Truth Without Reference: The Use of Fictional Names

Abstract

Singular terms without referents are called empty or vacuous terms. But not all of them are equally empty. In particular, not all proper names that fail to name an existing object fail in the same way: although they are all empty, they are not all equally vacuous. “Vulcan,” “Jacob Horn,” “Odysseus,” and “Sherlock Holmes,” for instance, are all empty. They have no referents. But they are not entirely vacuous or useless. Sometimes they are used in statements that are true or false. We are basically referentialists about proper names. The ordinary semantic function of a proper name is to refer to an object, and to do it directly, that is, without semantically providing any identifying condition that the object should meet to be the referent. To put it differently, we agree that statements containing proper names express singular propositions, i.e., that their truth-conditions involve the referent of the proper name, if it exists, and not any identifying condition of it. Now, since empty names lack a referent, and therefore would not express such a singular proposition, how do we explain that many, if not all, statements containing them have a truth-value? Answering this question for the case of fictional names, in particular, is the aim of this paper.

1 Introduction

Singular terms without referents are called empty or vacuous terms. But not all of them are equally empty. In particular, not all proper names that fail to name an existing object fail in the same way: although they are all empty, they are not all equally vacuous. “Vulcan,” “Jacob Horn,” “Odysseus,” and “Sherlock Holmes,” for instance, are all empty. They have no referents. But they are not entirely vacuous or useless. They convey information. They are empty in different ways, and thus some of them at least are not-so-vacuous. Consider the following statements:

(1) Vulcan causes the perturbations of Mercury’s orbit.

(2) Odysseus was set ashore at Ithaca while sound asleep.

(3) Zeus was chief of the gods.

(4) Jacob Horn was an important person in Colonial America.

(5) Sherlock Holmes lived in 221B Baker Street, London, England.

Any of these might appear on a true/false exam, in the appropriate class. In an Astronomy Class, the right answer for (1) would be “F.” In an American History Class, the right answer for (4) would be “F.” But in Literature or Classics courses, the right answer to (2), (3) and (5) would be “T.” This doesn’t settle the issue of whether statements with empty names have truth-values. But it shows that there are correct and incorrect answers, and that the mere fact that the names contained in them are empty is not enough to decide on this issue.

Let us analyze each statement in a bit more detail. (1) is a scientific hypothesis about a planet called “Vulcan,” which, as it turned out, doesn’t really exist. (1) is nowadays taken to be false, and the discovery of its falsity is a piece of astronomic knowledge. Though empty, “Vulcan” is not so vacuous. One might agree that empty names have no place in science, if by “science” one means “finished science.” But they play key roles in scientific practice and they have played important roles in the history of science.

(2) belongs to Homer’s poem and is one of Frege’s famous examples. He says that “it is a matter of no concern to us whether the name ‘Odysseus’ … has a Bedeutung, so long as we accept the poem as a work of art.” But, if the name is empty, then (2) lacks a truth-value: “…since it is doubtful whether the name ‘Odysseus’ … has a Bedeutung, it is also doubtful whether the whole sentence does” (Frege 1997 [1892], p. 157). Frege takes (2), and any simple statement containing empty names, to be neither true nor false. Russell takes them to be false.

Things are a bit more complicated. Our intuitions about the truth-value of statements containing empty names diverge. Sometimes, we tend to agree with Frege that they lack a truth-value, but often we tend to take them as true or false, depending on whether the names purport to refer to actual, mythological or fictional “objects.” The scare quotes are due, of course, to our resistance to calling the referents of fictional or mythological names “objects.” In our view, all these names are empty, i.e., there is no object to which they refer, but from that it does not follow that all statements containing them are false, or that they lack truth-value.

Take (3). We are now fairly confident that “Zeus” does not refer to an existent object, but presumably many ancient Greeks thought it did, and believed that (3) was true. Once we identify Zeus as a myth and “Zeus” as empty, however, we generally keep the intuition that (3) is nonetheless true, because it is an accepted part of the myth. So, in contrast,

(6) Aphrodite was chief of the gods,

is false. In other words, in a Classics class “T” would be the right answer for (3) and “F” for (6). Actually, things can get even more complicated. While in a Classics class “T” might be the right answer for (3), in a Basic Dogma class in a Catholic High School, the right answer would be “F.”

That is not what happens with (4). With the publication of The Horn Papers by William Horn in 1945 as the diary of Jacob Horn,¹ people came to believe (4) to be true. But, when it was discovered that the diary was a hoax, that Jacob Horn didn’t actually exist, everyone took it not only to be untrue—for lack of a truth-value, as Fregeans would conclude—, but false. They realized they had been fooled, and that what they believed to be true was not.

Again, with (5), things are quite different. We all know that Sherlock Holmes is a fictional character created by Arthur Conan Doyle, and that there was no actual person to whom the name “Sherlock Holmes” refers. Our intuition, however, is that (5) is true, while (7), for instance, is false:

(7) Sherlock Holmes lived in 221B Boulanger Street, Paris, France.

Fictional names—like “Sherlock Holmes”—seem to have more in common with mythological names—like “Zeus” or “Aphrodite”—, than with hoax names—like “Jacob Horn.” At least in the sense indicated here. Finally, all seven statements above differ from,

(8) Ignatius P. Thunderbold had five arms.

We just made this name up. It isn’t part of a myth or a story or a legend or a fraud. It is not even part of a story we are telling. You might suppose that the sentence must be correct, since we made up the name, and we said it. But we didn’t assert it. It’s just an example of a completely and utterly vacuous name. As such it differs from the ones in the list. There is no standard for correctness, there is no class where an answer to (8) might be correct or incorrect.

We are basically referentialists about proper names. The ordinary semantic function of a proper name is to refer to an object, and to do it directly, that is, without semantically providing any identifying condition that the object should meet to be the referent. To put it differently, we agree that statements containing proper names express singular propositions, i.e., that their truth-conditions involve the referent of the proper name, if it exists, and not any identifying condition of it. Now, since empty names lack a referent, and therefore would not express such a singular proposition, how do we explain that some, if not all, the statements (1)–(7) have a truth-value? Answering this question is the aim of this paper.

We begin with a preliminary discussion of fictional and non-fictional names in Sect. 2. Then, in Sect. 3, we present our account of ordinary uses of ordinary (non-fictional, non-mythological, non-empty) names, which is based on Critical Referentialism (Perry 2012a, b [2001]), and Critical Pragmatics (Korta and Perry 2011). In Sect. 4, we elaborate the explanation to cover the cases of ordinary uses of empty names. In Sect. 5, we present our account of fictional names, and, in Sect. 6, we conclude.

2 Fictional and Non-fictional Names

To repeat, not all empty names are equally empty. For present purposes, we will take fictional names to be those proper names that are initially used in works of fiction to fictionally refer to particular people, places, objects and so on, and later to talk, outside fiction, about those fictional people, places, objects and so forth.² Our particular focus will be on the uses of fictional names in statements like (2), (5) and (7) above, which purportedly express fictional truths (or falsities), addressing mythology, frauds and empty names in ordinary discourse only as important preliminaries.

A fictional truth concerns a fictional story and the facts according to the story. These facts can be explicit or implicit in the story. We call statements about fictional facts made, so to speak, outside fiction parafictional statements. They are actual statements made by actual speakers (or writers) about fictional states of affairs, i.e. about states of affairs explicitly or implicitly depicted in a work of fiction; states of affairs that are part of the story told. Parafictional statements are not fictional statements. Neither (5) nor (7) are lines from any of Sherlock Holmes’s stories. Watson never uttered (5) or (7), neither as narrator nor as a character. No other character or narrator ever fictionally uttered (5) or (7).

We call the statements (or pseudo-statements, or pretense-statements, if you prefer) made by the author, the narrator or the characters of a work of fiction fictional statements. Fictional statements are part of the work of fiction; parafictional statements are not; they designate fictional states of affairs that belong (or not) to the fictional story.

Any ordinary (full or empty) proper name, say, “Gaius Julius Caesar” or “Vulcan” can become fictional, if it’s used in a work of fiction to identify a fictional character or a planet, for instance. That is, if they are used in fictional statements. And then, exploiting what we call a name-notion-network of co-reference, we use those names in statements about the events in the work of fiction, that is to say, in producing parafictional statements.

Consider how the proper names in a very well-known fiction were first introduced. The first mention of “Sherlock Holmes” is in the title of Chapter I (“Mr. Sherlock Holmes”) of A Study in Scarlet, whose Part I is presented as a non-fictional diary of John H. Watson, M.D., setting the latter as the narrator.

The first mention of Baker Street is part of a fictional utterance by Sherlock Holmes in his first conversation with Watson, in their first appearance in a fictional story:

“I have my eye on a suite in Baker Street,” he [Holmes] said, “which would suit us down to the ground.”

(Conan Doyle 2010 [1887], p. 4)

The narrator—Watson—says a little later

We met next day as he had arranged, and inspected the rooms at No. 221B, Baker Street, of which he had spoken at our meeting.

(Conan Doyle 2010 [1887], p. 4)

London appears in the first sentence of Chapter I: “In the year 1878 I took my degree of Doctor of Medicine of the University of London,” (Conan Doyle 2010, p. 1).

This is how the fictional names “Sherlock Holmes” and “[Doctor] John H. Watson” are introduced. As names for fictional characters made up by Arthur Conan Doyle,³ whose adventures will be narrated in a number of novels and short stories from 1887 to 1927.⁴ In our view, this is the beginning of a wide network of what we call “coco-reference” that reaches to the present day.⁵

There are a huge number of novels, movies, and TV series based on these two characters, with the very same names and—roughly—similar features. One might wonder whether the uses of “Sherlock Holmes” related to, for instance, the recent BBC series do co-refer with the uses of the name in statements about Conan Doyle’s stories; that is to say, whether we are talking about the same fictional character or not. We are not going to discuss this question in detail, but our intuitions lead us to suspect that, despite their many similarities, in these particular examples the two Sherlock Holmeses are two different fictional characters. For instance, the parafictional statement.

(9) Doctor Watson and Mary Morstan got married in 1889,

is true if the names correspond to the characters in Conan Doyle’s stories, but false if they correspond to the characters in the BBC series, which situates the action in the twenty-first century. To be sure, this is not a definite argument for the “two-character” view, but it certainly triggers intuitions in its favor.

Coming back to Conan Doyle’s stories, are “221B Baker street” and “London” fictional names? Are they empty, or do they actually refer to actual London, and actual 221B Baker Street? In this respect, we side with Frege (1997 [1897]) and against Kripke (2013): we think they do not refer to actual cities and streets.⁶ And, consequently, “London” and “221B Baker Street,” both in parafictional statements like (5) and in fictional statements uttered by fictional characters, are as fictional as “Sherlock Holmes.” The fact that there was and still is a real city called “London”, in a real country, and in the real world, doesn’t mean that this city is the one referred to by all uses of “London”, and it certainly doesn’t mean that real London is the city referred to with the use of “London” in Conan Doyle’s stories or in parafictional statements like (5). Similarly, the existence of a real person called “Sherlock Holmes,” if there is one, would not make all uses of “Sherlock Holmes” refer to this real person and it wouldn’t mean that all uses of the name in Conan Doyle’s stories refer to this real person.

To be sure, at the time when Conan Doyle published the stories, addresses in Baker Street did not go as high as 221, so the name “221B Baker Street, London” in (5), if taken as an ordinary non-fictional name, is empty at the relevant time, so (5) should be deemed false. In any case, even if there was such an address in the real world, the idea of a fictional character living in an actual flat in London is nonsense.⁷ In fictional stories and parafictional statements, then, we assume that proper names are about fictional people, places, pets, et cetera.⁸ Of course, it remains to explain how this works, what fictional characters are, and what we mean by “aboutness” if the names are empty.

We should distinguish the following kinds of statements:

• (i) ordinary “literal” statements containing proper names that are purported to refer to actual objects,⁹ and whose truth concerns actual facts;

• (ii) fictional statements by authors, narrators and characters in works of fiction;

• (iii) parafictional statements that concern the fictional facts; and

• (iv) metafictional statements. These are ordinary statements about actual facts about the fiction. For instance,

(10) Sherlock Holmes is a fictional character created by Arthur Conan Doyle.

This is a true statement because it corresponds to actual facts. We have to be careful to make these distinctions appropriately, because sometimes our intuitions about the truth-value of these kinds of statements depends on whether we interpret them as ordinary, fictional, parafictional or metafictional. Take

(11) Sherlock Holmes does not exist.¹⁰

If taken as an ordinary statement, it seems perfectly true, but if considered as a parafictional statement, it is false—Sherlock exists and lives in 221B Baker Street, in the fiction. And perhaps it is also false as a metafictional statement—Sherlock, the fictional character, exists since it was created by Conan Doyle. Actually, things are more complicated, since we can have ordinary/parafictional/metafictional mixed statements. Take

(12) Sherlock Holmes does not exist—he’s a fictional character.

In the most charitable interpretation, the first statement should be taken as an ordinary statement while the second should be interpreted as metafictional. But, if taken as parafictional, for instance, they would be both false. Eventually, a full account of fictional names would have to cover all these cases,¹¹ but we’ll concentrate on their use in parafictional statements.

Also, notice that all these distinctions should also apply to mythology, that is, to ordinary vs. mythological vs. paramythological vs. metamythological statements. The main difference is that mythological names, at least in many cases, were once thought to name existing objects. But this should not be too relevant for our present purposes. In any case, to introduce our take on fictional names, we have to start discussing the case of ordinary non-empty names.

3 Ordinary Non-empty Names

As referentialists, we assume that an utterance of

(13) Cicero was a Roman Senator,

is true if and only if a certain person, the referent of the use of “Cicero” in (13), had the property of being a member of the Roman Senate. These truth-conditions involve an existing object (a certain person) and a property (we can ignore tense and time issues for present purposes), and, using common philosophical parlance, we can represent it as a singular proposition; a proposition that, if true, would remain true whatever the name of the object, and whether or not (13) was ever produced as an utterance. Philosophers often represent this proposition as

(14) <Roman-Senator, c>,

and say that (13) expresses it.

Notice, however, that this way of putting things might obscure the fact that different kinds of truth-conditions are in place. For (13) to be true the following conditions must obtain:¹²

E (existential condition). That c exists.

R (reference condition). That the use of “Cicero” in (13) refers to c.

S (satisfaction condition). That c is a Roman senator.

The satisfaction condition S, which is basically equivalent to the proposition (14), requires that “Cicero” refers to c (R), which, in turn, requires that c exists (E). Distinguishing among these sorts of truth-conditions is critical, for instance, when we want to distinguish between the truth-conditions of (13) and (15):

(15) Tully was a Roman senator.

As referentialists, we acknowledge that (15) also expresses the proposition (14), with object c and property Roman-senator as constituents. In other words, the same satisfaction condition S applies. The existential condition E is also the same—c must exist. But the reference condition is different. R puts conditions on the use of the name “Cicero” but the truth of (15) doesn’t require anything from the name “Cicero.” Rather, what we find in (15) is a condition on the use of the name “Tully”, i.e. that it must refer to c.

Acknowledging this is not tantamount to the abandonment of referentialism about proper names; we are not assuming that an identifying condition is associated with names as part of their meaning. We are just assuming that proper names are not semantically associated with a single actual object. Following Strawson, we claim that proper names do not refer, people do, by using names and other expressions (indexicals, demonstratives, descriptions), or even with no expressions (with referents as unarticulated constituents).¹³ But of course, people err, and sometimes, even when we intend to refer, we do not do so. So, in our view, the reference condition is not straightforwardly given for any use of a name. But then, what are the conditions for reference?

Our complete picture of reference by proper names is Perry’s conception of notion-name-network, which is inspired in turn by Chastain’s (1975), Donnellan’s (1972), Evans’s (1973) and Kripke’s (1980) view of chains of reference. In this paper, however, we skip many details of the proposal, to focus on the most relevant ones for present purposes: the concepts of a network and a notion.¹⁴

3.1 Networks

Proper names are easy to use. We don’t have to have much information about the referent to be able to refer to it using a name. Except for the few occasions in which we choose a name for a child, a pet or a nickname for a friend, we rely on previous uses of the names we heard. And learning the name is precisely a way to know about it by asking questions or googling.

Imagine María tells you,

(16) Juana is coming,

you have no idea who she is referring to, apart from the fact that she’s probably referring to a female person.¹⁵ Regardless of this, if you want to get more information and you ask “Who is Juana?,” you are referring to the same person that María is referring to, assuming she exists. If María answers “My mum,” then you acquire some information about her, for instance, that she is alive. According to Critical Referentialism, your utterances of “Juana” refer, co-refer, and coco-refer (that is, conditionally co-refer, see Korta and Perry 2011), because with your use of “Juana” you intend to refer to the same person that María’s use of “Juana” refers to, assuming again that there is such a referent.¹⁶

These relations of coco-reference form a network of coco-reference for the name, a convention that by coco-reference allows one to refer to the origin of that network.¹⁷ When her parents decided to name María’s mum “Juana,” they established a convention that allowed speakers to refer to her with that name, creating a network of coco-reference (and co-reference, in this case) with her as origin. But what happens with empty names? That is, how do we account for networks that do not have an object at their origin?

Consider the following scenario. John tells you (4) “Jacob Horn was an important person in Colonial America.” You don’t want to show your ignorance about US history—while you show, without noticing, your ignorance about Donnellan’s 1974 paper—so you just smile and go to Wikipedia. You are re-directed to “Jakob Horn”, (https://en.wikipedia.org/wiki/Jakob_Horn, accessed August 9, 2017), who is said to be a German mathematician, famous for his Horn functions. It rings a bell from your Logic courses. But that’s not the guy John is referring to by his use of the name. You somehow arrive at the entry for The Horn papers, where you read that

The Horn Papers were a genealogical hoax consisting of forged historical records pertaining to the northeastern United States for the period from 1765 to 1795. They were published by William F. Horn of Topeka, Kansas between 1933 and 1936, and presented as a transcription of documents of his great-great-great grandfather, Jacob Horn (died 1778), and other members of the Horn family. (https://en.wikipedia.org/wiki/Horn_Papers, accessed August 9, 2017)

This is the Jacob Horn John is talking about or, rather, the one he is not talking about, given that this Jacob Horn never existed. So perhaps John is teasing you, or wants to discuss Donnellan’s paper, or both.

This case illustrates two things. First, that we have many networks associated to the same name. There are many Juanas in the world. We can use “Juana” exploiting the particular network of coco-reference that has María’s mother as origin, or a different one, picking a different referent for our utterance. In other words, proper names are nambiguous, and it is a matter of a speaker’s intentions which network she is trying to exploit. Consequently, we can use “Jacob/Jakob Horn” exploiting a network of coco-reference with the German mathematician as origin and, thus, refer to him; or we can exploit another network, coco-referring with John’s use in (4), and fail to refer to anyone. And this is the second point. The second network, exploited by John, has no origin or, as Donnellan (1974) puts it, ends in a block. In other words, it is an empty name. In this case we have a network of coco-reference, with utterances that do not co-refer, because they don’t refer. And this leads us to the main theme of the paper. Before elaborating our account of empty names, though, let us briefly present our account for the name “Cicero”, as used in (13) above.

Simplifying a bit, we can assume that there is a wide network of coco-reference of the name “Cicero,” with the Roman citizen Marcus Tullius Cicero (106 BCE–43 BCE) as its origin. Call it N_C. There is a smaller network for the name “Tullius” (sometimes anglicized as “Tully”), with the same individual as its origin. Call it N_T. They are not the only networks associated to these names, however. Both “Cicero” and “Tullius” are nambiguous. The speaker may intend to refer to his brother Quintus Tullius Cicero, for instance, via smaller networks with Quintus as its origin. Leaving this complication aside, we can elaborate the reference condition R of (13), making explicit the conditions on reference via networks, as follows:

RN (reference-network condition). (i) That there is a network N_C such that the use of “Cicero” in (13) exploits N_C; and (ii) has c as its origin.

The E and S conditions are the same as before. In the case of (11), we have a different name, “Tully” and a different network, N_T, which explains why, even if (13) and (15) are true if and only if Roman-senator, c is a fact, a competent speaker/hearer of English might believe that one is true and not the other. They share the same E and S conditions but they have different R conditions that, though true, might be unknown to such a speaker/hearer. To complete our picture of reference and cognitive significance, we need to invoke another important concept of Perry’s (2012b [2001]) critical referentialism: notions.

3.2 Notions

Notions are mental files that store information (and misinformation) about the individuals we think and talk about. The three of us have our own separate notions of Cicero, which were born when each of us was separately—at different times on different continents—reading Ancient Roman history. Notions include ideas, and our notions of Cicero probably include similar ones, that is, ideas of being a powerful senator and brilliant orator. Those of us who took Latin courses would have richer notions, and might have later included in the notion the famous lines “Quo usque tamdem abutere, Catilina, patientia nostra…” that reportedly open his speech against Catiline.¹⁸ But not all of us had included them in our Cicero notion, not until now, at least. Anyway, for one of us to utter (11), s/he doesn’t need to have a rich notion of Cicero; proper names are especially apt tools to refer to individuals with whom we do not have any opportunity to interact directly, because they are dead, like Cicero, or because they are remote in space, like Venus.

Other singular terms are particularly apt for referring to individuals, when we have a specific kind of notion that involves particular cognitive fixes of them: demonstratives, for instance, are suited to exploit perceptual information; indexicals to exploit utterance-bound fixes. Proper names are paradigmatically associated to standing notions that participate in intersubjective notion-networks.¹⁹ Notion-networks provide a natural account of the cognitive significance of statements containing proper names. They explain a speaker’s particular choice of a singular term (a demonstrative over a proper name, e.g. “she” instead of “Livia”; a proper name over another, “Tully” instead of “Cicero”), and they explain also why a hearer can get different information from utterances of different co-referential terms. Notion-networks also offer an account of how the reference, if any, of a proper name is determined: it is the origin of the notion-network. Typically, however, it is just the origin of the notion-network, if there is any, which counts for the truth of what the speaker says.

As we said above, what a speaker says when she utters (13) is the same as what she says when she utters (15), and what she says involves the origin of the name-notion-network, c, and the property predicated of it—the S condition—but not the names and the notion-networks associated to them. This is the typical case. But things are different with empty names.

An empty name is a name whose notion-network ends in a block. Perry (2012b [2001], p. 205–208) distinguishes three kinds of blocks:²⁰

(E-1) blocks that are created by perceptual experiences wrongly taken as being caused by an object;

(E-3) blocks that are created by utterance comprehension processes that interpret a referring act when there is none; and

(E-5) blocks that are freely created.

The notion-networks associated to fictional names are of the last kind. So are hoaxes like Jacob Horn. There is no wrong perception or any other kind of mistake, rather, there was an act of (free) creation of notions without origins. The case of myths is a bit more problematic. They can be taken to be cases of E-5 too, but most likely they should be taken as E-1 blocks; as cases in which perceptual experiences of events were wrongly interpreted to have “supernatural” individuals as their causes. Finally, it seems there are also cases of (E-3) blocks with myths too.²¹ The important point, however, is that when using fictional (and mythical) proper names in parafictional (and paramythical) statements, notions become exceptionally (and, perhaps, exclusively) important for their truth. Unlike ordinary proper names, fictional names are not purported to refer to their origin. They are not purported to refer to a notion either; they are not purported to refer at all.²² But a particular notion is involved in the truth-conditions of the parafictional statement such that the statement can be true or false, even if the notion has no origin, and, hence, the name is empty.

We call fictional notions those notions that are freely created by authors without an origin in a work of fiction. In this sense, what Conan Doyle did by writing the Holmes stories was to create a fictional notion associated to the proper name “Sherlock Holmes,” filling it with many explicit and implicit ideas. Once created, and because it is a fictional notion, it gets more ideas from the application of the Reality Principle, which assumes that everything that is true in reality is also fictionally true unless it is excluded (explicitly or implicitly) by the work. That is why it is fictionally true that Sherlock Holmes has a great-grandfather, even if he is never mentioned in the Holmes stories (as far as we can remember).

Fictional names are importantly different from other empty names like “Jacob Horn” and “Vulcan,” which once were taken to have a referent, but didn’t. We will start by discussing the latter before turning to fictional names.

4 Ordinary Empty Names

Le Verrier explained the perturbations of Uranus’s orbit by the presence of a planet that was unknown at the time, Neptune. The success of his hypothesis gave much credit to the hypothesis expressed by (1) “Vulcan causes the perturbations of Mercury’s orbit.” According to our previous account,²³ this hypothesis has the following truth-conditions:

E. That v exists.

RN. (i) That there is a network N_V such that the use of “Vulcan” in (1) exploits N_V; and (ii) has v as its origin.

S. That v causes the perturbations of Mercury’s orbit.

Now, the hypothesis turned out be to false. There is no such planet. There is no v, so all conditions turn out to fail.²⁴ This also means that there is no proposition <Cause-the-perturbations-of-Mercury’s-orbit, v>, but at most an incomplete or gappy proposition like <Cause-the-perturbations-of-Mercury’s-orbit, [empty slot]>. Referentialists that assume a unique proposition—the proposition expressed—to encode the truth-conditions of statements containing proper names, have a problem with this, for various reasons. But we won’t press the issue here. Let us just note here that that’s no problem for our account. The truth-conditions of (1) are complete, but, because E is not met, R and S are not met either. In our account then (1) is clearly not true, but: is it false, as Russell would conclude? Or does it lack a truth-value, as claimed by Frege?

Both options are available to our account. If we side with Frege and Strawson, we could take condition E to be a presupposition of (1), so that when the utterance fails to meet this condition, the proposition would be neither true nor false. This would accord with the idea that actually there is no proposition expressed. We would have to explain away, however, the strong intuition that (1) is false, and its negation is true. We would also need to explain away the intuition that (17) is true, and its negation false:

(17) Vulcan does not exist.

So we rather align with Russell, without needing to treat proper names as definite descriptions in disguise. (1) is false because if we take the negation of (1) as taking wide scope, it would amount to denying that the conditions E, RN and S are met; which is obviously true. With narrow scope, it would amount to asserting that E, R are met but not S, which is obviously false.

The truth conditions of negative existentials like (17) are a bit special, of course. They don’t include anything like the E condition or the S condition but just the RN. The affirmative counterpart of (17) has the following ones:

RN. (i) That there is a network N_V such that the use of “Vulcan” in (17) exploits N_V; and (ii) has v as its origin.

Now (17) denies RN, which is false, so (17) is true, its positive counterpart false. But arguably RN(i) is left intact, since the speaker of (17) is exploiting the name network of coco-reference N_V created with Le Verrier’s hypothesis, and not any other network associated to that name (like the one that is associated to a planet in the BBC series Doctor Who or in the American series Star Trek). In uttering (1) (“Vulcan causes the perturbations of Mercury’s orbit.”) or its negation, (17) or its negation, the speaker does not suspend her commitment to RN(i), the existence of a network N_V associated with the name “Vulcan” that she is exploiting.²⁵

We suggest that this commitment to the existence of a particular network of coco-reference associated to the name is generalized to all uses of proper names in statements, except perhaps in the quite extraordinary cases of initial baptism, which constitute the very creation of the network for a name.

The application of this account to the case of Jacob Horn and statement (4) above is left as an exercise for the reader. We turn now to fictional names.

5 Fictional Names

We turn back to Sherlock Holmes now. How can we take (5) to be true and (7) to be false, if Sherlock Holmes is a fictional character, i.e., if he doesn’t exist?

(5) Sherlock Holmes lived in 221B Baker Street, London, England.

(7) Sherlock Holmes lived in 221B Boulanger Street, Paris, France.

If we apply our account taking them as ordinary statements, we would render the truth conditions of (5) as follows:

E. that sh exists.

RN. (i) that there is a network N_SH such that the use of “Sherlock Holmes” in (5) exploits N_SH; and (ii) has sh as its origin.

S. that sh lived in 221B Baker Street, London, England.

There is no sh, so (5) is false, just as (1) is. This seems favored by the fact that (11) seems as true as (17) (“Vulcan does not exist”), for the same reasons:

(11) Sherlock Holmes does not exist.

(11) lacks S conditions, and it asserts that that RN(ii) condition is not met—which is true.

If we interpret (5) as an ordinary statement as above, this would be an appropriate account. (5) would be false, as false as (7)—if this is also interpreted as an ordinary statement—as long as condition E is not met, and, consequently, conditions RN(ii) and S are not met.

It is only if (5) and (7) are taken as parafictional statements that we take them to be true and false, respectively. The question now is, how does the fact that a statement is parafictional affect its truth-conditions?

A first immediate answer comes from the notions of mimesis, pretense or make-believe. According to such a view, in making a parafictional statement we wouldn’t be making an actual statement, but only pretending or acting as if we were making it. We would know that conditions E, RN(ii) and S do not obtain, but we would pretend they do. Perhaps this makes sense for fictional statements (except for the fact that we are not the authors, narrators or characters in a work of fiction). As Searle (1975) puts it, “the author of a work of fiction pretends to perform a series of illocutionary acts” (p. 325) such as statements, assertions, descriptions, and so on. But it’s not clear that it is the right approach for parafictional statements, which, to repeat, are real statements by real speakers. Do speakers of parafictional statements pretend that they refer to a person, even if they are not doing so?

It seems clear that the sincere speaker of the parafictional statement (5) is not committed to conditions E, RN(ii) and S: she does not believe that sh exists, so she doesn’t believe that it is the origin of any network, or that sh lives anywhere. She is committed to RN(i)—the existence of the network N_SH associated to her use of “Sherlock Holmes”—, that’s quite clear. But what do we do with the other conditions? That is where fictional notions enter forcefully into the picture.

In our view, fictional names are not purported to refer to any individual. Any speaker or hearer who believes that the uses of “Sherlock Holmes” in (5), (7), (9), (10) and (11) are purported to refer to a real person—in other words, who takes them to be ordinary statements— is not getting at the right truth-conditions of those utterances.

Take (5). The informed speaker does not intend to refer to a nonexistent individual sh; but she is assuming the existence of a network N_SH and a notion n_sh at the end of the network (just before the block), associated to the name “Sherlock Holmes”, such that the notion includes the idea expressed by the predicate “lived in 221B Baker Street, London, England.” Or, to put it differently, the truth-conditions for (5) are the following ones:

E. That a notion n_sh exists

RN. (i) That there is a network N_SH such that the use of “Sherlock Holmes” in (5) exploits N_SH; and (ii) has no origin, but ends in the notion n_sh.²⁶

S. That the notion n_sh includes the idea “lived in 221B Baker Street, London, England.”

As we saw, the fictional notion n_sh includes the idea “lived in 221B Baker Street, London, England.” The information is given by Doctor Watson himself playing the role of the (fictional) narrator.²⁷ That’s why we take the parafictional statement (5) to be true. And this is why we take (7) to be a false statement: the S condition for (7) is not met, n_sh does not include “lived in 221B Boulanger Street, Paris, France.” It does not include it, not because Conan Doyle or anyone in the works tells us otherwise, but because it contradicts what is explicitly true in the stories, that is to say, (5).

Take now the case of (10), “Sherlock Holmes does not exist.” If we interpret it as an ordinary statement, its truth-conditions are just the RN-conditions.

RN. (i) That there is a network N_sh such that the use of “Sherlock Holmes” in (10) exploits N_sh; and (ii) N_sh has no origin.

which is obviously true, since the name is empty. If we interpret it as a parafictional statement, however, the name is still empty, but the fictional notion n_sh comes to the fore, and so for (10) to be true, the following truth-conditions must obtain:

RN. (i) That there is a network N_SH such that the use of “Sherlock Holmes” in (5) exploits N_SH; and (ii) has no origin, and doesn’t end in the notion n_sh.

The network N_SH does end in the notion n_sh, so (10) is false in this interpretation.

6 Conclusions

To recapitulate, our account of fictional proper names in parafictional statements takes them to be really empty: that is, names whose uses don’t refer to anything, and, in particular, they don’t refer to ficta or any other individual entity of dubious metaphysical status.

We want to emphasize that, in our view, they don’t refer to (fictional) notions either. Notions are a critical element of the truth-conditions of parafictional statements, but that doesn’t make them the referents of fictional proper names. Are we saying that what we call fictional characters are notions? Yes, we are. Writing a novel, after all, seems to basically consist in creating a bunch of characters with their properties and relations among them, in imagined places and times with imagined events happening to them. All of these are freely created notions with no origins. The work of fiction consists in those notions, but it is not about them.

Speakers using a proper name are not talking about notions or networks, but make (implicit) use of them. Knowing how to use ordinary proper names shows (implicit) knowledge of networks, notions and origins in a play of coco-referring (conditional co-referring). Using fictional proper names in parafictional statements properly is similar except that the existential condition, along with reference, is dropped and notions come to play a role in its truth.

To be sure, there is an important difference between truth in fiction (or the truth of parafictional statements) and truth tout court. Perry says that in fiction

…we drop realism and completeness—the principle of excluded middle. No one supposes there is a truth to the matter of how many hairs Sherlock Holmes had on the back of his neck on his forty-fifth birthday, or whether Holmes’s great-grandfather had more grandchildren that Professor Moriarty’s great-grandfather did. (Perry 2012b [2001], p. 231.)

That’s why he proposes to distinguish between truth, which is a matter of correspondence with (actual) facts, and accuracy or fit, which would be a matter of “agreement with the contents of canonical representations” (Perry 2012b [2001], p. 231). In the terminology we use here, we would say that the truth of parafictional statements is a matter of correspondence with fictional facts, that is, facts as established by the author of the (canonical) work explicitly or implicitly via the creation of notions without origins, which are also fed by the reality assumption. Realism is dropped, that’s quite clear. That completeness is also dropped is not so clear. Arguably, Sherlock Holmes either had 10,000 hairs on the back of his neck on his 45th birthday or he hadn’t; and either Holmes’s great-grandfather had more grandchildren than Moriarty’s or he hadn’t. As parafictional statements, it is pretty obvious that the disjunctions are true; but it seems to us that the truth (or falsity) of any of the disjuncts is not only epistemically undetermined but also metaphysically so. What the consequences of these facts are is a matter we are not going to pursue in this paper.