true, in line with creationists’ expectations. Consider the following (8):

1. (8) Had Biden’s parents not met, the person Biden would not have been born. Had Doyle not written any novels, the fictional character Holmes would not have come into existence. So, contrary to what actually happened, both Biden and Holmes might not have existed.

Since the predicate “exist” in the last sentence applies to the plural subject “Biden and Holmes” and does not seem to be a case of syllepsis, as seen in sentences like “Both this color and this feather are light,” we have strong reason to believe that the sentence attributes to both Biden and Holmes the possibility of nonexistence in exactly the same sense. ⁸ Therefore, given that the last sentence in (8) semantically expresses the proposition about the possibility of Biden’s literal nonexistence, the same proposition is also about the possibility of Holmes’s literal nonexistence.

Indeed, the natural and straightforward creationist analysis of (8) is that (8) is literally true because both Biden and Holmes are contingent beings: their existence conditions might not have been satisfied. Of course, since Biden and Holmes belong to different species of objects, that is, one is a person and the other is a fictional character, their existence conditions would be different, and (8) describes two different possible situations where the existence conditions for each object are not satisfied.

Note, on the other hand, that the following sentence is infelicitous:

1. (9) Contrary to what actually happened, there might have been no such thing as the king of the U.S.

The straightforward explanation for the felt infelicity of (9) is that the non-existence of the king of the U.S. is not contrary to what actually happened: there is no such thing as the king of the U.S. in the actual world. Together with the linguistic datum from (8), then, this observation suggests that the ontological status of Sherlock Holmes is different from that of the king of the U.S.: while Biden and Holmes are contingent beings that exist in our actual world, the king of the U.S. does not exist at all, and the description “the king of the U.S.” fails to denote.

Again, I am not denying that anti-realists might find a way to explain away the intuition about the literal truth of (8). My point is simply that there are linguistic data on FNEs such as (6)–(8) that favors creationism over anti-realism. While anti-realists need an additional error-theoretic explanation for them, distinct from that for bare FNEs, creationists can provide a straightforward, non-revisionary explanation. ⁹ This constitutes one theoretically appealing aspect of creationism.

In this section, I have argued that creationism is not in a worse starting position—perhaps creationism is even in a better starting position—than anti-realism, even if we restrict our linguistic data to FNEs. However, having a better starting position for creationism, achieved by having a smaller number of revisionary works, does not guarantee that creationism is indeed the best theory. To establish its superiority, creationists must demonstrate that there is a plausible way to explain the apparent truth of bare FNEs like (4) and (5). ¹⁰ If it turns out that there is none, creationism will lose its theoretical appeal. Therefore, from the next section, I will explore how creationists can successfully address the problem of bare FNEs.

3. Against Contextualism about the Existence Predicate

In this section, before defending my pragmatic approach, I will first criticize the so-called contextualism about the existence predicate (“CEP” for short), a version of the semantic approach to explaining the apparent truth of FNEs. ¹¹ I describe this view as follows:

Contextualism about the Existence Predicate: A speaker S’s utterance of the singular negative existential “N does not exist” in a context C semantically expresses the proposition that N does not exist as a Kc (where a Kc is a salient kind in C). ¹²

For simplicity, I will treat CEP as a hidden indexical view about the existence predicate. But my criticism equally applies to other semantic approaches such as the ellipsis view, according to which when a speaker utters “N does not exist,” she elliptically utters “N does not exist as a Kc,” ¹³ and the modulation view (Spewak, 2016), according to which the semantic content of “exist” is determined by the pragmatic process of modulation.

Now the question is which ontological kinds can be assigned to “Kc.” Thomasson (1999, pp. 112–113) proposes that “Kc” is assigned a kind to which N in a relevant fiction in C belongs. In many ordinary contexts, for example, our utterances of (4) semantically express the proposition that Sherlock Holmes does not exist as a person. However, Everett (2007, pp. 65–71) offers a compelling criticism against Thomasson’s view. Suppose an author wrote a story about Yugo where it is indeterminate and unclear to which kind Yugo belongs; what we as readers can only know from the story is that Yugo is a fictional object. But even in this case, the following sentence seems to be true:

1. (10) Yugo does not exist.

Since there is no kind to which Yugo belongs in fiction, Thomasson’s view has a problem explaining the apparent truth of (10).

In response to Everett, von Solodkoff (2014) provides a sophisticated defense of CEP. ¹⁴ Her idea is that there is no reason to regard the value of “Kc” as restricted to a kind to which N belongs in fiction. Rather, it seems plausible that a kind like the concrete object, that is, a so-called ontologically fundamental kind of object, can also be a salient kind in most contexts. Indeed, assigning concrete object to “Kc” seems to predict the right result in most contexts. For example, in many ordinary contexts, our utterances of (4) will semantically express the true proposition that Sherlock Holmes does not exist as a concrete object. Similarly, the apparent truth of (10) is well explained because our utterances of (10) semantically express the true proposition that Yugo does not exist as a concrete object. Moreover, it also seems correct to predict that the utterance of (4) will sometimes semantically express a false proposition in contexts where the value of “Kc” is fixed to a kind that includes an abstract object or artifact, that is, in such a context where the utterances of (6)–(8) are true.

I think von Solodkoff’s approach sheds some light on the problem of FNEs. But her view still seems to face some serious problems that are not easily solved. In what follows, I will offer two criticisms against von Solodkoff’s version of CEP.

3.1. The problem of ordinary negative existentials

Note that CEP is a general claim about negative existentials: it is a claim that the existence predicate is context-sensitive. That is, CEP applies not only to FNEs, but also to ordinary negative existentials (“ONEs” for short), which include non-fictional ordinary names. Now consider the following ONE:

1. (11) Biden does not exist.

The intuitive falsity of (11) is well explained by von Solodkoff’s CEP: our utterances of (11) semantically express the false proposition that Biden does not exist as a concrete object in many ordinary contexts.

However, given that most creationists believe that abstract objects and concrete objects are two equivalent and exhaustive categories of objects, it does not seem easy for von Solodkoff to explain why there seems to be no context where our utterances of (11) semantically express the true proposition that Biden does not exist as an abstract object, where the value of “Kc” is fixed to abstract object. That is, if our utterances of FNEs including fictional names that refer to existing abstract objects, like (4), can express true propositions when uttered in a context where the value of “Kc” is concrete object, there seems to be no prima facie reason to deny that our utterances of ONEs including ordinary names that refer to existing concrete objects, such as (11), can express true propositions when uttered in a context where the value of “Kc” is abstract object. But, however, salient the kind abstract object may be in a particular conversational context, there seems to be no way to semantically express a true proposition by simply uttering (11).

Proponents of CEP might claim that every time we utter an ONE in a context where the kind abstract object is salient, our utterances themselves bring about the context shift, and the value of “Kc” comes to be fixed to concrete object. However, they should then provide a non-ad hoc explanation for why context shifts occur only when we utter an ONE, even if abstract and concrete objects are two equivalent categories of objects. I do not mean to suggest here that such an explanation is impossible. Still, given that the explanation contextualists must provide should not be something that anti-realists can exploit to defend their own position, it seems quite challenging for them to provide such an explanation.

3.2. The problem of plural subjects

Let us now turn to my other criticism, which I consider more serious and is based on Everett’s (2007, pp. 60–62) observation. Everett proposes the apparent truth of the following sentence as a linguistic datum that refutes CEP:

1. (12) Biden exists, but Sherlock Holmes and the round square do not.

The VP elision in the right conjunct suggests that (12) both affirms the existence of Biden and denies the existence of the round square in an ordinary sense, whatever that may be. But then we seem forced to admit that (12) also denies the existence of Sherlock Holmes in the same ordinary sense. Everett takes it that linguistic data like (12) point to a constant semantic content of the existence predicate across contexts. In what follows, I will defend Everett’s anti-contextualist view by showing why proponents of CEP cannot provide a non-ad hoc response to Everett’s criticism.

Note that von Solodkoff’s version of CEP seems immune to Everett’s original objection. In a context where the value of “Kc” is fixed to concrete object, the utterance of (12) semantically expresses the true proposition that Biden exists as a concrete object, but Sherlock Holmes does not exist as a concrete object and the round square does not exist as a concrete object. ¹⁵ Therefore, (12) affirms the existence of Biden and denies the existence of Sherlock Holmes and the round square in the same sense. ¹⁶

Consider, however, the following seemingly true sentence: ¹⁷

1. (13) Biden, the U.S., and the number 3 exist, but Sherlock Holmes, Wakanda, and the largest number do not.

In this case, von Solodkoff’s contextualist strategy for accommodating the intuitive truth of (12) does not work: it is not the case that the U.S. and the number 3 exist as concrete objects. Arguably, they are abstract objects (or at least there does not seem to be any fictional creationist who denies that they are abstract objects). Moreover, given that Wakanda is a fictional country, it does not seem that the utterance of (13) simply claims that Wakanda does not exist as a concrete object because the U.S. does not exist as a concrete object either. Rather, it seems to claim to the effect that Wakanda is not a real country.

Is there any kind other than concrete object for the value of “Kc” that can accommodate the intuitive truth of (13)? Von Solodkoff (2014, pp. 341–242) suggests a so-called serious object, defined as a non-pretend object: the existence of a serious object has nothing to do with any acts of pretense. Her suggestion seems to predict the right result: Biden, the U.S., and the number 3 exist as serious objects, whereas Sherlock Holmes, Wakanda, and the largest number do not exist as serious objects (i.e., each either does not exist at all or is not a serious object).

What then of so-called mythical objects like Vulcan? According to mythical creationism, Vulcan was inadvertently created by Le Verrier’s act of mistaken theorizing; ¹⁸ Vulcan is not a pretend object, as its existence has nothing to do with any acts of pretense. Thus, if the name “Vulcan” is added to the plural subject of the right conjunct of (13), it seems that the strategy of using serious object does not apply. Von Solodkoff, however, gets around this problem in the following way:

[E]ven if [the realist] denies that Vulcan is a pretend object, the realist could expand the category of non-serious objects to include any mythical object. (von Solodkoff, 2014, p. 344)

Nevertheless, I do not think that this move brings any substantial improvement. Expanding the category of non-serious objects to include mythical objects would define non-serious objects disjunctively as pretend or mythical objects. But this looks highly arbitrary and ad hoc.

To illustrate this point, let us consider so-called imaginary objects and hallucinatory objects. ¹⁹ Von Solodkoff’s maneuver makes it too easy for us to explain away the apparent truth of negative existentials regarding those objects. We could simply expand the category of non-serious objects to include imaginary objects and hallucinatory objects (or narrow down the category of serious objects so that it excludes them), thereby defining non-serious objects disjunctively as pretend, mythical, imaginary, or hallucinatory objects. Regardless of the kinds of objects we encounter that pose the problem of the apparent truth of negative existentials about those objects, expanding the category of non-serious objects to include them offers us a cheap solution. But it is obviously ad hoc.

What von Solodkoff should provide is a substantive explanation of what constitutes (non-) serious objects. However, what we can derive from her suggestion seems to be nothing but the following:

Serious Object: An object o is a non-serious object if and only if in most contexts ordinary people tend to utter “N does not exist” to express something true, even if a name “N” refers to o.

This clearly shows why the strategy of employing the concept of serious objects is ad hoc. Again, I do not intend to imply here that it is entirely impossible to provide a non-ad hoc, substantive explanation for serious objects. But it does seem to be quite challenging for her to provide such an explanation.

The discussion so far supports the anti-contextualist view that the existence predicate expresses a constant semantic content across contexts. (12) and (13) seem to affirm and deny the existence of various objects in exactly the same primitive sense of “exist.” Given the problems for CEP, which we have seen in this section, I believe we now have a motivation to explore an alternative approach in explaining the apparent truth of FNEs. In the next section, I will defend a sophisticated version of the pragmatic approach to FNEs that aligns with the view that the existence predicate is not a context-sensitive term.

4. The Pragmatic Approach to Fictional Negative Existentials

I propose the following version of the pragmatic approach to FNEs (“PAF” for short): ²⁰

Pragmatic Approach to FNEs: When a speaker S takes “N” to be a fictional name that refers to a fictional object, S’s utterance of the singular negative existential “N does not exist” (or “There is no N”) usually pragmatically conveys the descriptive proposition expressed by the sentence “N as a Kc does not exist” (or “There is no N as a Kc”), which results from substituting “N” with “N as a Kc” (where a Kc is a salient kind with respect to N in a context of utterance). ²¹

For simplicity, I will use “N as a Kc” as an abbreviation for the description “the thing that is both N and a Kc” (or “the Kc N” for short). For example, “Sherlock Holmes as a person” will be used as an abbreviation for the description “the thing that is both Sherlock Holmes and a person” (or “the person Sherlock Holmes”). ²² Now, according to my PAF, the relevant pragmatic mechanism typically comes into play when a speaker takes “N” to be a fictional name referring to a fictional object, because in this case, she generally intends to deny the existence of an eccentric object that is both a fictional character and a Kc, for example, Sherlock Holmes as a person (or as a concrete object). On the contrary, if the speaker believes that “Sherlock Holmes” is an ordinary name referring to a person who is already dead, she will generally intend to convey only the literal meaning of her utterance of “Sherlock Holmes does not (or no longer) exist.” ²³

This pragmatic mechanism, of course, needs further explanation. Let us first note that, according to creationism, the existence and persistence of fictional objects supervene on our mental activities. Thus, when a speaker takes “Sherlock Holmes” to be a fictional name and utters (4) or (5), if she succeeds in having singular thought about Sherlock Holmes and in using the name “Sherlock Holmes” to refer to Sherlock Holmes, then this very success of her mental activities implies the existence of Sherlock Holmes and also contributes to the persistence of Sherlock Holmes. Then it follows from this that her utterance of (4) or (5) cannot be literally true, if her use of “Sherlock Holmes” refers to Sherlock Holmes. ²⁴ If the hearer realizes this, he will think that the speaker’s real intention is not to convey the literal meaning, which is false, but to convey another proposition that the speaker believes to be true. Finally, the hearer succeeds in pragmatically grasping the true descriptive proposition that Sherlock Holmes as a person does not exist.

The same process, of course, does not take place with respect to concrete objects like Socrates, whose existence and persistence conditions are independent of our mental activities. Our thinking of or referring to Socrates has no influence on his survival or death: even if he presently no longer exists, we can think of him, use the name “Socrates” to refer to him, and sincerely assert that Socrates no longer exists, without any difficulty.

Our situation is rather comparable to the fact that there is arguably no ordinary context of utterance where a speaker’s utterance of the following sentence (14) is literally true; her successful utterance of (14) normally implies her existence. ^{25 , 26}

1. (14) I do not exist.

When we hear someone says (14), we almost always expect that the speaker, who is sensible enough, does not just intend to convey the literal meaning of her utterance. In most contexts, (14) is used to pragmatically convey something else. Similarly, creationists can naturally expect that, since there is no context where utterances of bare FNEs like (4) and (5) are true when referring is successful, these utterances are usually used to pragmatically convey something else. ^{27 , 28}

Now, the main advantage of PAF over contextualism is that PAF can account for our observation that the existence predicate expresses a constant semantic content across contexts, that is, the observation that our utterances of (12) and (13) seem to affirm and deny the existence of various objects in exactly the same primitive sense of “exist.” For example, according to PAF, if a speaker takes “Sherlock Holmes” and “Wakanda” to be fictional names referring, respectively, to the fictional character Sherlock Holmes and the fictional country Wakanda, then her utterance of (13) pragmatically conveys the proposition expressed by the following sentence (13’), which is obtained by substituting the names “Sherlock Holmes” and “Wakanda” in (13) with the description “Sherlock Holmes as a person” and “Wakanda as a country,” respectively:

(13’) Biden, the U.S., and the number 3 exist, but Sherlock Holmes as a person, Wakanda as a country, and the largest number do not.

For the sake of argument, let us adopt the orthodox Russellian analysis of a description. Then (13’) is equivalent to the sentence “There is (or exists) Biden, the U.S., and the number 3, but there is (or exists) no unique thing that is both Sherlock Holmes and a person, no unique thing that is both Wakanda and a country, and no unique thing that is a largest number.” That is, we affirm and deny the existence of various objects in the same primitive sense of “exist.”

In a similar vein, according to PAF, there is no need for any ad hoc work, to which CEP is committed, to identify a single “Kc” that appropriately applies to every subject in (13). It is sufficient to replace each name that the speaker takes to be fictional with a corresponding description to obtain the correct result. Even if the name “Vulcan” is added to the plural subject of the right conjunct of (13), as long as the speaker takes “Vulcan” to be a mythical name, she will intend to deny the existence of the eccentric object that is both a mythical object and a planet, namely Vulcan as a planet. Therefore, we can apply exactly the same approach as PAF here: simply replacing “Vulcan” with “Vulcan as a planet” leads to the correct result. ²⁹

At this point, it should be noted that PAF is an approach to illuminate a general, but not every, mechanism for pragmatically conveying something true through the utterance of an FNE. I do not deny that there is a case where a speaker uses the existence predicate “exist” to pragmatically convey <existing as a Kc>. ³⁰ Indeed, there seems to be no reason to deny this, given that the sentence “N does not exist as a Kc” is equivalent to “N as a Kc does not exist.” Rather, what I deny is that the existence predicate is a context-sensitive term, so that the use of an FNE semantically expresses the proposition that N does not exist as a Kc in a given context. Moreover, as we have seen, in cases where a speaker utters (13), it becomes problematic to assume that the speaker uses the existence predicate for the plural subject in order to pragmatically convey <existing as a Kc>, where Kc is a single salient kind in a given context. That is precisely why we need a general approach like PAF.

Thus far, I have argued that PAF satisfactorily explains the apparent truth of a bare FNE. But there is one remaining concern: given that ordinary speakers seem to have a natural intuition that (4) and (5) are literally true, I seem to have to admit that there is a kind of meaning confusion among ordinary speakers. Note, however, that my previous explanation only works when a speaker has a referential intention, whether explicit or implicit, for her use of a name to refer to a fictional object. While I expect that in most cases, ordinary speakers at least implicitly have such a referential intention when they think that they are using a fictional name and thus uttering an FNE, it seems evident that this is not true for all conversational situations. In uttering an FNE, ordinary speakers sometimes seem to have no referential intention to refer to any object, believing instead that their use of a fictional name refers to nothing. ³¹ If so, it follows that not every way in which a speaker uses a bare FNE has been explained yet. From the next section, to complete an exhaustive explanation for an FNE, I will explore a case where a speaker has no referential intention. I believe this will shed important light on the source of ordinary speakers’ meaning confusion concerning FNEs and why the problem of FNEs is notoriously difficult to address.

5. No Referential Intention for Fictional Names in Negative Existentials

Suppose Gi-hun utters (4), but he has a clear intention for his use of “Sherlock Holmes” to refer to nothing, believing that the name “Sherlock Holmes” refers to nothing. ³² Suppose further that the hearer, Jun-ho, grasps Gi-hun’s intention. As we have seen, PAF is silent on this case. Nevertheless, it still seems that even in this case, Gi-hun can successfully convey something true to Jun-ho. Certainly, it cannot be the literal meaning of (4), which is false according to creationism. Consequently, our remaining task is to elucidate how Gi-hun conveys something true and to identify the proposition he conveys.

My proposal is to invoke the distinction between the speaker’s referent and the semantic referent of “Sherlock Holmes.” ³³ As a creationist, I suggest that the semantic referent or content of “Sherlock Holmes” is constantly fixed to the abstract artifact Sherlock Holmes, so that in any conversational situation, our use of “Sherlock Holmes” semantically refers to the very abstract artifact Sherlock Holmes. There are at least two reasons supporting this view. First, it offers its own explanatory power. If the semantic referent of “Sherlock Holmes” is indeed the abstract artifact Sherlock Holmes, then, as we have seen, our intuition about the literal truth of external metafictional sentences like (1)–(3) and even of FNEs in (6)–(8) is straightforwardly explained. Second, it possesses the virtue of theoretical simplicity; it is simpler than the view that the semantic referent of “Sherlock Holmes” is indeterminate or ambiguous between an abstract object and nothing. Thus, if my simpler view successfully provides a convincing analysis of the phenomenon that we seek to explain, as I will argue in this section and the next, there would be no reason to adhere to the alternative, more complex view. ³⁴

On the view I have taken, then, even if an ordinary speaker sometimes uses “Sherlock Holmes” to refer to nothing when uttering (4), this deviant use does not determine the semantic referent of “Sherlock Holmes” in (4). ³⁵ In other words, while the speaker’s referent of Gi-hun’s use of “Sherlock Holmes” is nothing, its semantic referent remains the fictional object Sherlock Holmes. Gi-hun is mistaken in thinking that the semantic referent of “Sherlock Holmes” is nothing, and as a result, he does not intend for his use of “Sherlock Holmes” to refer to anything.

The situation is analogous to cases where we make a mistake in using an ordinary name. For example, suppose I attend a film festival with my friend and encounter the movie star Colin Firth. I intend to inform my friend about him, but I mistakenly use a similar name of another movie star and say, “Colin Farrell was fantastic in the Kingsman movie series!” In this case, the speaker’s referent for my use of “Colin Farrell” is Colin Firth, while its semantic referent is Colin Farrell. Furthermore, despite my mistake, I can still pragmatically convey something true to my friend, namely the proposition that Colin Firth was fantastic in the Kingsman movie series, provided my friend immediately grasps my intention.

Similarly, Gi-hun can pragmatically convey something true to Jun-ho. I propose that the proposition conveyed by Gi-hun is the negation of a so-called atomic gappy proposition: gappy propositions are propositions that lack their essential constituents, and they are expressed by sentences containing genuinely non-referring names. ³⁶ For explanatory purposes, let us assume the existence of a structured singular proposition. Then we can represent the proposition semantically expressed by (1) with the ordered pair <Holmes, being a fictional character>. Likewise, (4) semantically expresses the false proposition <<Holmes, existing>, NEG>. But since Gi-hun intends for his use of “Sherlock Holmes” to refer to nothing when uttering (4), even though (4) is literally false, he pragmatically conveys something true, that is, the negation of the atomic gappy proposition <<__, existing>, NEG>, given that an atomic gappy proposition is always untrue. ³⁷

We now have an explanation for how Gi-hun conveys something true and which proposition he conveys. When uttering (4), Gi-hun mistakenly thinks that the semantic referent of “Sherlock Holmes” is nothing, leading him to intend to refer to nothing. As a result, the speaker’s referent for Gi-hun’s use of “Sherlock Holmes” becomes nothing, and Gi-hun conveys the true gappy proposition <<__, existing>, NEG>, mistakenly believing that what he conveys is the literal meaning of his utterance. If Jun-ho grasps Gi-hun’s intention, he can get the proposition conveyed by Gi-hun. ^{38 , 39 , 40}

With the account provided above, we finally have a theoretically complete explanation of how the speaker can convey something true by uttering (4): If the speaker has the intention for her use of “Sherlock Holmes” to refer to a fictional character, she pragmatically conveys the true descriptive proposition that Sherlock Holmes as a person (a detective, or a concrete object) does not exist. If, on the other hand, the speaker has no referential intention (by mistake), she pragmatically conveys the true gappy proposition <<__, existing>, NEG>. In either case, the speaker can convey something true. ⁴¹

But what about the actual linguistic practice of ordinary speakers with respect to bare FNEs? Can the theoretical explanation provide thus far be readily applied to our actual linguistic practice? If most speakers consistently maintain a single transparent referential intention in every utterance of FNEs, whether that intention is to refer to a fictional object or to refer to nothing, then there would be no issue in directly applying our explanation to the actual linguistic practice. Reality, however, does not seem to be quite so straightforward. In fact, I believe that the final crucial step in explaining the apparent truth of a bare FNE is to address this very issue.

6. Ordinary Speakers’ Inconsistent Intentions

I mentioned earlier that I expect that in most cases, when ordinary speakers utter FNEs, they at least implicitly have the intention to refer to a fictional object. However, that does not mean that they do not at the same time have any implicit intention to refer to nothing. If they are not careful about maintaining their own referential intention when they utter FNEs, they are likely to become confused and fail to have a single transparent intention. They might implicitly and vaguely have two inconsistent intentions without consciously realizing they have both intentions. It might be the case that they implicitly have the intention to refer to a fictional object when they begin to utter but think that they have no referential intention at the end of their utterance, or vice versa. I believe that this is a real source of difficulty in resolving the problem of FNEs.

Indeed, David Braun has previously highlighted that ordinary speakers’ beliefs about fiction are confused and inconsistent. He wrote:

Most ordinary speakers’ beliefs about fiction really are (deep down) confused and inconsistent. This should not surprise us. Fiction, after all, raises hard issues in semantics and metaphysics. We should not expect ordinary speakers to (tacitly) believe a coherent, unproblematic theory of fiction, when philosophers have so many difficulties formulating an explicit one that fits their pre-theoretic intuitions. (Braun, 2005, p. 613)

I agree with Braun’s view with respect to bare FNEs, as I expect that, more often than not, ordinary speakers are indeed confused and implicitly have two inconsistent referential intentions when they utter bare FNEs. Unlike Braun, however, the kind of inconsistency I ascribe to ordinary speakers is a local one; it arises only when they utter FNEs and is not a widespread one extending to every use of all kinds of sentences about fiction. For example, I do not see a compelling reason to attribute inconsistent referential intentions to ordinary speakers when they utter external metafictional sentences like (1)–(3) or modal FNEs like (6)–(8). That is, my theory minimalizes the attribution of inconsistency and confusion to ordinary speakers as much as possible, which is preferable if it provides a satisfactory explanation for the phenomena.

Suppose the speaker implicitly has two inconsistent intentions when she utters (4). In such a case, there are two true propositions that she can convey to the hearer: one is the proposition that Sherlock Holmes as a person does not exist, and the other is the proposition <<__, existing>, NEG>. If, after her utterance, she pays attention to her intention to refer to a fictional object, she will believe that she has pragmatically conveyed the first proposition. On the other hand, if she pays attention to her other intention to refer to nothing, she will believe that she has literally conveyed the second proposition. ⁴² But without any conscious reflection on her own mental state, the speaker’s thought remains disorganized and confused. She believes, of course, that she has conveyed something true anyway, but she has no clear idea of which intention(s) she has, exactly which proposition(s) she conveys, and whether the proposition(s) she conveys belongs to an implicature or a literal meaning. The same applies to the hearer as well. Although the hearer believes that the speaker conveys something true to him, he too has only a vague idea of which intention(s) the speaker has, which proposition(s) he gets from the speaker, and whether the proposition(s) he gets belongs to an implicature or a literal meaning. I suggest that all of this contributes to the source of ordinary speakers’ meaning confusion with respect to bare FNEs, if there is any ordinary intuition that bare FNEs such as (4) and (5) are literally true.

In this regard, the difficulty in resolving the problem of FNEs arises from this inconsistency and confusion of ordinary speakers. Suppose, without exception, that ordinary speakers invariably intend to refer to a fictional object using a fictional name, even when they utter bare FNEs such as (4) and (5). Under this supposition, since they will recognize that the true descriptive proposition that Sherlock Holmes as a person does not exist is conveyed only pragmatically, it becomes challenging to explain why many have the mistaken ordinary intuition that FNEs such as (4) and (5) are literally true, even though they are in fact literally false.

But it is also unsatisfactory to assume that, unlike other kinds of metafictional sentences, bare FNEs are exceptional in the sense that ordinary speakers always have no referential intention whenever they utter FNEs. This is because, not to mention modal FNEs in (6)–(8), there are certain linguistic data on bare FNEs which support that the speaker has a referential intention to refer to a fictional object. For example, suppose I utter “Sherlock Holmes does not exist” and a little later I utter “Spider-Man does not exist either.” Then it seems that I am literally talking about two different fictional characters (not just about two different fictional names) and expressing something different by each utterance. ⁴³ But if I utter “Peter Parker does not exist” and a little later I utter “So, neither does Spider-Man,” it seems that I am literally talking about one and the same fictional character. ⁴⁴

All of this suggests that we should admit that, in our actual linguistic practice, ordinary speakers frequently have inconsistent referential intentions when they utter bare FNEs. If we overlook this and assume that ordinary speakers’ uses of fictional names in FNEs are consistent, we will always be left with some unexplained ordinary intuitions about FNEs. Embracing the reality of this inconsistency and confusion among ordinary speakers enables us to find a way to account for all the various ordinary intuitions about FNEs.

I suggest that the inconsistency arises because ordinary speakers have ambivalent ordinary intuitions about fictional objects. Indeed, the following seems clear to me: In many cases, ordinary speakers’ uses of fictional names presuppose that there are fictional objects, but it is also hard to deny that they have an ordinary anti-realist intuition about fictional names and objects. To the extent that this anti-realist intuition is pre-theoretic and ordinary, it is well explained away by creationism, as we have seen so far. But what about cases where a speaker explicitly believes and expresses her belief that fictional characters are nothing at all, just like an anti-realist? ⁴⁵ Before finishing, let me make a diagnosis of this case.

Consider the following conversation:

Jennie: Sherlock Holmes does not exist.

Lisa: You mean Sherlock Holmes is not a real person, right?

If Jennie casually replies, “Yes, because Sherlock Holmes is a fictional character,” then no problem arises, and they are staying in an ordinary conversational context. However, suppose Jennie replies as follows:

1. (15) No, I do not just mean that Sherlock Holmes is not a real person. What I mean is that Sherlock Holmes is nothing at all, because he is a fictional character. Fictional characters do not exist, not just as real people, but they are simply nothing.

I suspect that in this case, if Jennie’s view, as expressed by (15), is literally and explicitly the same as the view held by anti-realists, then Jennie is expressing her premature philosophical view, which could develop into anti-realism. But if a speaker’s anti-realist intuition becomes no longer pre-theoretic, it is surely wrong according to creationism. While some speakers may sometimes say something like (15), creationists need not paraphrase (15) as expressing true and consistent propositions. Indeed, creationists need not paraphrase the philosophical views of anti-realists and their utterances as true and consistent. ⁴⁶ Thus, the fact that there is no way for creationists to paraphrase (15) does not provide any motivation for giving up creationism. According to creationism, it is simply wrong to say that fictional characters are simply nothing. ⁴⁷

7. Concluding Remarks

In this paper, I have argued that FNEs do not pose any serious problem for creationism. I first pointed out that modal FNEs provide linguistic evidence that favors creationism over anti-realism, as anti-realists require additional revisionary work, whereas creationists can straightforwardly explain them. I then diagnosed that the difficulty in resolving the problem of FNEs arises from the inconsistency and confusion of ordinary speakers. By acknowledging this, we can satisfactorily explain all the various ordinary intuitions regarding FNEs.

From this, I hope to have demonstrated that creationism remains a viable and compelling view. At the very least, I hope that my discussion in this paper contributes to a deeper understanding of the problem of FNEs, which stands as one of the most perplexing and significant issues that has constantly troubled philosophers.