Fictional Names and Co-Identification

2.1 Notions, Producers, and Consumers

Many philosophers - including Donnellan (1974); Perry (2001); Taylor (2003); Sainsbury (2005), (2015); Everett (2013); and Friend (2014) - have agreed that

Empty names, like referring names, are embedded in practices of communication that link uses of the names together, and it is natural to think that such communicative practices play a key role in accounting for identification and co-identification. (Friend, 2014, 308)

Among such approaches, what Friend (2014) calls “name-centric” views explain co-identification in terms of chains of communication originating in name-introducing acts, as in the tradition from Kripke (1980) and others. By contrast, “information-centric” views, in Friend’s terminology, appeal to information associated with names, following Evans (1973). Here I focus chiefly on the name-centric account of Sainsbury (2005), (2015) and the information-centric account of Friend (2011), (2014). I begin by reviewing the main points of the latter.

Friend adopts Perry’s (2001) framework of notion networks. On this view, names are associated with bodies of information called “notions”, corresponding to Evans’s (1973) “dossiers” of information and what others call “mental files”.⁶ Name use is sustained by information distributed throughout interpersonal networks of notions. Speakers in the network have notions of Oprah Winfrey comprising information about her which guide their use of “Oprah Winfrey” in constraining what they say and how they understand what others say.

Yet, on this view, reference is not determined satisfactionally to be the individual satisfying some descriptive information - as on the descriptivist view famously rebutted by Kripke (1980). Rather, as Evans (1973) suggested, reference is determined relationally as the individual that is the predominant source of the relevant information.⁷ What makes “Oprah Winfrey” refer to Winfrey is that she is the predominant source of the information in the network.

When it comes to fictional names, Friend (2014) argues that there is no source of the information in the network:⁸

In the case of Emma Bovary there simply is no dominant source […]. Instead the notion network originates with Flaubert’s freely created notion, associated with invented information, which guides Flaubert’s identification of Emma in the novel. (Friend, 2014, 323–324)

In particular, we cannot say that Flaubert is the source. Given that, for the information-centrist, reference is fixed relationally as the source of the information in the network, to say that Flaubert is the source of the information in the Emma Bovary network implies that “Emma Bovary” refers Flaubert. Analogously, suggesting that the source is the novel Madame Bovary or Flaubert’s notion of Emma Bovary both have wrong results: neither is a suitable referent for the name. Hence, the challenge is to explain why “Emma Bovary” identifies Emma Bovary, since there is no source of the kind that information-centrics point to as fixing reference or identification for our notions of her, however rich they may be.

Following Evans (1982), Friend distinguishes between producers and consumers within an information network. Flaubert was the producer of the information about Emma Bovary. You and I are consumers. There are two ways in which consumers identify fictional characters.⁹ Participating consumers “identify Emma so long as the information associated with their Emma-notions is dominantly derived from Flaubert’s mental file”.¹⁰ Parasitic consumers “simply refer to or identify whatever others in the practice refer to or identify”.¹¹

In other words, Friend allows that one may identify Emma Bovary with “Emma Bovary” even if one associates little or no information with her, for instance, if one has only overheard the name used by others. Yet such parasitic identification is purely second-hand. Parasitic speakers “only refer or identify in virtue of deference to the practice of producers and participating consumers, and here information is central”.¹² What ultimately secures identification is that the notions of participating consumers are derived from producer information.

2.2 Non-Authorial Information

When it comes to referential names, Dickie (2011) has objected to this kind of information-centrism. Consider her example:

Chaucer lived from about 1343 to 1400. He was well known in his lifetime. But in the centuries after his death […] the pool of claims made using Chaucer’s name became flooded with invented attributions of literary works to him, and fabrications about his life […]. As a result of this flood of invention, there was a period of several hundred years […] during which even Chaucer experts had ‘Chaucer’ files most of the information in which was derived from fabrications made long after Chaucer’s death. (Dickie, 2011, 53–54)

Dickie argues that this case shows that one may be a participating consumer even if one’s information is not (predominantly) derived from producer information.¹³ I agree with this verdict. The misguided Chaucer experts refer to Chaucer, and they are not just parasitic consumers in Friend’s sense. There is no non-arbitrary way of counting them as simply deferring to others, as with a name merely picked up from overhearing.

Analogously, imagine the following course of events:

Corrupted Flaubert

Flaubert’s Madame Bovary was published in 1857 and was well known in its day. But in the subsequent centuries, due to the scandal sur rounding the novel, almost all original copies gradually disappeared, and instead a flood of corrupted texts, bad plagiarisms, and so on, appeared. As a result, there was a period of several hundred years during which even Flaubert experts had “Emma Bovary” files derived from corrupted versions of the novel.

As in the Chaucer case, the corrupted speakers are participating consumers and they clearly identify Emma Bovary with their uses of “Emma Bovary”. For instance, suppose one of the corrupted Flaubert experts says,

• (3) Emma Bovary was a Parisian widow.

Clearly (3) is false, even in their mouths, the reason being that they are identifying the character from Flaubert’s original novel.

As a shorthand, call information like that associated with “Emma Bovary” in this case, non-authorial information. Cases like Corrupted Flaubert show that non-authorial information does not always undermine identification, even for participating consumers.

2.3 Origination and Chains of Communication

A natural reaction is to accept that, while associated information plausibly plays a role in guiding use, identification itself depends on, in Kripke’s words, “the actual chain of communication”.¹⁴ To a first approximation, the speakers in Corrupted Flaubert co-identify Emma Bovary due to the chain of communication leading back to Flaubert’s introduction of the name.

Sainsbury (2005), (2015) defends a name-centric view of identification of this kind. In paradigmatic referential cases, names are introduced by baptisms of individuals. Someone (so we assume) bestowed the name “Oprah Winfrey” on Winfrey, thereby initiating a practice of use, along the lines of what was suggested by Kripke (1980, 91–93). Sainsbury broadens this to include other ways of introducing names, bona fide baptisms being just one kind of originating use. For instance, someone may acquire a nickname that sticks and becomes a conventional name, even if this was unintended.¹⁵

Later speakers engage in non-originating uses “animated by some kind of conformist or deferential intention: to use the name as it is used in the relevant community […]”.¹⁶ Such intentions may be more or less explicit, and need not be directed at specific previous uses. Moreover, one need not be able to identify the originator or originating event.

Further, Sainsbury argues that

In some cases, a name is introduced when there is no bearer, perhaps as a result of error or perhaps as part of deliberate fiction. […] The intention to introduce a (new specific) name is normally successful, even if the simultaneous intention to introduce a name with a bearer is not. (Sainsbury, 2015, 200)

In the case of fiction, there is typically no intention of the latter kind. Flaubert’s act of writing Madame Bovary introduced the name “Emma Bovary”, even though he did not, and did not intend to, baptize any existing individual. Thereby, he originated the practice by which later speakers can use the name. This practice is sustained by intentions to use the name in the same way as earlier speakers, reaching back to Flaubert’s originating act of writing the novel.

This name-centric view predicts that the speakers in Corrupted Flaubert identify Emma Bovary. Even though the later speakers associate non-authorial information with “Emma Bovary”, they are nevertheless embedded in the relevant kind of communicative chains. Further, this view agrees that the information associated with Emma Bovary by different speakers varies in richness with no hindrance to identification. Two speakers may be linked to Flaubert’s origination even though one has a rich notion of Emma, while the other has little or no information about her.

2.4 Identification Shift

Parallel to the original problems brought out by Evans’s (1973) examples of reference shift, a central challenge for the name-centric comes from cases of identification shift.¹⁷ Sainsbury describes the following scenario:

Jocular Inversion

A small group, as a kind of “in” joke, decides to use “Holmes” for Watson and “Watson” for Holmes. Then all the texts are destroyed in some cataclysm. People learning from the small group don’t realize there has been a jocular inversion. These benighted souls become the only users of the specific names. (Sainsbury, 2015, 213)

As used by the jocular speakers and their successors, “Holmes” identifies Watson and “Watson” identifies Holmes. Yet the names are those Doyle introduced, and clearly there is a chain leading back to Doyle’s originating uses. Hence, the challenge for the name-centric is to allow for the shift in identification.

In response, Sainsbury distinguishes two kinds of intentions involved in name use.¹⁸ He discusses the following example from Evans (1973):

Two babies are born, and their mothers bestow names upon them. A nurse inadvertently switches them and the error is never discovered. It will hence forth undeniably be the case that the man universally known as ‘Jack’ is so called because a woman dubbed some other baby with the name. (Evans, 1973, 196)

According to Sainsbury, when the mother uses “Jack”, she has a syntactic intention to use the same name as before. Further, she has a semantic intention to use the name with the same meaning.

The former intention is successful throughout, and the name remains the same. By contrast, the semantic intention is initially unsuccessful. Initially, “Jack” still semantically refers to the original bearer, while the baby the mother is bringing up counts as the speaker referent. Yet gradually a convention emerges. The shift in reference is then explained by the suggestion that, at least in the relevant cases, “The ‘semantic reference’ of a name, as used in a community, is its conventionalized, stabilized or normalized speaker-reference in the community”.¹⁹ In other words, for Sainsbury,

the conditions on being the same name diverge from the conditions on having the same bearer. […] the causal facts that ensure identity of names may fail to ensure identity of reference. (Sainsbury, 2015, 209)

Cases of reference shift are examples of this divergence.

Similarly, on Sainsbury’s view, the Jocular Inversion case turns on a divergence between the speakers’ syntactic and semantic intentions. Syntactically they intend to use the names Doyle introduced. But semantically they intend to use “Holmes” to identify Watson, and vice versa. In turn, second-generation speakers use the names with their new conventionalized identification. While the names are the same as those Doyle introduced, their identification has shifted.

Friend objects that this view fails to account for how identification, post-shift, is fixed:

Although […] the theory tells us how an individual’s use of a name comes to have a referent (if any) - by deference to ‘conventionalised speakers’ reference’ - it provides no explanation of how the latter is determined, and thus no explanation of what ultimately determines either reference or identification. (Friend, 2014, 318)

In particular, Friend complains that, given that the name-centric distinguishes name propagation from reference-propagation, it follows that “using the same name, participating in the same practice, is insufficient by itself to guarantee that we are referring to or identifying the same thing”.²⁰ But if so, the objection goes, what does secure identification?

The name-centric should reply by agreeing that using the same name is insufficient for sameness of reference or identification. But she should deny that this means that “participating in the same practice” is insufficient for sameness of reference or identification. Given the distinction between syntactic and semantic intentions, there are really two sides to a name-using practice. Deferring syntactically to earlier uses is sufficient for using the same name but not for sameness of identification. Deferring semantically is sufficient for sameness of identification but not for using the same name.

More intuitively, the jocular speakers can be described as intending to use “Holmes” the way Doyle and subsequent speakers used “Watson”, and vice versa. That is, they have a syntactic intention concerning the name “Holmes”, which Doyle introduced. At the same time, they have a semantic intention to use that name in a particular way. Namely, the way Doyle and others used “Watson”.

This is an account of in virtue of what a name identifies what it does. As is standard, this metaphysical question should be distinguished from the epistemic question about how audiences figure out what a name identifies. The name-centric account answers the metaphysical question by arguing that the existence of a practice of using the relevant name with deferential intentions, rather than information speakers associate with the name, determine identification. At the same time, one can grant that information plays a central epistemic role. Sainsbury writes,

when we wish to know to which practice a given use of a name belongs, or what the referent of a practice is, it is rare that we can reach an answer by first identifying the baptism. Normally our evidence is associated information, even though this is evidence only, and does not make a practice the practice it is. (Sainsbury, 2005, 106)

Correspondingly, one can grant that associated information plays a guiding role for name users. Even so, the name-centric holds that identification is determined by communicative chains, irrespective of associated information, authorial or non-authorial.

We should conclude that, contrary to Friend’s challenge, the name-centric has a way of accounting for identification shifts in terms of her distinction between syn tactic and semantic intentions.

2.5 Co-Identification and Truth Conditions

The challenge concerning co-identification was to explain how disparate occurrences of a fictional name identify the same character. We have found reason to favor the name-centric view that co-identification depends on chains of speakers deferring to earlier uses. Consider again the dialogue in (1).

• (1) John: Anna Karenina is Polish.

  Sue: Anna Karenina is Russian.

John and Sue are talking about the same character because they both intend (more or less explicitly) to defer to earlier uses, ultimately reaching back to Tolstoy’s originating act in writing the novel. Name-centrism, in this sense, is a metasemantic theory of why fictional names mean what they do.²¹ Our first-order semantics and pragmatics for fictional names should reflect this metasemantics, thereby bolstering our reasons for accepting each.

Moreover, we noted that co-identification is a factor in why John is wrong and Sue is right. That is, in the truth conditions of their utterances. Recall the comparison with (2).

• (2) John: Oprah Winfrey is American.

  Sue: Oprah Winfrey is Canadian.

The truth conditions of these utterances rest on (at least) the following two facts:

• (a) “Oprah Winfrey” refers to Oprah Winfrey.

• (b) Oprah Winfrey is American.

John is right in (2) because his use of “Oprah Winfrey” refers to Oprah Winfrey, and Oprah Winfrey is American, and Sue is wrong for the same reason. Analogously, the truth conditions of the utterances in (1) rest on (c) and (d), loosely described:

• (c) “Anna Karenina” identifies Anna Karenina.

• (d) Anna Karenina (the novel) portrays Anna Karenina as Russian.

A semantics and pragmatics for fictional names should not only deliver a metasemantically plausible theory of (c), but also explain how the truth conditions of utterances like those in (1) depend on (c) and (d). This is the aim of the account I develop in the rest of this paper.

3.1 Roles and Characters

My theory of the semantics and pragmatics of fictional names relies on a metaphysical view of fictional characters. According to this view, a character is a role that can be occupied by different individuals.²² Such roles are comparable to offices like the president of the European Commission or the CEO of Apple whose occupants vary across worlds and times. I will not offer a defense of this metaphysical view of fictional characters here. But since my theory of fictional names appeals to roles and occupants, some comments on the role view are in order.²³

The role view is an instance of metaphysical realism about fictional characters. Realists hold that there are fictional characters.²⁴ One brand of realism, typically called “Meinongian”, agrees that there are fictional characters but insists that fictional characters do not exist. In other words, views labelled “Meinongian” in this area of theorizing distinguish being from existence: there are things that do not exist, such as fictional characters.²⁵ Other realists, however, reject the Meinongian claim that there are things that do not exist.²⁶ Instead, anti-Meinongian realists accept that fictional characters exist.²⁷

Role theorists are realists in that they agree that there are fictional characters, namely roles. Like many role theorists, I take roles to be sets of properties. Crudely, for instance, Anna Karenina is a role that we identify with the following set of properties:

{is called “Anna Karenina”, is Russian, is a countess, is married to Karenin, is the sister of Oblonsky, …}

So far the role realist is not committed to either accepting or rejecting Meinongianism. At least prima facie, sets of properties like the one above might be thought to actually exist or not. In the former case, the role realist will be seen as an anti-Meinongian, in the latter case as a Meinongian realist.²⁸

My arguments in this paper do not require taking a stand on this issue. Even so, my arguments do rely on the claim that there are worlds in which someone has all the Anna Karenina properties.²⁹ Yet, since the Anna Karenina role is actually unoccupied, this realist can make sense of the intuition that Anna Karenina does not actually exist.³⁰ (To be sure, someone might falsely believe that “Anna Karenina” refers to a concrete, actual individual, and they might talk as if it did. I return to this in 4.5 below.)

A world in which someone has all the Anna Karenina properties is a world in which someone occupies the Anna Karenina role. It is natural to say that they are Anna Karenina in this world, just as Ursula von der Leyen actually is the president of the European Commission in 2023. Still, there is a sense in which they are not, since no individual is identical to the role, or set of properties. Von der Leyen is not identical to the office she occupies. Whether actually existing, concrete individuals like you and me can occupy Anna Karenina at non-actual worlds depends on one’s views of transworld identity and essentialism.³¹

Qua sets, roles are abstract entities that are uncreated and eternal. Nevertheless, for the role theorist, authors can be said to introduce characters, or bestow fictionality on particular sets, by producing stories. Other realists hold that characters are genuinely created by authors.³² As we will see, my view puts emphasis on the introduction of characters, i.e. roles, by authors. Yet it does not per se turn on rejecting creationism about fictional characters, as long as something other than sets of properties could be made to function similarly in the semantics and pragmatics I propose.

It might be thought that the challenge of co-identification is dispelled if one accepts that fictional characters are non-existing objects denoted by fictional names.³³ Indeed, for the Meinongian, co-identification is just co-reference. If “Anna Karenina” refers to the non-existing object that is Anna Karenina, our different uses of “Anna Karenina” genuinely co-refer after all, and (1) is on a par with (2).

Yet it has been argued that even if there were such objects, it is far from clear that they would help explain co-identification. Friend (2014, 309) rightly notes that “the mere postulation of a realm of abstract or non-existent objects does not by itself resolve the problem of determining which such object we are talking about when using a name”. Moreover, Sandgren (2018, 723–724) argues that “there are too many exotic [i.e. non-existing] objects and not enough facts to which we can appeal when determining which exotic objects are assigned to which attitudes”.

This challenge also applies to the role realist. Simply postulating that fictional characters are roles does not explain which role “Anna Karenina” identifies. Indeed, as sets of properties, there are infinitely many roles. The name-centric’s metasemantics meets this challenge by pointing to chains leading back to originating uses, and moreover, as we will see, she argues that name-origination can be successful even if no actual individual is baptized.

3.2 Discourse Referents

The semantic-pragmatic theory of fictional names I defend here is couched within the dynamic picture of discourse information pioneered by Karttunen (1976); Heim (1982), (2002 [1983]a), (2002 [1983]b); Kamp (2002 [1981]); and others.³⁴ A central tenet of the dynamic paradigm is that understanding a discourse requires keeping track not simply of which worlds are “live”, as on Stalnaker’s (1999 [1970]), (1999 [1978]) ancestral picture, but also of a range of variables over individuals called discourse referents.³⁵

This is clearly seen in cases of anaphora. For example, confronted with (4), the listener or reader must decide whether “he” is Sam or someone else.

• (4) Sam opened the door. He turned pale.

We represent these alternatives using indices:

• (5) a. Sam₁ opened the door. He₁ turned pale.

  b. Sam₁ opened the door. He₂ turned pale.

Orthodoxy regards the indexing choices in (5a-b) as differences at the level of logical form (LF).³⁶ As such, the co-valuation of “Sam” and “He” in (5a) is fixed by the grammar.

The dynamic tradition takes indices of this kind to represent discourse referents. A discourse referent is a variable that gets associated with various information as the discourse develops. On the reading represented by (5a), (4) involves one discourse referent, labelled “1”, and conveys the information that 1 is called “Sam”, opened the door, and turned pale.

This information rules out possibilities in which no individual has all these properties. Formally, this is represented by pairing assignments of values to indices with possible worlds. So the informational content of (5a) is represented by the following set (where g is a function from indices to individuals):

• (6) {<g, w>: g(1) is called “Sam” in w, g(1) opened the door in w, g(1) turned pale in w}

(6) is the set of possibilities relative to which some individual is called “Sam”, opened the door, and turned pale. That is, the possibilities compatible with (5a).

Heim (1982) proposed that we think of the information delineated by sets like (6) as a file, understood as a collection of indexed file cards bearing entries about discourse referents. Since there is only one discourse referent in this case, (5a) determines a one-card file:

By contrast, (5b) determines a file with two cards: As I explain below, this framework provides a powerful way of understanding fictional discourse and roles.

3.3 Keeping Track of Characters

Just as in ordinary discourse, understanding a fictional text or oral narration proceeds by keeping track of discourse referents and associated information. Take the following excerpt from the opening of Dostoyevsky’s The Brothers Karamazov:³⁷

• (7) [Alexey Fyodorovitch Karamazov]₁ was the third son of [Fyodor Pavlovitch Karamazov]₂, a land owner well known in our district in his₂ own day, and still remembered among us owing to his₂ gloomy and tragic death […]. He₂ was married twice, and had three sons, the eldest, Dmitri₃, by his₂ first wife, and two, Ivan₄ and Alexey₁, by his₂ second. (Dostoyevsky, 2003 [1880], 15)

Understanding the novel requires tracking indices across the text and maintaining the information associated with each discourse referent. That is, to keep a file with information recorded on various cards.

Here is a simplified rendition of the file determined by (7):

When you encounter Alexey later in the novel, you co-index the relevant terms with earlier ones, ultimately reaching back to the first occurrence in (7). As this shows, you track discourse referents independently of tracking bona fide reference. You co index occurrences of “Alexey”, “Alyosha” and other terms like pronouns, even though you are aware that there is no actual, concrete individual as referent for the name.

I propose to understand fictional characters qua roles in terms of information associated with discourse referents introduced in fictional discourse. Take the card for Alexey Karamazov labeled “1” above. The Alexey Karamazov role is the set of properties recorded on the card at the end of the novel.³⁸ This role is the character Alexey Karamazov, which was introduced by Dostoyevsky by his writing of the text.

Information associated with discourse referents is conveyed linguistically. The four-card file above (simplistically) represents the information conveyed by the pas sage in (7) in virtue of the meaning of the words — given disambiguation, indexing, and composition. Concretely, therefore, we see fictional characters, i.e. roles, as introduced and incrementally specified by fictional works.

4.1 The Basic System

Heim’s original work ambitiously proposed that the dynamic theory replace traditional truth-conditional semantics by construing the meaning of declarative sentences as potentials for updating files.³⁹ I follow Mandelkern (in press) in imposing a division of labor between semantics and pragmatics such that the semantics assigns static, truth-conditional meanings, while updating of files is seen as a purely pragmatic matter. Below I sketch a version of this theory with some modifications to suit our purposes of developing a framework for understanding fictional discourse that facilitates a theory of fictional names.

As always, semantic evaluation depends on the context of the conversation. Here a context is a common ground, that is, what is commonly taken for granted by the participants.⁴⁰ As suggested by Heim (1982, 285–289), (2002 [1983]b, 255–256), we can think of the common ground as itself a file, called the context file. So a context c is a set of pairs < g, w >, intuitively, a file consisting of a number of cards.

Given this, we define denotations like the following (where F’ is the extension of F and “#” means undefined):

To illustrate, take (10).

• (10) He₁ is Russian.

Given (9), (10) requires that for all pairs in the context file, the referent of the pronoun, g(1), is male. This means that (10) says that g(1) is Russian while presupposing that g(1) is male. Formally as in (11).

As is standard, we take g to represent the speaker’s intentions. For an ordinary utterance of (10), the speaker has a particular individual in mind, say Alexei Navalny, who she wants to talk about. In this case g(1) = Alexei Navalny. In other cases the speaker’s intentions are less specific, as in (12).

• (12) Someone₁ left the door open. They₁ must have been in a hurry.

In (12) the speaker intends to contribute information about a discourse referent, rather than to speak of some individual she has in mind. As we said in 3.3, a discourse can accumulate information about a variable even though a real-world referent for it has not been identified, or does not exist.

So our semantics is a classical, static system that assigns 1 or 0 to sentences (if defined). Understanding 1 as satisfaction and 0 as non-satisfaction, we define truth relative to a world w and an assignment g as follows:⁴¹

Note that (13) entails that truth value gaps arise when presuppositions are not in fact satisfied, regardless of what is common ground.⁴²

For instance, even if it happens to be common ground that the referent is male, (13) does not assign a truth value to (10), unless g(1) actually is male. Where w* is the actual world, (10) has the following truth conditions:

Hence, if the intended referent of the pronoun is not male in the actual world, w*, (10) is neither true nor false at w*. Otherwise, (10) is true (false) relative to g and w* if and only if g(1) is (not) Russian in w*.

Pragmatically, utterances update common grounds, that is, context files. Generally, updating produces a new, trimmed down context (where + represents up dating):

For example, updating c with (10) means narrowing c to the pairs according to which g(1) is Russian:

Moreover, since the pronoun presupposes that its referent is male, it will be common ground that g(1) is male. Formally, g(1) will be male according to each pair in c’. As usual, presuppositions can be satisfied in two ways. It may either already be common ground that the referent of “he₁” is male, or this may be accommodated.

4.2 Fictional Names

We can now give a treatment of fictional names as variables constrained by presuppositions concerning roles. I use the notation ◦i_s◦ for the set of properties associated with i by the file s, or what is written on the i card in s. Take the simplified four card file for (a fragment of) The Brothers Karamazov above. Call this file k. Then ◦3_k◦ is the Dmitri Karamazov role, the properties associated with the discourse referent 3 by The Brothers Karamazov.

Further, I use ι[◦i_s◦]w for the unique x (if there is one) such that, in w, x has all the properties in ◦i_s◦. So ι[◦3_k◦]w is the unique Dmitri Karamazov occupant in w, if there is one. Specifically, if ∃!x(f₁(x), …, fn(x)) is false at w, where f₁, …, fn are the properties in ◦i_s◦, then ι[◦i_s◦]w has no value.⁴³ In other words, if there is no unique occupant of the role in w, ι[◦i_s◦]w is empty.

Given this, a fictional name will be a variable that presupposes that its value is the unique occupant of the relevant role. Formally, suppose the fictional name n was introduced in the file s with the index j. Then we define the denotation of n indexed with i as follows:

This means that n presupposes that its value is the unique individual who has all the properties in ◦j_s◦, or satisfies all the entries on the j card in s. By s in (16) we mean the final file, as it is at the end of the story (or text). As stated by the second clause, if this presupposition is either satisfied or accommodated, n denotes that individual. If the presupposition is not satisfied, the name has no value.

As this brings out, on this view, fictional names do not denote or refer to roles. Rather, roles constrain the values of fictional names at different worlds. Fictional names denote individuals which have certain properties at certain worlds, the occupants of the role. (I return to this in 5.2).

4.3 Using Fictional Names

In the rest of this section, I expand on this theory by showing how it treats some ways of using fictional names. Here I will distinguish fictive, factual, metafictional, and theoretical uses.⁴⁴ Take (17).

• (17) Anna Karenina is Russian.

First, (17) can be used fictively, as part of telling a fictional story. This is how Tolstoy used the sentences in writing the novel. Second, (17) can be used factually. Someone who mistakenly thinks Anna Karenina is a work of history might utter (17) as an assertion about how things actually are. Third, (17) can be used metafictionally, to say something about Anna Karenina.⁴⁵ This is how the sentences are used in our initial example of (1).

Next, consider (18).

• (18) Anna Karenina is a fictional character.

In (18) “Anna Karenina” is used theoretically, as I shall say.⁴⁶ (18) is not a metafictional statement about what happens in Anna Karenina, nor is the speaker confused about whether Anna Karenina is a real-life person.

Below I go through each in turn. The main point to note is that, on this theory, fictional names have the same meaning on all of these uses.⁴⁷

4.4 Fictive Uses

When used fictively, fictional names serve to build up information about discourse referents, as described in 3.3. Consider the following toy example:

• (19) Ben₁ fell asleep. He₁ snored.

In (19) the sentences are used to build up a file for the Ben story. Call this file b. Hence, since “Ben” originates in b with index 1, according to (16), the denotation of “Ben₁” is:

So as it occurs in the first sentence of (19), “Ben₁” presupposes that g(1) has all the Ben properties. Yet no Ben properties have been determined yet. Instead, as the first sentence of the story, it triggers various processes of accommodation. First, a discourse referent is introduced, labeled 1.⁴⁸ Intuitively, a card is opened up. Second, even before updating with “fell asleep”, various things might be added to the card by accommodation. I assume that at least “is called ‘Ben’” is added just in virtue of the occurrence of the name. Given this, the file is updated with the content of the first clause of (19).

Hence, once we get to the second clause, ◦1_b◦ includes at least “is called ‘Ben’” and “fell asleep”. Next, since “he” is co-indexed, we update the card with “snored”. So at the end of the story, we are left with a one-card file as below:

Concretely, then, the author’s fictional uses of the name throughout the story determines a role, namely the information on the card.

4.5 Factual Uses

Imagine that we have both read Anna Karenina. Yet neither of us realized that it was fictional. We both thought it was a work of history. So when we talk to each other, we presuppose that Anna Karenina actually exists. Suppose you say (17), now indexed.

• (17) [Anna Karenina]₁ is Russian.

Your utterance of (17) is factual: an assertion about the actual world. We both falsely believe that Anna Karenina actually exists, and you want to impart to me the information that she is Russian. Even so, “Anna Karenina” in your mouth does not refer to an actually existing individual the way “Oprah Winfrey” does. I follow orthodoxy in treating non-fictional names as rigid and directly referential.⁴⁹

We distinguish non-fictional names from fictional names by appealing to chains of communication.⁵⁰ (See Section 5.) Since “Anna Karenina” is a fictional name, despite what you and I falsely believe, when you utter (17), its semantics is an instance of (16). In particular, let a be the file for Anna Karenina, and suppose “Anna Karenina” in a has the index 65. So the name has the following meaning:

Accordingly, your utterance presupposes that there is a unique Anna Karenina occupant. Given our context, this presupposition is satisfied. We do presuppose that there is an actual individual who is uniquely described by the novel, just as one does when reading, say, a biography of Benjamin Franklin. So, in this sense, your assertion of (17) is successful in our context.

At the same time, your utterance should be neither true nor false because Anna Karenina does not actually exist, regardless of what we believe or presuppose. Our theory predicts this result. Consider the relevant instance of (13):

According to (23), (17) is neither true nor false at the actual world. Regardless of which assignment g you care to pick, it is not the case that g(1) is the individual who uniquely has all the properties in ◦65_a◦ in w*. There is no such individual. Hence, as an assertion about the actual world (17) is neither true nor false, even though it may pragmatically succeed in adding to the common ground, given our false beliefs and presuppositions.

4.6 Metafictional Uses

Next we turn to metafictional uses of fictional names. Imagine that Sue says (17), speaking metafictionally about Anna Karenina.

• (17) [Anna Karenina]₁ is Russian.

On the standard view, inherited from Lewis (1983 [1978]), metafictional utterances have the same content as the corresponding sentences prefixed with an operator like “In Anna Karenina”, as in (24).⁵¹

• (24) In Anna Karenina, Anna Karenina is Russian.

The theory I have sketched allows for a treatment along these lines. In particular, fictional names will have the same meaning under operators like “In Anna Karenina” as on other occurrences. To demonstrate, I will spell out a version of Lewis’s (1983 [1978]) “Analysis 1” of truth in fiction.

Let us use Ƒ_s for the metafictional operator “In s”. As before, suppose n is a fictional name introduced in s with index j. We then understand

In other words, we derive the following semantics for the metafictional reading of (17):

According to (26), (24) is satisfied at w if and only if it is satisfied by a pair <g’, w’> in the Anna Karenina file a such that w’ is closer to w than any world belonging to some pair in the Anna Karenina file by which (17) is not satisfied.

Given (26), (24) carries no presuppositions. So, for the actual world, (this instance of) our definition of truth reduces to:

(24) is true relative to the actual world, w*, if and only if there is a pair in the Anna Karenina file that makes the unique Anna Karenina occupant Russian, and whose world is closer to w* than any world belonging to a pair in the file that makes the unique Anna Karenina occupant not Russian. This correctly predicts that (17) is actually true, on the metafictional reading. Clearly many nearby worlds feature Anna Karenina occupants who are Russian, and clearly many (perhaps all) of these worlds are closer to the actual world than any world with a non-Russian Anna Karenina occupant.

It is not crucial here whether this version of Lewis’s Analysis 1 predicts the right truth values in all cases. Consider this dialogue:⁵²

• (28) Anna: [Gregor Samsa]₁ was turned into a cockroach.

  Bill: No, he₁ was turned into a beetle.

As Friend (2011) notes, Kafka’s “The Metamorphosis” tells us that Samsa was turned into a “vermin” (Ungeziefer) but does not tell us which one. As such, it is at least less clear intuitively whether Anna or Bill is right, if either of them is. Correspondingly, it is at least not intuitively clear whether what Anna said is true or false, and similarly for Bill’s utterance.

If one thinks that it is impossible for any human to turn into a vermin of any kind, both utterances in (28) are false according to (25).⁵³ One may or may not be satisfied with such results. The semantics for can be amended to fit one’s preferred theory of truth in fiction.

4.7 Theoretical Uses

Now consider again the theoretical utterance in (18).

• (18) Anna Karenina is a fictional character.

Many agree that theoretical statements cannot be treated as metafictional.⁵⁴ (18) does not make the false claim that, in Anna Karenina, the heroine is a fictional character. Rather, if one thinks that fictional characters are roles, the natural way to understand (18) is along the lines of (29).

• (29) The Anna Karenina role is a fictional character.

To implement this within our framework, we follow the standard idea that The F is G presupposes that there is a unique F and asserts that the unique F is G.⁵⁵ So we take (18) as presupposing that there is a unique Anna Karenina role, and in turn, as asserting that this role is a fictional character.

How should we understand the presupposition that there is a unique Anna Karenina role? A natural idea is to see it as presupposing that there is a unique set of properties such that anyone who has all of them is an Anna Karenina occupant. To spell this out we define an operator d835dcaf that generates theoretical readings. As for metafictional cases, will operate on the same meaning of the fictional names as follows (again where n is a fictional name introduced in s with index j):

As always, it is easiest to see how this works by examples. Here is the instance of (30) for (18):

This means that (18) presupposes that there is a unique role r such that, necessarily, someone has all the r properties if and only if they have all the Anna Karenina properties.⁵⁶ That is, necessarily, someone is an r occupant if and only if they are an Anna Karenina occupant. If this presupposition is satisfied (or accommodated), (18) says that r is a fictional character in the actual world. So we have analyzed (18) as (29).

This kind of treatment also provides satisfactory ways of understanding cases like (32) and (33).

• (32) [Oprah Winfrey]₁ admires [Anna Karenina]₂.

• (33) [Anna Karenina]₁ was created by Tolstoy₂.

(32) is the ordinary kind of statement we routinely make about people’s affections for a fictional character, and (33) exemplifies equally familiar claims concerning the origin of a character.⁵⁷

(32) will be analyzed as follows (given the obvious extension of (30) to two-place predicates):

In other words, we understand (32), roughly, as “Oprah Winfrey admires the Anna Karenina role”, which for the role theorist is to say that Oprah Winfrey admires the fictional character Anna Karenina. Similarly, we will treat (33) as the claim that the Anna Karenina role was created by Tolstoy, analogously to the semantics in (34).

5.1 Originating and Non-Originating Uses

In virtue of what does “Anna Karenina” mean what it does? I have endorsed the name-centric metasemantics according to which users of names defer to earlier speakers by intending to use the name with the same meaning, reaching back to an originating name-introduction.

The account of fictional names proposed here directly implements this metasemantic view. First, by treating originating uses as fictive utterances introducing discourse referents, and by the same token, roles. Second, by treating fictional names, on non-originating uses, as having the same semantic meaning as on originating uses. Below I offer some remarks on each of these points in turn.

For the name-centric, name-origination can be successful even if no individual is baptized. There are several dimensions of such an event, including the speaker’s intentions and authority.⁵⁸ From a linguistic point of view, the dynamic picture provides a compelling analysis of origination in this sense. Consider an example discussed by Sainsbury:

Take a case of error, Leverrier’s introduction of “Vulcan”. The originating episode perhaps started with a false quantified thought on the lines “There must be another body there affecting the orbit of Mercury”. This can sustain a grammatically singular thought, involving a definite having no bearer: “Let’s call it Vulcan”. (Sainsbury, 2015, 200)

This is a paradigm case of discourse reference.⁵⁹ The indefinite “another body” introduces a discourse referent, call it j, which is picked up anaphorically by the pronoun:

• (35) There must be [another body]j there affecting the orbit of Mercury. Let’s call itj “Vulcan”.

Consequently, there is a Vulcan card in the file for this discourse recording information like “affects the orbit of Mercury”, “is called ‘Vulcan’”, and so on.⁶⁰ Similarly, for fictional names, originating uses are fictive utterances introducing discourse referents. Sometimes a discourse referent is introduced first and the name added to the relevant card later, analogously to (35). This happens in (36):

(36) Once upon a time there was a rich man who lived happily with his wife for a long time, and they had [one little girl]j together. […] Since shej always rummaged in dust and looked dirty, they named herj “Cinderella”. (“Cinderella” in Grimm & Grimm, 2014)

Sometimes the name occurs first, itself introducing the discourse referent, as in (37), the first sentence of Alice’s Adventures in Wonderland.

• (37) Alicej was beginning to get very tired of sitting by herj sister on the bank, and of having nothing to do […]. (Carroll, 1998 [1865])

Sometimes introduction proceeds by other means.

A fictional name is originated when the name is associated with a discourse referent by a fictive utterance used to tell a fictional story — either by the name itself introducing the discourse referent, or the name being attached to a previously introduced discourse referent, or in some other way. As demonstrated in 4.4, our theory explains such originating uses of fictional names in terms of their semantics involving a role that gets fleshed out through accommodation and updating as the story progresses. In turn, the difference between fictional cases and cases of error, like “Vulcan”, is that the latter are not introduced in fictional discourse, but in ordinary, factual discourse.

Further, we have seen that fictional names have the same meaning on factual, metafictional, and theoretical uses as when they occur fictively. When used metafictionally by you or me, “Alice” has the same meaning as it has in (37). These are non-originating uses for which the speaker deferentially intends to use the name in the way of the relevant community. Such non-originating uses are not directed at files for fictional stories, and therefore do not expand on the relevant role. Rather, they presuppose the role that was introduced by the author’s originating act, or acts.

5.2 Co-Identification

Our account of metafiction bears out the observation that the truth conditions of the utterances in (1) rest on (c) and (d):

• (1) John: Anna Karenina is Polish.

  Sue: Anna Karenina is Russian.

• (c) “Anna Karenina” identifies Anna Karenina.

• (d) Anna Karenina (the novel) portrays Anna Karenina as Russian.

On this view, (c) means that “Anna Karenina” presupposes that its value is the unique Anna Karenina occupant. In other words, we see identification not as fictional names denoting roles, but as triggering presuppositions concerning roles determined by fictional works. John and Sue co-identify Anna Karenina in that both their uses of “Anna Karenina” trigger presuppositions concerning ◦65_a◦ (again, as a dummy for the actual card in the actual file). In turn, (d) means that ◦65_a◦ includes the property of being Russian. We thereby integrate the explanation of the truth conditions of metafictional utterances with the explanation of identification, as wanted.

A further issue concerns utterances of “Anna Karenina” by characters in the story.⁶¹ Imagine that the sentence in (38) appeared in Anna Karenina.

• (38) “Anna, I love you”, said Vronsky.

In this scenario (38) is uttered (written) fictively by Tolstoy. At the same time, the metafictional (39) is true about this version of the novel.

• (39) In Anna Karenina, Vronsky refers to Anna by using “Anna”.

Many other things are true in this version of Anna Karenina, as in the actual novel. Among them, that humans have kidneys and that Moscow is east of Petersburg. Perhaps it is also true in this Anna Karenina that Vronsky’s referring to Anna is underpinned by a chain of communication leading from his use of “Anna” to a baptism of Anna (within the fiction, not by the fiction). Or perhaps this Anna Karenina turns out to be a fiction in which reference works differently from how it actually works. The answer will depend on one’s views on truth in fiction.

5.3 Identification Shift and Non-Authorial Information

Consider again the case of Jocular Inversion. The name-centric maintains that the initial jocular speakers intend to use “Holmes”, the name Doyle introduced, in the way he used “Watson”. The next generation intend to use “Holmes” the way their predecessors did. Hence, “Holmes” in the mouths of both generations is linked with the Watson card in the original file, or files.⁶²

For convenience, call the Watson properties, i.e. role, ◦Watson◦. Accordingly, the first generation’s uses of “Holmes” have the meaning in (40).

Given this, we predict the correct results. For example, as used meta-fictionally by the jocular speakers and their descendants, (41) is true.

• (41) Holmes is a doctor.

Since “Holmes” in (41) is linked to Doyle’s use of “Watson”, its presuppositions concern ◦Watson◦. Hence (41) is true. This will be the result of the account of metafiction outlined in 4.6.

To illustrate further, consider (42).

• (42) Holmes is called “Holmes”.

There are different ways of reading (42). On a metafictional reading, it is a statement about the Holmes stories. Given that, as used by both the initial and the later jocular speakers, “Holmes” identifies ◦Watson◦, we predict that this use of (42) is false. Yet (42) might be read in other ways. Arguably, there is a metalinguistic reading of (42) on which it is trivially true.

This account also implies that the second-generation jocular speakers will be mistaken about the truth values of certain metafictional utterances. Ex hypothesi, the second-generation speakers are not aware that the role they identify with “Holmes” includes the property of being called “Watson”, even though, in Sainsbury’s scenario, they are aware that this role includes being a doctor. Hence, most likely, they would think that (42) is true metafictionally. Yet they are mistaken. Their uses of “Holmes” identify ◦Watson◦. So the metafictional reading of (42) is false when uttered by them. Still, it is natural to think that most of what they say, metafictionally, when using “Holmes” will be true things, such as (41).

Finally, consider the case of Corrupted Flaubert, in which speakers identify Emma Bovary despite non-authorial information. Again, suppose one of these speakers says,

• (3) Emma Bovary was a Parisian widow.

As a metafictional utterance, (3) is false, regardless of what the corrupted speakers believe about Emma Bovary. Our account predicts this result. The corrupted use of “Emma Bovary” in (3) is linked with the discourse referent that Flaubert introduced, and the name triggers presuppositions about ◦Emma Bovary◦, the role determined by Flau-bert’s original text.