A Leibnizian Antirealist Account of Fictional Characters

Abstract

Alberto Voltolini advocates a syncretistic account of fictional entities, asserting that it satisfies all the desiderata for a suitable account of fictional entities. This article presents an alternative account of fictional characters that meets these criteria, while circumventing the problems with Voltolini's account. On my Leibnizian antirealist account, a fictional character can be identified by the collection of predicates attributed to its name. And this account offers the benefit of the bundle theory, avoiding the issue of bearers of fictional names, while also explaining why a fictional character is more than just a collection of predicates.

Résumé

Alberto Voltolini préconise une analyse syncrétique des entités fictives, affirmant qu'elle satisfait tous les desiderata d'une analyse appropriée des entités fictives. Cet article présente une analyse des personnages fictifs qui réponde à ces critères, tout en évitant les problèmes que rencontre l'analyse de Voltolini. Selon mon analyse antiréaliste et leibnizienne, un personnage fictif peut être identifié par la collection de prédicats attribués à son nom. Cette analyse offre le bénéfice de la théorie des faisceaux, puisqu'elle écarte la question des porteurs des noms fictifs, tout en expliquant pourquoi un personnage fictif est plus qu'une simple collection de prédicats.

1. Introductory Remarks

Alberto Voltolini (2006, 2015) advocates a syncretistic account of fictional entities on the grounds that it satisfies all the desiderata for a suitable account of fictional entities. This account combines elements from both neo-Meinongian accounts of fictional entities, such as Terence Parsons (1980) and Edward Zalta (1983, 1988), and artifactualist accounts, such as Amie Thomasson (1999, 2003, 2015a, 2015b), Nathan Salmon (1998), and Stephen Schiffer (1996).¹ Let me first briefly explain these desiderata. For brevity, I will focus on fictional characters among fictional entities.

1. (1) The non-existence of fictional characters: Fictional characters like Sherlock Holmes cannot be located in space-time. Therefore, a suitable account of fictional characters should acknowledge that fictional characters do not exist spatiotemporally.

2. (2) The causal inefficacy of fictional characters: Fictional characters do not exist spatiotemporally. Accordingly, we cannot causally interact with them. Therefore, a suitable account of fictional characters should acknowledge that nothing can causally modify a fictional character.

3. (3) The incompleteness of fictional characters: A complete object is such that for any property P, either it has P or it does not have P. But fictional characters do not satisfy this condition. For example, we cannot definitely say that Holmes has a mole on his left foot, nor can we definitely say that he lacks a mole on his left foot. For Arthur Conan Doyle didn't mention anything about a mole on Holmes's left foot in any of his Holmes stories. Therefore, a suitable account of fictional characters should explain why fictional characters are not complete objects.

4. (4) The created nature of fictional characters: Fictional characters, such as Holmes, are the creations of their authors.² Therefore, a suitable account should explain why fictional characters can be considered to be the creative products of their authors.

5. (5) The actual possession of narrated properties by fictional characters: Fictional characters possess narrated properties. For example, according to the Holmes stories, Holmes is a detective. So, a suitable account should explain why it is correct to say that Holmes is a detective, whereas it is incorrect to say that Holmes is a ballet dancer.

6. (6) The unrevisable ascription of narrated properties to fictional characters: An author of a work of fiction holds the ultimate authority regarding properties assigned to fictional characters. For example, given that Conan Doyle wrote a story in which Holmes is a detective, Holmes is a detective. To put the point another way, whether a fictional character has a certain property is prescribed by an author in such a way that there is nothing readers can do about it. Therefore, a suitable account should be able to explain this fact.

7. (7) The necessary possession of narrated properties by fictional characters: It appears that fictional characters necessarily possess properties attributed to them in relevant stories. For example, it does not seem appropriate to say that Holmes might not have been a private detective. Of course, we can easily imagine a possibility that Conan Doyle might have written a somewhat different story using the name “Sherlock Holmes.” But, if such were the case, strictly speaking, we would have a different story, and so a different fictional character. Therefore, it is desirable to explain why fictional characters necessarily possess the properties attributed to them.

After pointing out the above seven desiderata, Voltolini defends a syncretistic account of fictional entities that combines the virtues of neo-Meinongianism and artifactualism by claiming that it fulfills all these desiderata.

To begin, Voltolini's syncretistic account shares with neo-Meinongianism the idea that fictional objects are correlates of property sets.³ This idea provides a necessary condition for the identity of a fictional entity: x and y are the same fictional entity only if they share the same property set.

Voltolini's syncretistic view also shares with artifactualism the idea that acts of make-believe enable us to correlate fictional objects with property sets. Based on this idea, Voltolini explains why fictional entities are created entities whose existence depends on creative acts of story-tellers. In particular, on his view, fictional entities begin to exist only after relevant make-believe processes have taken place, and they exist non-spatiotemporally and so are abstract entities.

For the above reasons, on Voltolini's syncretistic account, fictional entities are abstract, hybrid entities, which are made of two components: a certain make-believe narrative process, and the set of properties that one such narration mobilizes.

I agree with the aforementioned seven desiderata. But, since Voltolini's syncretistic account combines elements of neo-Meinongianism and artifactualism, it inherits some drawbacks from each of its constituent theories. So, this article presents an alternative account of fictional characters that meets these criteria, while circumventing the problems with Voltolini's account.

This article proceeds as follows. In Section 2, I briefly discuss three important problems with Voltolini's view. In Section 3, I present a Leibnizian account of fictional characters, according to which a fictional character can be identified by the collection of predicates attributed to its name. In particular, I argue that this account offers the benefit of the bundle theory, avoiding the issue of bearers of fictional names, while also explaining why a fictional character is more than just a collection of predicates. Finally, in Section 4, I argue that my Leibnizian antirealist account can accommodate the aforementioned seven desiderata, while avoiding the problems with Voltolini's syncretistic view. In particular, I argue that it explains why fictional characters can be considered to be the creations of their authors, without committing to the existence of fictional characters.

2. Three Problems with Voltolini's View

As mentioned, Voltolini's syncretistic account of fictional entities combines elements of both neo-Meinongianism and artifactualism. Consequently, it inherits some drawbacks from each of its constituent theories. In this section, I briefly discuss three important problems of this account: counterintuitivity, ontological overpopulation, and indigestible creationism.

On Voltolini's syncretistic account, fictional entities are abstract, hybrid entities, which are made of two components: a certain make-believe narrative process, and the set of properties that such a narration mobilizes. So, the first important problem with this account is that the existence of such hybrid entities is very counterintuitive. Voltolini (2015, pp. 137–138) addresses this problem by biting the bullet. He argues that a counterintuitive metaphysical view can be justified if it proves explanatorily fruitful. It is an indisputable merit of his syncretistic theory that it satisfies all the criteria for a suitable account of fictional entities. However, it would be preferable to avoid such counterintuitive metaphysics if at all possible. In this article, I will present an alternative account that fulfills these criteria without resorting to counterintuitive metaphysics. If there is indeed such an account, the problem of counterintuitivity would be a significant weakness in Voltolini's account.

The second important problem with the syncretistic account is that it unnecessarily proliferates fictional entities. As mentioned, Voltolini accepts the neo-Meinongian view that fictional objects are correlates of property sets. On this view, x and y are the same fictional entity only if they share the same property set. As a consequence, if an author makes even a slight alteration to their draft, a character being depicted would change, leading to an unnecessary proliferation of fictional entities. Voltolini (2015, pp. 138–140) also bites the bullet regarding this problem. He denies that fictional characters persist across different stories. Instead, he reinterprets this at the level of a non-committal make-believe process. More specifically, he claims that this should be understood in terms of make-believe protraction. On this proposal, in storytelling, an author typically makes believe that there is a particular individual having certain properties. Later on, the same author or a different author can resume the make-believe process by attributing additional properties to the very same character that was originally pretended to have some properties. According to Voltolini, this is a case of intentional identity to be grounded not in the existence of an individual, but in the identity conditions of a make-believe process along with its protraction.

But two fictional characters with the same name can exhibit significant differences, making it inappropriate to claim that the make-believe process related to one fictional character is a protracted extension of the make-believe process related to the other fictional character. For instance, Holmes appearing in A Study in Scarlet apparently reappears in other works like A Samba for Sherlock by Brazilian author Jô Soares. While there are notable similarities between the Holmes depicted in A Study in Scarlet and the one in A Samba for Sherlock, there are also many significant dissimilarities. For example, compared with the former, the latter is fumbling and nearsighted, resulting in his failure to solve crimes. Due to such significant differences between the Holmes in A Study in Scarlet and the Holmes in A Samba for Sherlock, it is not appropriate to consider the latter as a protracted extension of the make-believe process related to the former.

More importantly, Voltolini bites the bullet regarding the problem of unnecessarily proliferating fictional entities. The reason is clear: he embraces realism concerning fictional characters. And he contends that any divergence in properties between two fictional characters warrants considering them to be distinct fictional entities.⁴ But it would be desirable to avoid the problem of ontological overpopulation, if possible.⁵

The third problem with Voltolini's syncretistic view concerns the question of how fictional entities can really be created. He agrees with artifactualists that a make-believe process is necessary to create a fictional entity. But he believes that the existence of a certain type of make-believe process alone is not enough to bring a fictional entity into existence (see Voltolini, 2006, pp. 76–78). On his view, fictional entities are the outcomes of correctly seeing relevant make-believe processes, and so a reflexive stance on relevant make-believe processes is also necessary. He writes:

In actual fact, those rules do not involve the relevant make-believe process as such, but rather the reflexive stance that allows one to take that process as involving a certain property set. By seeing a certain make-believe process as involving a certain property set, a certain fictional entity is eo ipso generated. Seeing the process in this way amounts to having a fictional entity at one's disposal is the relevant conceptual truth expressing a constitutive rule for ficta. (Voltolini, 2015, p. 143)

But it is unclear how ontologically real objects can be generated through such a reflective stance on relevant make-believe processes.⁶ So far, no one, including Voltolini, has provided a satisfactory explanation of the mechanism through which we, existing in space-time, can create abstract entities outside of space-time. Therefore, if possible, it would be desirable to have an account that does not face the issue of how fictional entities can really be created.⁷

3. A Leibnizian Account of Fictional Characters and the Anaphoric Theory of Reference

In this section, I present a Leibnizian antirealist account of fictional characters, according to which a fictional character can be identified by the collection of predicates attributed to its name.

To begin, in “Discourse on Metaphysics,” G. W. Leibniz (1686/1989, §8) defines an “individual substance” as a single subject to which several predicates are attributed, while this subject is attributed to no other. So, he regards an individual substance as a subject of predication. And he upholds the Complete Individual Concept Account of Substance (hereafter, CIC account), according to which every individual substance has one complete individual concept, that is, a concept that contains within it all predicates which can be correctly attributed to the individual substance. Furthermore, Leibniz upholds the Principle of the Identity of Indiscernibles (hereafter, PII), according to which x and y are identical if every predicate true of x is also true of y, and vice versa. And this principle can serve as the principle of individuation for individual substances.⁸ Nowadays, Leibniz's CIC account is widely regarded as implausible. Nevertheless, in the remainder of this section, I will argue that it is useful in providing a suitable account of fictional characters.

To begin, as mentioned earlier, on Leibniz's view, an individual substance is a subject of predication. In a similar vein, a fictional character, such as Holmes, is a subject of predication. For example, a predicate “is a detective” can be correctly attributed to a fictional name “Holmes.”

In addition, on Leibniz's view, God can create individuals in such a way that the truths of statements about them are completely determined by Him (see Leibniz, 1686/1989, §13, pp. 45–46). This is why, on his view, every individual substance has one complete individual concept, which includes all predicates true of the individual substance. And Leibniz's PII can serve as the principle of individuation for individual substances. Similar points can be made about fictional characters. Fiction authors can create fictional characters in such a way that the truths of statements about them are completely determined by their authors. For this reason, we can understand, for example, the concept of Holmes as the concept that contains within it all the predicates that Conan Doyle has attributed to the fictional character. And Leibniz's PII can serve as the principle of individuation for fictional characters.

For the above reasons and others, it would be worth considering the following proposal: a fictional character can be identified by the collection of predicates that can be correctly attributed to its name. As we will see in the next section, this account satisfies the aforementioned seven desiderata for a suitable account of fictional characters, while avoiding the problems with Voltolini's syncretistic view.

Before proceeding, however, there are two key questions that must be addressed for it to be considered a viable account: What binds some predicates (or properties) together to make up a fictional character? And how can a fictional character, as understood by this account, serve as a subject of predication?

Let us start with the first question. The substratum theory holds that a substratum is necessary to unify properties and form a single object. By contrast, the bundle theory holds that the compresence relation can achieve this without the need for a substratum. So, both theories aim to provide a unifying device.⁹ But the bundle theory does so without requiring the existence of a substratum as a fundamental component of reality.

One significant benefit of the bundle theory is that it can avoid the problems associated with substrata (or bare particulars). One significant issue with the substratum theory concerns the role of a substratum as a subject of predication. Take the statement “Joe Biden is a person.” The substratum theory holds that every substance includes a substratum that serves as the subject of its properties. So Joe Biden's substratum is supposed to be the subject of his properties. Then, we should be able to say, for instance, that Joe Biden's substratum is a person. But “Joe Biden” here is the name of a particular person. So, if something is not a person, then it cannot be Joe Biden. In addition, Joe Biden's substratum is his constituent. Therefore, the substratum theory leads us to ask: “How can a constituent of a person be a person?” or “How can we correctly attribute the property of personhood to a constituent of a person?” These are challenging questions for substratum theorists to answer.¹⁰

However, the bundle theory must address another problem in its place. This theory holds that a substance is nothing more than a bundle of properties. But, as Bertrand Russell (1900, p. 49) points out, there is a logical reason for distinguishing an individual substance from a bundle of properties. An individual substance is a subject of predication, which is logically distinct from any predicate. In other words, there is a role of a substance that is not exhausted by its properties, namely, its role as a subject of predication.¹¹ Likewise, there is a role of a fictional character that is not exhausted by its properties, namely, its role as a subject of predication. At least for this reason, a fictional character must be more than just a collection of predicates.

In the remainder of this section, I will present a Leibnizian antirealist account of fictional characters, which offers the benefit of the bundle theory, avoiding the issue of bearers of fictional names, while also explaining why a fictional character is more than just a collection of predicates.

On Sellars-Brandom's inferentialist semantics, the meaning of an expression is constituted by relevant language norms that determine its correct use.¹² Furthermore, as I have argued in detail elsewhere (Lee, 2022), we can understand the meaning of a fictional name in a similar way, that is, in terms of relevant language norms that determine its correct use. And this account can explain the meaning of a fictional name without attributing a bearer to the name. Let me explain its main idea.

To begin, the inferentialist account of fictional names adopts a Brandomian anaphoric theory of reference, which allows us to understand reference in term of anaphoric word-word relation, rather than in terms of substantial word-world relation. What then is an anaphoric word-word relation? Consider the following statement:

If Mary wants to leave on time, she should leave now.

In this conditional statement, “she” is a pronoun that is used instead of the proper name “Mary” in the antecedent. Consequently, the token of “she” bears an anaphoric word-word relation with the token of “Mary.” And such an anaphoric relation is a commitment-preserving link in the following sense: if anyone treats two word tokens as anaphorically related, then they are thereby committed to treating both as having the same inferential significance. Robert Brandom argues that we should understand the notion of reference in the same way. Suppose that someone named “Joe” makes the following statement to another person with the name “Jim”:

I should have known better than to let the mechanic Binkley work on my car. That airhead misadjusted the valves.

Suppose also that Jim forgot the name “Binkley,” but he nonetheless remembers that Joe called the mechanic as “that airhead.” Then he may say:

For car repair, don't go to the mechanic Joe referred to as “that airhead.”

According to Brandom (1994, p. 305, 2005, pp. 265–266), in this discourse, the description “the mechanic Joe referred to as ‘that airhead’” is a lexically complex pronoun that takes the token of “that airhead” originally used by Joe as its anaphoric antecedent. Brandom calls such a description “an anaphorically indirect definite description.” If the antecedent and dependent tokens are sufficiently close to each other in time, space, or audience attention, one may use lexically simple pronouns such as “he,” “she,” or “it.” As for distant antecedents, however, one might be required to use such indirect definite descriptions, which give us more information about their antecedents. On Brandom's view, we can understand the expression “refers” as a pronoun-forming operator that is used to form such an anaphorically indirect definite description. Accordingly, in the above case, the token of “the mechanic Joe referred to as ‘that airhead’” bears an anaphoric word-word relation with the token of “that airhead,” and such an anaphoric relation is a commitment-preserving link; that is, since Jim uses the former instead of the latter, if Jim is committed to holding that the one Joe referred to as “that airhead” is F, then he should also be committed to holding that the one he referred to as “the mechanic Joe referred to as ‘that airhead’” is F. On the anaphoric theory of reference, therefore, we can understand the expression “refers,” not in terms of a substantial relation between a linguistic expression and an extra-linguistic entity, but rather in terms of an anaphoric word-word relation.

In addition, the anaphoric theory of reference allows us to explain the meaningfulness of fictional names, even if these names lack bearers. The most important thing to note in this regard is that, even if a fictional name lacks a bearer, a token of the name can initiate an anaphoric chain, which can be continued by other tokens of the same name or tokens of a pronoun. For example, one can start to make up a story using sentences like the following: “Sherlock Holmes is a detective. He lives with Dr. Watson in Baker Street. He is interested in early English chapters,” and so on.

Here, a token of the name “Sherlock Holmes” initiates an anaphoric chain that is continued by tokens of the pronoun “he.” It is (partly) by virtue of such an anaphoric chain that the correctness conditions for the use of “Sherlock Holmes” are established. And it is by virtue of such an anaphoric chain that different tokens of a fictional name are all about the same fictional character.

For the above reason, an author can make up a story by constructing anaphoric chains of fictional names in the way suggested above, even if these names lack bearers. And the anaphoric theory of reference enables us to use (or understand) a fictional name correctly, in accordance with relevant language norms, even if the name lacks a bearer. This is again because the sameness of reference can be achieved by an anaphoric relation, and reference here should be understood in terms of an anaphoric word-word relation, instead of a substantial referential relation between a linguistic expression and an extra-linguistic entity. Therefore, the meaningfulness of a fictional name does not depend on the condition that it has a bearer. For the meaning of a fictional name is constituted by relevant language norms that determine its correct use, and these language norms do not depend on the condition that the fictional name has a bearer.¹³

It is beyond the scope of this article to fully defend the anaphoric theory of reference.¹⁴ So, let me here confine myself to emphasizing that the anaphoric theory of reference is a viable theory of reference.

Now, with the anaphoric theory of reference on the table, let us consider again the following two key questions: What binds some predicates together to make up a fictional character? And how can a fictional character, as understood by my Leibnizian account, serve as a subject of predication?

To begin, anaphoric relations between tokens of a fictional name can play the role of binding various predicates together into predicates that can be correctly attributed to the fictional name. Due to such anaphoric relations, for example, “is a detective” is correctly applied to “Holmes,” while “is a ballet dancer” is not. Therefore, the anaphoric theory of reference enables us to adequately address the first question, without attributing bearers to fictional names. In addition, on the inferentialist account, fictional names are genuine names that serve as subjects of predication. Therefore, this account also has no difficulty in explaining why fictional names like “Holmes” are logically distinct from any collection of predicates.

For the above reasons, my Leibnizian antirealist account of fictional characters can answer the first question without attributing bearers to fictional names, and it can explain why fictional names, as subjects of predication, are logically distinct from any collection of predicates. Therefore, we can identify a fictional character by a collection of predicates attributed to its name, while avoiding the need for an implausible ontology of fictional entities. To put the point another way, my Leibnizian antirealist account offers the benefit of the bundle theory, particularly by circumventing the issue of bearers of fictional names, while simultaneously avoiding the main problem with this theory by explaining why a fictional character is more than just a collection of predicates.

Three cautionary notes might be necessary here.

First, I am not advocating Leibniz's CIC account concerning non-fictional individuals. And my intention is not to defend Leibniz's own view on fictional names but rather to present a Leibnizian account of fictional characters inspired by his CIC account. For example, it is worth recognizing that the anaphoric theory of reference is not Leibniz's own view.

Second, the realism-antirealism debate concerning fictional characters is deeply intertwined with the semantic analysis of fictional terms. Much of this debate has been grounded in the representationalist approach to meaning. For instance, Voltolini adopts truth-conditional semantics that aligns with a realist stance on fictional entities. However, it is important to recognize that representational semantics and inferential semantics are rival approaches in contemporary philosophy of language. While the debate has traditionally assumed the representationalist approach to meaning, this assumption is not justified in the current context. Consequently, the real debate should centre on determining which approach provides a more comprehensive and less problematic explanation overall. And it can be argued that adopting the inferentialist approach, rather than the representationalist approach, can better address at least some key challenges in the realism-antirealism debate concerning fictional characters.

Third, and related to the second point, although the realism-antirealism debate concerning fictional characters is deeply intertwined with the semantic analysis of fictional terms, it is still necessary to distinguish between semantic questions and ontological questions. What is the meaning of a fictional name such as “Sherlock Holmes”? I answer this semantic question based on the aforementioned inferentialist account of fictional names. On this semantic account, the meaning of a fictional name is constituted by relevant language norms that determine its correct use. And it enables us to use fictional names correctly without committing to an implausible ontology of fictional entities. What then is the ontological nature of a fictional character like Holmes? I answer this ontological question based on my Leibniz-inspired bundle theory. On this theory, a fictional character can be identified by the collection of predicates attributed to its name. It is important to note that the inferentialist account of fictional names constitutes a crucial component of my Leibnizian antirealist account of fictional characters, but the latter is primarily concerned with the ontological nature of fictional characters. I will say more on this point in the next section.¹⁵

4. Meeting the Seven Desiderata for Fictional Characters

In this section, I argue that my Leibnizian antirealist account of fictional characters explained in the previous section satisfies the seven desiderata for a suitable account of fictional characters.

As mentioned in Section 1, the first desideratum is the non-existence of fictional characters, and the second desideratum is the causal inefficacy of fictional characters. These desiderata are easily satisfied by my antirealist account of fictional characters, because it denies the existence of fictional characters.

The third desideratum is the incompleteness of fictional characters. For example, we cannot assert that Holmes has a mole on his left foot, nor can we assert that he does not have a mole on his left foot. This desideratum does not pose a problem for my antirealist account. On my account, Holmes is identified by the collection of predicates attributed to his name. In Conan Doyle's Holmes stories, there is no mention of a mole on Holmes's left foot. Consequently, neither the statement “Holmes has a mole on his left foot” nor “Holmes does not have a mole on his left foot” can be definitively attributed to the character. Therefore, it is correct to assert neither that Holmes has a mole on his left foot, nor that he does not.¹⁶

The fourth desideratum is the created nature of fictional characters. My account can meet this desideratum as well. On my account, if a work of fiction written by an author has been successfully published, certain norms of fiction are thereby established, and these norms are such that, if one engages with a fictional story, one is thereby prescribed to imagine as the story says; in other words, one is subject to norms of the following form: “Imagine that p if, according to the story, p.”¹⁷

In addition, with the establishment of such norms, which did not exist previously, we can assert that fictional characters, like Holmes, have been created. In this regard, it is important to note that creating such a fictional character does not imply that the character's name refers to an ontologically real entity. Let me draw an analogy. Suppose that a game called “Rock, Paper, Scissors” is devised with three rules: Rock (a closed fist) beats Scissors (a “V” shape made with the index and middle fingers) but loses to Paper (a flat hand); Paper beats Rock but loses to Scissors; and Scissors beats Paper but loses to Rock. If this game is frequently played in accordance with these rules within a society, we can say that it has been “created.” However, this creation doesn't imply that the name “Rock, Paper, Scissors” refers to an ontologically real object. As pointed out in the previous section, Sellars-Brandom's inferentialist semantics allows us to use some names correctly, even if they lack bearers. As a consequence, we can use the name “Rock, Paper, Scissors” to talk about the above game without assuming that it refers to an ontologically real object. A similar point can be made about fictional characters. If a work of fiction featuring a fictional name, such as “Sherlock Holmes” has been successfully published, we can say that a fictional character named Sherlock Holmes has been “created.” And we can use the name “Sherlock Holmes” to talk about this character without assuming that it refers to an ontologically real object. For this creation pertains not to an ontologically real entity but to the norms governing the correct use of the fictional name. Let me elaborate on this point.

Above all, norms are categorically different from objects. First, norms are such that they can be evaluated as correct (or justified) or incorrect (or unjustified). By contrast, objects themselves, whether concrete or abstract, are not such that they can be correct or incorrect. Second, we can act in compliance with norms we adopt, but we can also violate them, although violating them may result in punishment or other negative sanctions. By contrast, objects themselves are not such that we can violate them. Third, and most importantly, norms belong to the logical space of reasons, whereas objects themselves do not. As Wilfrid Sellars (1963) argues, the logical space of reasons should be distinguished from the realm of law. The realm of law is the domain of science and empirical inquiry, where we seek to understand the objective structure of the world. Specifically, this realm concerns the causal, scientific order of the world, which is to be explained on the basis of scientific principles and empirical evidence, rather than normative reasons. By contrast, the logical space of reasons concerns the realm in which we engage in rational activities such as reasoning, justifying, and giving and asking for reasons. Therefore, norms belong to the logical space of reasons. Recall that norms can be evaluated as correct (or justified) or incorrect (or unjustified). However, we have no good reason to think that objects themselves, whether concrete or abstract, belong to the logical space of reasons.

In addition, the normative force of a norm depends on our normative attitude. Note that if we all do not accept a norm, then it cannot exert normative force on us. Accordingly, we can make a norm no longer real by refusing to accept it any longer. Note also that norms exert normative force on us through the enforcement of sanctions on violators, and we are compelled to follow norms not by alleged abstract entities associated with norms but by actual individuals who impose sanctions when non-compliance occurs. By contrast, whether a certain thing is a real object is a matter of ontology. And we can hardly say that we can make the object no longer real just by refusing to accept it any longer.

Moreover, we can use names to talk about norms (or rules), without making ontological commitments to them. To illustrate, consider the rule of modus ponens, according to which we can infer “q” from two premises “if p then q” and “p.” Note that we can use the name “modus ponens” to talk about this rule of inference. So, we may say that “modus ponens” is a genuine name that we can use to talk about a certain rule of inference.¹⁸ But, to say that we ought to infer in accordance with modus ponens, we don't have to commit to an implausible ontology of modus ponens.

One additional thing worth mentioning is that there is no compelling reason to believe that we can create ontologically real objects by adopting certain rules. Two key points require emphasis. First, artifactualists have yet to adequately explain the mechanism through which we can create abstract entities that supposedly exist beyond the confines of space and time. Second, while we can adopt and adhere to certain norms of fiction, this does not demonstrate our ability to create fictional characters as abstract entities.

For the above reasons, we don't have to regard norms such as rules of inference as ontologically real objects. In a similar vein, we don't have to regard language norms (or rules) as ontologically real objects. And recall that when a fictional character is said to be created, what is really created is not an ontologically real object, but a set of norms about how we are to imagine the character. Therefore, my account explains why fictional characters can be regarded as creations of their authors, without requiring an implausible ontology of fictional entities.¹⁹

The fifth desideratum is the actual possession of narrated properties by fictional characters. For example, according to the Holmes stories, Holmes is a detective. So, a suitable account must be able to explain why it is correct to say that Holmes is a detective, whereas it is incorrect to say that Holmes is a ballet dancer. This desideratum is also satisfied by my account. Consider the following three statements:

1. (1) Holmes is a detective.

2. (2) According to The Hound of the Baskervilles, Holmes is a detective.

3. (3) Holmes is a fictional character created by Conan Doyle.

We can distinguish between a perspective within fiction and a perspective outside of fiction. A fictional statement like (1) is to be understood from a perspective within fiction. By contrast, a metafictional statement like (3) is to be understood from a perspective outside of fiction, that is, from a real-world perspective. To put the point another way, the name “Holmes” in (1) is used fictionally, that is, from a perspective within fiction. By contrast, the name “Holmes” in (3) is used metafictionally, that is, from a real-world perspective. Note that Holmes is understood as a flesh and blood individual from a perspective within fiction, whereas Holmes is understood as a fictional character from a real-world perspective.

On my inferentialist account, when a fictional name is used from a perspective within fiction, it is governed by norms of the following form: “Imagine that p if, according to the story, p.” As a consequence, we are prescribed to imagine that Holmes is a detective. In this sense, we can take (1) to be true as a fictional statement. By contrast, we are not prescribed to imagine that Holmes is a ballet dancer. In this sense, we can take “Holmes is a ballet dancer” to be false as a fictional statement.

Let us now turn to statement (2). According to François Recanati (2018), in addition to fictional and metafictional uses, there is the third type of use for fictional names: parafictional uses. For instance, consider statement (2), which describes something true within a fiction but is not part of the original narrative. Such a statement, classified as parafictional, exhibits characteristics of both fictional and metafictional statements. Like metafictional statements, parafictional ones, such as (2), can be evaluated as true or false from a perspective outside of fiction. However, akin to fictional statements, parafictional statements attribute properties to the putative referent of a fictional name — such as being a detective or playing the violin — that are typically associated with real individuals rather than abstract entities. Along these lines, Recanati argues that in statements like (2), fictional names are used parafictionally rather than fictionally or metafictionally.

On my view, however, we may consider parafictional statements as a species of metafictional statements. For, as pointed out above, parafictional statements, like metafictional ones, can be evaluated as true or false from a real-world perspective.²⁰ And from a real-world perspective, we can assert that (2) is true on the grounds that Conan Doyle wrote a fictional story entitled “The Hound of the Baskervilles,” in which Holmes is portrayed as a detective. In this regard, it is important to note that it is an empirical question as to whether a certain author wrote a novel in which a certain predicate is attributed to a certain fictional character. For this reason, we can consider (2) as reporting a fact about a specific work of fiction and its content. However, the statement “Holmes is a ballet dancer” does not meet the conditions to be considered such a report. Along these lines, my inferentialist account of fictional names explains why it is correct to say that Holmes is a detective, whereas it is incorrect to say that Holmes is a ballet dancer.²¹

The sixth desideratum is the unrevisable ascription of narrated properties to fictional characters. In other words, a suitable account should explain why the author of a fictional work holds the ultimate authority on the properties assigned to their characters. This desideratum can also be explained by the inferentialist account of fictional names. For example, given that Conan Doyle wrote a story in which Holmes is a detective, the meaning of “Sherlock Holmes” is partly constituted by the language-language rule: “x is Sherlock Holmes” → “x is a detective.” Therefore, as we engage with the story, we are prescribed to imagine that Holmes is a detective. Beyond this, our only choice is to engage with the story or not.

Finally, the seventh desideratum is the necessary possession of narrated properties by fictional characters. As a consequence, for example, it is not appropriate to assert that Holmes might not have been a detective. This desideratum also does not pose a serious problem for my account.

To begin, with regard to a fictional name such as “Sherlock Holmes,” we cannot do the following: by pointing to a certain fictional character around us, and saying “This one is Sherlock Holmes,” we first fix that fictional character as the bearer of this fictional name, and then start describing hypothetical scenarios, continuing to use this name to talk about the same fictional character. This is because “Sherlock Holmes” is not a name of any real object in the world, and so there is no real object that we can fix as the bearer of this name at the beginning in order to make de re modal claims about it. Of course, we can imagine a possibility that Conan Doyle might have written a somewhat different story using the name “Sherlock Holmes.” But, if such were the case, strictly speaking, we would have a different story, and so a different fictional character.

In addition, making the alleged de re modal claims about a fictional character, such as “Holmes might not have been a detective,” assumes a possibility that a fictional character can undergo changes or lose certain properties without losing its identity. But, as argued before, a fictional character can be identified by the collection of predicates that can be correctly attributed to its name. To put the point another way, the concept of a fictional character is determined by the collection of predicates attributed to its name. As a consequence, for example, being a detective is partly constitutive of the concept of Holmes. For this reason, any fictional character that is not a detective cannot be our Holmes. Hence, on my account, it is not appropriate to say of the Holmes character that he might not have been a detective.

As has been emphasized, my account holds that a fictional character can be identified by the collection of predicates attributed to its name. However, this view may raise concerns about the identity of fictional characters across different stories. What is noteworthy in this context is that both Parsons and Zalta propose similar identity conditions for fictional characters. On Parsons's neo-Meinongian theory, there is a unique object correlated with every set of nuclear properties, so that fictional objects x and y are identical if and only if x and y have exactly the same nuclear properties (see Parsons, 1980, pp. 18–19, and p. 188). And Zalta's theory of fictional objects as abstract entities holds that two abstract objects (including fictional objects) are identical just in case they encode exactly the same properties (see Zalta, 1983, p. 13, and p. 93).

But the view that a fictional character can be identified by a collection of properties runs into the following problems raised by Thomasson (1999, esp. pp. 56–57). On neo-Meinongianism, the identity of fictional characters remains unaffected by the circumstances of their creation. Therefore, fictional characters with the same properties are considered identical, even if they were created independently by different authors.²² In addition, fictional characters that appear in different stories would not be identical because they are likely to have some different properties. Moreover, if an author makes even a slight alteration to their draft, a character being depicted would change, leading to the unnecessary proliferation of fictional entities. Thomasson argues that these consequences are very counterintuitive. But these objections do not pose any serious problem for my Leibnizian antirealist account of fictional characters. Let me explain.

First, it is possible that a fictional character with the same properties is created independently by different authors. On my account, whether there is a fictional character x depends on whether there is a fictional story that depicts x. And this does not rule out a possibility of a different fictional story being written, featuring a character with the same properties as x. In this scenario, we can say that the same fictional character appears in two different works of fiction. And when we engage with either work, we are prescribed to imagine the fictional character x in the same way. This situation is similar to a case where different societies adopt the same law at different times. In such a case, although these laws have different origins, they can still be considered to be the same law. Therefore, even if two characters are created by different authors, it does not necessarily mean that they are different fictional characters.

Second, my account denies that fictional characters can persist across different stories. Nevertheless, it can explain the intuition behind this view by appealing to a counterpart concept relation. To illustrate, consider again the character of Holmes as depicted in A Study in Scarlet and A Samba for Sherlock, respectively. The latter portrayal lacks some characteristic properties that can be correctly attributed to the former. This raises the question of whether these two characters are really the same character. It is important to observe at this point that, for any property Conan Doyle attributed to Holmes in his fictional story such as A Study in Scarlet, the same author or a different author can write a new fictional story in such a way that a fictional character with the name “Sherlock Holmes” lacks the property in question. Consequently, we can hardly argue that there are some essential properties that make two fictional characters in different stories one and the same character. Another thing worth noting in this context is that an author can create a fictional character in such a way that the distinction between necessity, possibility, and impossibility is not strictly kept. For example, as in The Metamorphosis written by Franz Kafka, if an author wants, they can write a story such that a human being is transformed into a non-human creature. This means that there is no real distinction between essential and accidental properties for fictional characters. Therefore, we can hardly provide a plausible cross-work identity criterion for fictional characters.

If, strictly speaking, there is no cross-work identity for fictional characters, what kind of relation holds between fictional characters that seem to reappear in different stories? According to Sellars (1974), our concept (or meaning) can undergo a change. For example, the concept of mass has changed during the transition from Newtonian mechanics to relativistic mechanics. In this case, the Einsteinian concept of mass is not simply other than the Newtonian concept of mass; for Newtonian mass and Einsteinian mass are so functionally similar that they can be regarded as varieties of mass. Along these lines, we may argue that these two concepts are closely related counterpart concepts. And it is due to this counterpart concept relation that we may say that the concept of mass underwent a change from Newtonian mass to Einsteinian mass, rather than saying that an old concept was simply replaced by a wholly different concept. On my view, this Sellarsian view of the counterpart concept could be extended in an analogous manner to fictional characters that seem to reappear in different stories. Therefore, we can say, for example, that the Holmes character in A Study in Scarlet bears a counterpart concept relation to the Holmes character in A Samba for Sherlock, although, strictly speaking, there is no cross-work identity between these two.²³

Along the above lines, we can argue that two fictional characters are identical only if they share all their properties, and so the Holmes in A Study in Scarlet and the Holmes in A Samba for Sherlock are, strictly speaking, different fictional characters. Instead, only a counterpart concept relation holds between these two. The same point applies to fictional characters belonging to a group of stories written by the same author. Therefore, we can say that there is no cross-work identity between the Holmes in A Study in Scarlet and the Holmes in The Hound of the Baskervilles. Instead, the former bears a counterpart concept relation to the latter. Therefore, on my account, if two fictional characters from different stories exhibit slight differences, they can be considered to have a strong counterpart concept relation. And due to such a strong counterpart concept relation, we may consider these fictional characters to be the same character in a loose sense, especially when these characters appear in a series of works by the same author.²⁴

Third, and most importantly, my account embraces antirealism concerning fictional characters, and hence it does not face the problem of unnecessary proliferation of fictional entities from the outset.²⁵

For the above reasons, my view that a fictional character can be identified by the collection of predicates attributed to its name does not generate any serious problem.

5. Concluding Remarks

In this article, I have presented a Leibnizian antirealist account of fictional characters that satisfies the seven desiderata for a suitable account of fictional characters and avoids the problems with Voltolini's syncretistic account.

On my Leibnizian antirealist account, a fictional character can be identified by the collection of predicates attributed to its name. So, this account must address the following two key questions: What binds some predicates together to make up a fictional character? And how can a fictional character, as understood by this account, serve as a subject of predication? My account can successfully address these two questions on the basis of the Brandomian anaphoric theory of reference discussed in Section 3.

To begin, anaphoric relations between tokens of a fictional name can play the role of binding various predicates together into predicates that can be correctly attributed to the same name. In addition, the anaphoric theory of reference also explains why fictional names can serve as subjects of predication. Therefore, the most important merit of this account is that it offers the benefit of the bundle theory by circumventing the issue of bearers of fictional names, while simultaneously avoiding the main problem of this theory by explaining why a fictional character is more than just a collection of predicates.

Another important merit of my account worth emphasizing is that it explains why fictional characters can be considered to be the creations of their authors without being committed to an implausible ontology of fictional entities. What should be recalled in this regard is that if a work of fiction written by an author has been successfully published, certain norms of fiction are thereby established, and those norms are such that, if one engages with a fictional story, one is thereby prescribed to imagine as the story says. In such a case, the things that are really generated are not ontologically real objects, but rather a set of language norms governing the correct use of fictional names. And we don't have to commit to an implausible ontology of fictional entities, because the meaningfulness of their names does not depend on the existence of their bearers.