The Diversity of Fiction and Copredication: An Accommodation Problem

Abstract

The paper presents an accommodation problem for extant semantic accounts of fiction. Some accounts of fiction are designed to accommodate one or another form of fictive statement exclusively, what I shall call in-fiction (roughly, statements made within fiction) and out-fiction (roughly, statements relating fictional things and events to extra-fictional things and events). Thus, typically, the accounts fail to do justice to their respective excluded form. A natural response, entertained by Kripke (Reference and existence. The John Locke lectures. Oxford University Press, Oxford, 2013) and in a different fashion by latter-day Meinongians, is to let the two different kinds of fiction have their respective accounts. It is very easy, however, to mix in and out predicates, which amounts to a species of copredication, where a single nominal receives simultaneous distinct interpretations relative to distinct predicates without creating ambiguity (Bond is a killer but remains as popular as ever). Copredication is a perfectly general phenomenon that arises for any nominal (The meal was delicious, but lasted all day/Bill read then burnt every book/Brazil is at last happy after the civil unrest). The simple problem now is that copredication appears to render all extant accounts of fiction either (1) essentially partial, in dealing with one kind of simple form of fiction but not another, or (2) merely wrong in imagining that fiction comes in categorically distinct forms, for the coherence of copredication is enough to show that this is not so. The aim of the paper is to show to resolve this accommodation problem by the radical expedient of suggesting that the fiction/non-fiction distinction is not really a semantic one, but more one between different ways in which sentences are made true or false.

1 Introduction

Let an adequate semantic account of fiction specify the truth conditions for fictive statements. A metaphysics of fiction will tell us what ficta are (maybe nothing at all), or what makes fictional discourse true or false. The present paper presents an accommodation problem for extant semantic accounts of fiction. Fictive statements come in two basic forms: in-fiction (roughly, statements made within fiction) and out-fiction (roughly, statements relating ficta to extra-fictional things and events). The accommodation problem is how to account for both kinds and their copredication in a single statement (a notion to be explained anon). Some accounts of fiction are as if designed to accommodate one or the other form of fictive statement exclusively. For example, accounts that render all fiction-involving statements false or gappy have some plausibility for in-fiction statements, but appear defeated by out-fiction. Equally, accounts that consider fictive terms to refer to artefacts appear considerably less natural when dealing with in-fiction. A natural response to such accommodation worries, entertained by Kripke (2013) and in a different fashion by latter-day so-called Meinongians, is to forgo a univocal account of fiction and settle for distinct accounts of in- and out-fiction. As intimated, though, in- and out-predications can be readily mixed up, a species of the general phenomenon of copredication, where a single unambiguous nominal receives simultaneous distinct interpretations relative to distinct predicates (Bond is a killer but remains as popular as ever) (cp., Chomsky 2000; Pietroski 2003; Asher 2011; Collins 2017). Copredication appears to render all extant accounts of fiction either (1) essentially partial, in accommodating one basic kind of fiction but not another, or (2) merely inadequate due to their endorsing the erroneous idea that fiction comes in categorically exclusive forms. Surprisingly, even though the distinction between in and out fiction has been widely discussed, copredication has received scant attention. Everett (2013) and Recaanti (2018) are exceptions.

As an alternative to the traditional options, the paper will offer a maximally simple account of the semantics of fictive statements designed to accommodate all types of fiction equally:

• (i) Fictional expressions as linguistic types or utterance tokens have no essential linguistic signature (bespoke semantico-syntactic machinery), such as uniform covert intensionality or a pretence or make-believe mode of speech.

• (ii) The truth values of fiction-involving statements depend on how the world/reality is without recourse to any ontology (ficta) one would otherwise not endorse.

• (iii) Fiction-involving statements, both in and out and their copredication, can be variously true or false

The view is inspired by Azzouni’s (2010, 2017) view that fictions can be variously true or false without an inflation of ontology. I shall approach the issue, though, by considering, as advertised, the accommodation problem accounts of fiction face, and show how this supports the maximally simple view, regardless of any metaphysical scruples.

2 The Insuperable Problem of the Diversity of Fictions

In this critical section, I shall survey major accounts of fiction and show that each one faces grave difficulties in separately accommodating in-fiction and out-fiction and their copredication.

2.1 All Fiction is False¹

This hypothesis is semantically attractive, for, if true, the category of fiction does not involve any semantic machinery otherwise unrequired—fiction devolves onto the false. It is metaphysically attractive, too, for, equally, it does not enjoin us to endorse any dubious extra-ontology. It might appear, then, that here we already have as simple account of fiction as could be desired. The proposal badly stumbles, however, when confronted with the accommodation problem.

As regards in-fiction, the hypothesis of universal falsity has an initial plausibility on two counts. Firstly, negative existentials for ficta appear to be true:

(1)	Prince Bolkonsky doesn’t exist

If the negative existentials are true, then how could in-fiction statements be true? If there are no ficta, then the stories told concerning the ficta can’t be true, or so it seems. Secondly, apparent in-fiction truths can be readily denied precisely on the basis of non-existence:

(2)	Prince Bolkonsky didn’t meet Napoleon, because Prince Bolkonsky never existed

Here it appears as if the mere use of a name triggers an existential presupposition, which can coherently—indeed, truthfully—be denied.² Without further ado, however, the universal falsity hypothesis faces two profound problems.

Firstly, there is a clear contrast within in-fiction discourse between what, at first blush, we would deem to be true and false claims:

(3)	a	Gregor Samsa turned into a giraffe
	b	Prince Bolkonsky met Mussolini

Of course, the source of the contrast is trivial enough. The cases in (3) are wrong about what Kafka and Tolstoy, respectively, say about their characters. Still, the present point is that a blanket claim that all fiction is false (Punkt!) fails to reflect contrasts between in-fiction statements, let alone explain them. (4), after all, seems at best obtuse, at worst a mistake about Tolstoy:

(4)	Prince Bolkonsky didn’t meet Mussolini, because Prince Bolkonsky didn’t exist.

Even if all in-fiction statements are false, some of them appear more false than others.

Secondly, the universal falsity hypothesis appears to be plain incorrect when it comes to out-fiction:

(5)	a	Tolstoy created Prince Bolkonsky
	b	Raskolnikov is Dostoevsky’s most famous character
	c	Humbert Humbert made Nabokov a millionaire
	d	I stayed in the building—now a hotel—where Bolkonsky first met Natasha

The crucial feature of these cases is that they do not solely concern characters and their properties, but relate them to their authors and other real-world entities. (5a), say, appears to be as true as Tolstoy wrote War and Peace.

There are variations on this basic falsity approach, such as the claim that in-fictions express gappy (but false) propositions (e.g., Kaplan 1989; Braun 1993) or have a ‘blocked’ reference, resulting in no proposition, outside of negative existentials (Donnellan 1974, pp. 20–21), or have a false metalinguistic construal (e.g., Hodgson 2018). Prima facie, however, such accounts are as ill-suited as Russell’s original proposal to accommodate out-fiction and copredication cases. Negative existentials do present a prima facie reason to endorse a falsity view. I shall seek to deflate their import in Sect. 3.5.

2.2 Fiction is Disguised Intensional Discourse³

The basic proposal here is to recognise that in-fiction statements can be true or false, even though the relevant ficta are non-extant. This is achieved by the statements in question being construed as occurring within the scope of an intensional operator, which we may render generically as According to such and such.⁴ Thus, semantically, the proposal departs from the previous account in order to distinguish fiction from the merely false, but retains a metaphysical animus towards any ontological inflation—whatever ontology the truth of fiction commits us to ought to be that we should otherwise endorse. Of course, it might be protested that we otherwise do have intensional operators available to us, so the position is not really semantically non-simple. The premise is correct, of course, but, according to the view, the fictional statements themselves are given a semantic analysis that distinguishes them from non-fiction, even if the means of doing so are not necessarily bespoke to fiction. In this sense, the account departs from the simplicity of the previous account. Thus:

(6)	a	[According to Tolstoy’s War and Peace] Prince Bolkonsky met Napoleon
	b	[According to Tolstoy’s War and Peace] Prince Bolkonsky met Mussolini

Here we see the in-fiction contrast between truth and falsehood explained; for the relevant unbracketed statements, if understood as being in the scope of the indicated operators, are not about reality (the actual world), or the world as it is anyway, but the world as it is according to some story or other. So, Prince Bolkonsky never existed, but when we utter truths about Bolkonsky we are in fact being truthful to the story; just so, false statements about ficta are marked by a lack of fidelity to the relevant story. Thus, the main clause of (6a) is understood to be true because the whole of (6a) is a rendering of what a speaker means by the main clause, and (6a) is true, whereas the main clause of (6b) is understood to be false because the whole of (6b) is a rendering of what a speaker means by the main clause, and (6b) is false. This also nicely explains why (4) is anomalous, for (7a) is true, whereas (7b) is quite confused, as if the nonexistence of the character explains something about Tolstoy’s story:

(7)	a	[According to Tolstoy’s War and Peace] Prince Bolkonsky didn’t meet Mussolini
	b	[According to Tolstoy’s War and Peace] Prince Bolkonsky didn’t meet Mussolini, because Prince Bolkonsky never existed

There are different ways of pursuing this intensional line, each of which face their peculiar problems. A complaint against the very idea, however, is that it constitutively fails to address out-fiction; indeed, Lewis (1978), who first developed the intensional view, was explicit about this neglect. For example, the cases in (5) appear to be true, but would be rendered false if construed as occurring in the scope of the relevant intensional operator. For example, (8) is false, for, inter alia, Tolstoy is not a character in War and Peace:

(8)	[According to Tolstoy’s War and Peace] Tolstoy created Prince Bolkonsky

Perforce, the account will fail on copredication cases.

2.3 Fiction is Pretence

As we saw in Sect. 2.1, the uniform falsehood view stumbles on out-fiction, but there appears to be something right about viewing in-fiction as false; after all, it is fiction! In this spirit, a ‘pretence’ account is endorsed for in-fiction by Kripke (2013), and elaborated by, among others, Walton (1990), Everett (2013), and Armour-Garb and Woodbridge (2015), although the idea perhaps goes as far back as Frege (1918/56). The lead manoeuvre is to recognise that fictions are false, if read as literal statements, while acknowledging that they appear to be variously true and false. Semantically, then, a pretence account does depart from maximal simplicity in introducing a certain pretence mode of speech or thought in order to accommodate the seeming truth of in-fiction. As with the intensional view, such a pretence mode is otherwise available to language users, but it is its special deployment in accounting for fiction that renders the account non-simple. The account remains metaphysically simple, though, for, like the other accounts, no bespoke ficta entities are posited.

The bare thought that in-fiction discourse is a form of making-as-if-to-describe various persons and events is not implausible. After all, authors and readers tend not to be confused over such matters, and, as Kripke (2013; cp., Donellan 1974) first noted, were there to be the actual persons and events described in some fiction, the fiction would not eo ipso be true. An author of fiction need not purport to be getting anything right, and invariably isn’t at the appropriate level of description. The key thing here, according to the pretence theorist, is to recognise an as if or pretend mode of assertion, in which truth and falsehood may get a grip, albeit under the cover of a pretence orientation of the speaker. It should be acknowledged that some such accounts do promise to explain the possibility of the appearance of in-fiction truth and falsehood. Still, the accommodation of out-fiction and copredicational statements appears to present a more intransigent problem.

Consider again the cases in (5). Speaker-hearers do not conceive of out-fictional statements about characters or made-up realms as being pretence or make believe in the same way they might concerning in-fiction; on the contrary, out-fictional statements are as liable to be true in as a flat-footed a way as any other factual statement concerning human creations.⁵ Without further ado, therefore, appeal to pretence/make-believe can at best only be an approach to in-fiction, not out-fiction, too, or copredication involving out-fiction.

Everett (2013), while rightly recognising the problem of copredication (he uses ‘mixed perspectives’ where I talk of copredication), thinks the pretence account easily eludes the complaint:

[Mixed perspectives] involve our participating in a pretence. In this we imagine that the world contains, along with its real denizens, the detectives portrayed in fiction, and that they have the property of being fictional together with the characteristics attributed to them by the fictions in which they occur… Within the scope of this pretence, both real entities and those from fiction count as flesh-and-blood people. They may bear the same properties (ibid., p. 176).

The general idea is that the ‘scope’ of pretence covers both in and out attributions, and so mixed perspectives or copredications are ‘straightforwardly’ accommodated.⁶ Extending pretence to cover out-fiction, however, appears to distort the content of the statements. In saying that Holmes is popular, say, one is not pretending that there is a detective along with real denizens of the world, for we can explain why what we say is true—look at the book sales, TV series, and movie franchise—without pretending that the stock of what there is has expanded. Pretence, as Kripke notes, is plausible precisely when telling a story, where all resemblance to the real world is accidental. Out-fictional statements, to the contrary, appear to be flat-footedly about the real world insofar as they are made true by it with essential reference to extra-textual facts. Likewise, if the constituent predications in copredications are treated as uniformly falling under a pretence mode, then it would be as if we are pretending, say, that a single entity is both popular and a spy or that a single entity has both found a new lease of life as a hipster in contemporary London and has a friend who served in the Second-Afghan War. Such pretences would appear to be incoherent. Everett (2013, p. 178) sounds a note of caution:

[O]ur discourse about fictional characters need [not] involve our actively or vividly imagining them to have the properties they are ascribed in the relevant fictions… But this does not alter the fact that [our] thoughts are still imaginings. Nor does it alter the fact that a great deal of our talk and thought about fictional characters involves us imagining them, in at least some sense, to be the way the relevant fictions portray them.

Let us accept that imaginings need not be fully conscious or willed. The problem I am raising is not that the relevant pretence must be sub-personal, in some sense, but that, if reflected upon, the relevant pretence would be utterly abjured, and for good reason. We don’t pretend such things as there being a person who is both popular and a spy, for we readily comprehend the difference between in- and out-fiction when thinking of what makes the relevant claims about Bond or Holmes true, even though a single nominal can support the conflicting predicates. The proposed imagining would be a poor pun, at best. So, we don’t need to pretend anything, and would reject any such pretence if offered to us. In the same way, we don’t pretend that a single thing is delicious and can last all afternoon when talking about lunch. It is this cognitive differentiation alongside a semantic univocity—Holmes or Bond is not ambiguous, just as lunch isn’t—that Everett’s uniform scope of pretence proposal elides. That said, I think Everett is right that the introduction of notions of distinct properties or predications or modes of being do not ameliorate the problem. We shall see this in the following sub-sections.

2.4 Fictions Refer to Artefacts⁷

The previous three accounts have especial problems with out-fiction, being more orientated towards in-fiction. An artefactual position is conversely orientated. Artefacts are familiar. They depend upon human cognitive activity for their design and creation but enjoy an independent life once extant, i.e., they are not mere mental entities. A tool is an artefact. So too are computer codes, sculptures, symphonies, and money. In light of the problems for alternative views of fiction, it might be that artefacts are sufficiently motley to include genuine referents for fictional terms. In this sense, an artefactual view may be seen as buying semantic simplicity—fictional truths are just like any other truths—at a minimal metaphysical cost. It might well be that we need artefacts as part of our ontology anyway, so the cost incurred is merely an addition to the ontological population rather than the introduction of an entirely novel kind of entity.

Ficta certainly fit the artefact bill to some extent. They are created by authors, and so depend upon human activity for all of their properties, but they develop a somewhat independent life, too, acquiring and loosing properties as a population’s use or understanding of the artefact changes. Ficta, in this sense, are what Thomasson (1999) calls abstract artefacts. They do not constitute anything ontologically weird or sui generis, for anything we want to say about ficta can be explained, much like any other artefact, by appeal to the human activity that creates and sustains the artefact as the thing it is (cp., Kripke 2013).

As intimated, I think the artefact view suffers from an out-fictional orientation that fails to capture in-fiction and so its copredication with out-fiction. In talking of a fictional character, we are, in a sense, talking about an artefact. Abstract artefacts are sufficiently complex entities that they can make sense of out-fictional discourse, being the kind of thing that can be created and be affected by extra-textual events. Yet when we turn to in-fiction, we appear to be talking about something other than an artefact. In telling of Prince Bolkonsky’s actions towards Natasha, such as his decision to break-up with her, and his death-bed reconciliation, Tolstoy is talking about a person, not an artefact (cp., the discussion of Everett 2013, above). To be sure, Bolkonsky wasn’t born of a woman, but via the pen of Tolstoy, but none of that is part of the story. Natasha didn’t dance with an artefact. The basic problem here is to determine which properties are apt to be possessed by an artefact and which ones aren’t. Prima facie, it is exactly the in-fiction properties that are not predicated of an artefact, unless the fiction involves an artefact, of course. Artefacts don’t fall in love, dance, realise the meaning of life, die during labour or via a stroke, and so on and so forth. On the other hand, artefacts are the kinds of things that can be created, be popular, make millions for someone, and so on and so forth. There is, it strikes me, two kinds of response a defender of ficta as artefacts might make.

Firstly, it might be thought that authors create artefacts, which can then acquire all kinds of properties given their public artefactual status. In short, right, artefacts can’t dance or fall in love, say, but they can be represented as doing so, which is precisely what Tolstoy does with his artefacts. I shall return to this issue in Sect. 2.5 to consider some semantic problems with implementing this idea for copredication cases, but we may note now that such a riposte does little to ameliorate the initial complaint. Whereas, out-fictionally, we may speak of a character, as if it had an independent existence once created, as being represented by an author as so and so or doing such and such, that is not how we understand the in-fiction statements. To think so would reduce the artefactual approach to a species of the intentional operator view. There is not the artefact Bolkonsky out there anyway, who Tolstoy depicts in various ways. Tolstoy wrote War and Peace, which makes true or false whatever we might say (in-fiction) about Bolkonsky, Natasha, etc. Thus, Bolkonsky as the out-fictional character, who is one of the romantic lead in the novel, is not represented in the novel by Tolstoy in any way whatsoever; the story is not about the made-up character, but simply constitutes that character, who can then be spoken of truly out-fictionally given how Tolstoy’s writing is embedded in the world. If that is right, then fictional artefacts are not represented by storytellers, but created by them by their writing. In response, one could think that creating an artefact and representing it as so and so are coeval, at least for abstract artefacts: the act of creation is representational. Maybe so, but such a conflation betrays an out-fiction perspective and appears not to capture the truth of in-fiction. There is also the problem of how semantically to implement the supposed representation of artefacts in in-fiction discourse, to which I shall turn, as mentioned, in Sect. 2.5.

The second response I imagine is a variation on the first response: instead of thinking of artefacts as being represented as such and such in fiction, we simply let them be such and such. Abstract artefacts, in other words, just are the kind of objects that can possess the motley attributes predicated of ficta in-fictionally and out-fictionally. The in-fiction ones are the effects of authors’ stipulations. The artefact also has out-fictional properties by its creation and subsequent social existence.

I don’t see how this variation improves the basic picture, for what we want to know is how an artefact can have or be thought of as possessing in-fiction properties in any way at all given that no artefact appears to occur in-fictionally (the artefact Holmes doesn’t occur in the stories). What we should say is that the in-fiction statements are true of, or are made true by, what is said in various stories. There are no free-floating Natasha and Bolkonsky artefacts, however, that serve as the referents of the names and who can fall in love and have children, etc., and who we understand Tolstoy to have written about. There is simply the story (however stories are constituted), which makes true or false whatever we say, in-fiction, concerning Natasha and Bolkonsky. Reciprocally, out-fictional statements are precisely those that are made true (in part) by the stories that make true the in-fiction statements. If that is right, however, then an ontology of artefacts is wholly otiose. Any statement, either in-fiction or out-fiction, will simply be made true by the relevant story or stories and their place in wider human affairs. There is no apparent need for a novel ontology of abstract artefacts with a unique suit of human and non-human properties, at least not in order to explain how fiction can be true. For each statement, it will be made true (/false) by how the world is as regards its storytelling and the social setting of the stories; abstract artefacts as extra denizens alongside other things are supererogatory beyond the stories themselves.

2.5 One Account for In-fiction, Another for Out-fiction

A natural remedy for the ills so far diagnosed appears ready to hand. The basic complaint I have made is that none of the accounts so far discussed appears to work for in-fiction and out-fiction; at best, they say something plausible about one or the other. So, why not settle for different accounts—one account for in-fiction and another for out-fiction? This, in effect, is the approach of Kripke (2013). We just addressed the option of treating ficta as artefacts, which appears most readily to accommodate out-fiction, but has problems with in-fiction. Another option, which I shall discuss in Sect. 2.6 is to treat ficta as abstract objects. For the moment, at least, I’m interested in the very idea of a dual account rather than any particular realisation of it. Still, the reader won’t go wrong if she assumes that one component of the dual account is an artefactual treatment of out fiction, just as Kripke (2013) proposes; for the abstract view promises its own peculiar take on the distinctions between fictions.

If in-fiction and out-fiction were separate discourses, as if the predicates of one discourse couldn’t co-occur with the predicates of another, then, ceteris paribus, a dual account would be attractive. The two species of fictional discourse, however, are inseparable, de jure and de facto. Such entanglement makes a dual account untenable.

The basic relevant linguistic fact is that in-fiction and out-fiction can copredicatively mingle in a statement by any grammatical means available in the language.

(9)	a	The ever popular James Bond continues to abuse women.
	b	Prince Bolkonsky is more popular than Vronsky, but he is still cruel.
	c	Humbert Humbert appears sophisticated to Charlotte Haze, but a cynical predator to us.
	d	Raskolnikov killed an old woman for no reason and set Dostoevesky up for life.

Each of the sentences, and innumerable similar cases, feature a single occurrence of a fictional name, but involve both in-fiction and out-fiction attributions. None of the statements, therefore, is solely in-fictional or out-fictional, but rather both. For example, to say that James Bond is ever popular is to make a remark upon the fiction, not what occurs in the fiction, such as his various murders, seductions, etc. (9a), therefore, says something simultaneously, as it were, both in and about the fiction. (9b) features the same mingling across the anaphora, with the first clause being out-fiction, whereas the second clause that refers back to the first via the anaphoric pronoun concerns, say, Bolkonsky’s behaviour towards his wife (Lisa) (Bolkonsky and Vronsky are not co-present in any Tolstoy fiction). With (9c), Humbert Humbert is simultaneously understood as a person viewed by another character as well as the generic reader. Similarly, (9d) speaks of Raskolnikov simultaneously in relation to other characters of the novel and to his author’s future life.

Suppose that we favoured an intensional view of in-fiction. How might that view apply to (9)? The putative operator can’t apply to the whole sentences as it is supposed to do in the simple in-fiction cases. This is because the sentences feature out-fictional predications that are not part of any fiction, so applying the operator across the whole sentence delivers the wrong truth value just as the operator applied to out-fiction alone does. Perhaps, then, the operator applies only to the in-fiction part of the sentences, leaving the out-fiction component to be accounted for in some other favoured way. Such a policy is impossible to implement. The basic reason is that there is no grammatically well-behaved way to separate in-fiction from out-fiction, such that the putative operator might only apply to the former. We may briefly see this from the above examples.

In (9a), the out-fictional predicate is an adnominal modifier (ever popular), which is constitutive of the subject of the in-fictional predicate, i.e., it is precisely the Bond who is ever popular who abuses women. There is just the one clause, and so the one subject; the putative operator, therefore, can’t selectively apply to one clause rather than another or to one argument (subject or object) rather than another. Since, of course, Bond is not ever popular in the stories (he is a spy!), the truth conditions of the sentence are not reflected by the application of the operator.

In (9b), there are two clauses. The first compares characters across War and Peace and Anna Karenina, while the second is in-fictional, concerning the actions of Bolkonsky. Perhaps, then, here at least, the intensional operator applies just to the second clause. That won’t do, however, because the second clause contains a pronoun anaphoric upon the antecedent out-fictional occurrence of Prince Bolkonsky. Thus, the same thing, whatever it is, is both cruel and more popular than Vronsky, but whoever is popular is not also cruel.

Similar considerations apply to (9c–d). In both cases, we have a pair of clauses, not linked by anaphora, but by ellipsis of the subject position, which must be construed as co-referential with the first subject. Again, though, not the one thing, at least in the present cases, can be both in the fiction (a person that seduces Charlotte Haze or kills an old woman) and be outside of the fiction as a thing that relates to us or the author.

The situation is as bleak if we go to press the pretence account into service to explain the in-fiction component of the copredicative cases of (9) (here I sideline Everett’s blanket pretence view). The problem is effectively the same as on the intensional view: it is impossible to separate what components of a statement are to count as pretence or make believe from the components that are to be treated in some favoured out-fictional manner. With (9a), we are, let’s suppose, to pretend to assert that Bond is an abuser, but genuinely to assert that Bond is ever popular, but (9a) doesn’t allow us to do that, for the very thing that is an abuser is said to be ever popular, so to assert one is to assert (or presuppose) the other by an utterance of (9a). Likewise, one may go to pretend to assert that Bolkonsky is cruel, but one cannot do so while sincerely asserting that the same thing that is more popular than Vronsky is cruel. Just the same considerations apply to (9c–d) as the reader may check for themselves. Of course, some relations are apt for pretence and others not; it is not the relations that are problematic, though, but the very thing that is constant across the sincere assertion and would-be pretence, viz., the fictitious character. The thing we are only pretending to have some property is the very thing that has the property we sincerely assert of it,

A different tack might be ventured. It will be recalled that on the artefactual view a distinction can be made between a fictional artefact and how it is represented. Thus, we might think of Tolstoy as representing Bolkonsky as a somewhat dark figure, while a BBC drama represents him as a more dashing figure. Thus, when it comes to copredication cases, we might seek to specify the truth conditions via a differentiation of the in-predications as represented as and the out-predications as simply holding of the artefact. Let ‘RA’ go proxy for some elaboration of a modification to the effect that ‘RA(F)’ means ‘(be) represented as F’.

(10)	a	The ever popular James Bond continues to RA(abuse women).
	b	Prince Bolkonsky is more popular than Vronsky, but he is still RA(cruel).
	c	Humbert Humbert RA(appears sophisticated to Charlotte Haze), but a RA(cynical predator) to us.
	d	Raskolnikov RA(killed an old woman for no reason) and set Dostoevesky up for life.

Suitable paraphrases might thus be:

(11)	a	The ever popular James Bond continues, as represented in the stories and movies, to abuse women.
	b	Prince Bolkonsky is more popular than Vronsky, but, as represented by Tolstoy, Prince Bolkonsky is still cruel.
	c	Humbert Humbert is represented by Nabokov as appearing sophisticated to Charlotte Haze, but is represented as a cynical predator to us.
	d	Raskolnikov killed an old woman for no reason, as represented by Dostoyevsky, and set Dostoyevsky up for life.

Perhaps better paraphrases than these can be cooked up. The problem with this stratagem, however, lies not so much in the nuance of its execution but in the very idea of distinguishing the in and out predicates via construing the former as relative to a means of representation while the latter are not. We can see this clearly in the problems that arise for (10) + (11).

The problem with (10a) + (11a) is that in (9a) it is the ever popular Bond that is said to abuse women, but Fleming or the stories don’t represent anyone popular as an abuser. In grammatical terms, the subject of being an abuser is the whole DP—the ever popular Bond—not just Bond, yet the artefact is popular. Likewise, the problem for (10b) + (11b) is that (9b) is a comparison in popularity between Bolkonsky and Vronsky, which Tolstoy doesn’t make in any novel, of course. Bolkonsky, then, who is in-fictionally cruel, must be the very thing that is more popular than Vronsky, but nothing here is both popular and cruel. (10c) + (11c) suffers from a different problem. The rendering has it that Nabokov represented Humbert Humbert in two different ways, as sophisticated to Charlotte Haze and as a predator to us, yet Nabokov didn’t represent Humbert Humbert as anything to us, the readers, in any way at all, for we are not part of the novel, yet the artefact is not a predator either. Finally, (10d) + (11d) renders (9d) as if the Raskolnikov artefact is represented as killing an old woman and is responsible for Dostoyevsky’s success, but this misses the claim of (9d) that it was Raskolnikov’s crime that led to Dostoyevsky’s success. It was not the representation, but the artefact itself, that is responsible for the success.

Recanati (2018; cp., Voltolini 2006) presents not so much a dual view as a triple view, with parafiction, as he calls it, constituting the third kind of fiction (According to such and such, P). Recanati rightly sees copredication phenomena as posing a problem for any easy assimilation of parafiction into one or the other fictions, insofar as fictional characters appear to be Janus-faced, as it were, both like us (in-fictional) but also created and whimsical (out-fictional) (cp., Everett 2013, p. 165). That is exactly right, but such an insight should lead to the more thoroughgoing conclusion that copredication poses a problem for both in-fiction and out-fiction accounts individually and jointly as a dual strategy.

It should be unsurprising, I think, that dual accounts face such issues with copredications. The very phenomenon of copredication in general is that one and the same nominal is an argument (in our case a subject, but the reasoning obviously transfers to objects) of predicates that appear to entail joint but categorically distinct construals of the one argument. A dual view is just a denial of the very phenomenon. I shall consider a uniform view next, a view that has some promise with copredication.

2.6 Fictions Refer to Abstract Objects of Some Kind

There are various ways of implementing such a view, often misleadingly labelled Meinongianism.⁸ In general, the view is rightly motivated to secure the truth of some fictions (not all fictions are false or pretend) and to accommodate both in and out species of fiction. Thus, while apparently metaphysically outrageous, it promises to offer a very attractive semantic proposal that is genuinely sensitive to the linguistic phenomena.

Since Bolkonsky et al. are not denizens of the world in the same way Trump is, it appears that fictions cannot be true, because Bolkonsky et al. don’t exist (assume artefacts are foregone). One answer is to posit more to reality than satisfies an existence predicate. Thus, there is Bolkonsky after all, along with Natasha et al., but that they are does not render them as more entities to be counted as extra citizens of Moscow circa 1812. Ficta are non-existent on some such view, but are entities we can nonetheless speak about and, in particular, say true things about. Minimally, then, it appears as if such a view of ficta allows for the truth of both in-fiction and out-fiction. It is difficult, however, to see copredication falling into line on such a view. The core problem copredication poses is that while we might view fictional statements as being about non-existents, it is difficult to see how any single entity could have both in-fiction and out-fiction properties as attributed copredicatively.

Within Meinogianism, as it were, the distinction between the two types of attribution has been much discussed and various proposals have been formulated, each having the aim of showing how the one entity can be subject to different sorts of predication or have distinct kinds of property.⁹ Space precludes an investigation of the numerous ways of spelling out the basic idea, and the problems each way might face. There is, however, a major obstacle to the very idea of distinguishing between properties or predication in terms of the status of the arguments of the predicate, which is just is the coherence of copredication. That is, the Meinongian is well placed to deal with in-fiction and out-fiction, without offering two distinct proposals of the semantics of fiction, but it doesn’t follow that she is well-placed to deal with copredication where the two discourses interact. We might, agreeing with the Meinongian, be tempted to think that the cases in (12) have a different status:

(12)	a	Bond is popular
	b	Bond is a spy

One might express this thought by the notion that (12a) is contingent, involving an ‘extra-nuclear’ property attribution in Parsons’s (1974, 1980) terms (it is hardly necessary that public tastes are as they are), while (12b) is necessary, involving a ‘nuclear’ property attribution, for James Bond cannot really be a taxi driver, say, who only daydreams of being a spy. That seems correct, and different semantic and/or metaphysical accounts for in-fiction and out-fiction might thus appear compelling. Copredication, however, shows that the two kinds of predication can simultaneously apply and be variously realised by the available syntactic means. For example:

(13)	a	The ever popular Bond is a spy
	b	Bond, who is popular, is a spy
	c	The spy Bond is popular

One may read the Meionongian as, in a sense, embracing copredication, for they precisely wish to draw the distinctions that respect the differences between, say, the cases in (12) in a way the other accounts fail to do. It seems, however, that the distinction is drawn in the wrong place.

Suppose the truth-value intuitions about (12) are correct. It follows that there is some difference in property or predication only if the same semantic value for Bond holds across the cases. The copredication cases lead us to think something different. We do not conclude from the truth of any of the cases in (13) that there is something that is both popular and a spy. To think so is effectively to fall victim to a pun, as if one thought that a meal that is delicious and a meal that lasts all afternoon is the very same thing. That is to say, the lesson of copredication is not that there are different modes of predication, but that there is not the one external semantic value or referent that may support the various predicates (see Collins 2017). Take Bolkonsky. The thing that is popular with readers and has been played by both Russian and American actors, say, just isn’t the very same thing, non-existent or not, that danced with Natasha and met Napoleon on the field of Austerlitz. Natasha didn’t dance with a thing that an actor could play, no matter how weird or abstract of gerrymandered one makes it. There is a sense in which we can coherently consider Bolkonsky, qua in-fiction, as an abstract object of a kind, as if we are reflecting on a possible non-extant being, or an object to be thought about, but unrealisable. Likewise, there is a sense in which a fictional character is abstract, for characters don’t so much exist in time and space and bump into things, but they still clearly possess properties. All of that is so. What is not so is that the same non-existent abstract object is understood to fulfil both roles, but that is precisely what a uniform Meinongian view demands. In short, it is a pun to think that the very same thing is a spy and popular, or fought at Austerlitz and was created by Tolstoy.

A Meinongian might try a different tack. Perhaps there are properties abstract ficta possess, such as being popular and being created, whereas they can also possess fictive properties, such as being fictionally a murderer. A reviewer points out that this kind of duality holds in the non-fiction realm. Tony, an actual person, can be both really not guilty and allegedly guilty. So, being a fictional murderer is a different property to being a murderer. On this sort of account, there are not two kinds of predication; there are two kinds of property, which both actual entities and putative ficta can possess.

Although correct in its separation of properties, the proposal will fall foul of the kind of considerations raised in Sect. 2.5 against the introduction of a ‘representation’ modifier. Take (13a), for instance. Suppose we say that the abstract Meinongian entity Bond is ever popular, but possesses the property of being fictionally a spy, much as I might possess the property of being alleged to be a criminal or suspected of being a plagiarist. The same problem now recurs. The entity that is popular is not the entity that is fictionally a spy, if that means it is represented as being a spy by Fleming or in the movies. The stories, trivially, are not about an entity that is popular and do not involve the representation of a popular entity as a spy. The situation is different for actual persons. I can be both not guilty and allegedly guilty, because I am independent of representations of me—I’m cussed, as it were; how I am and how I am represented to be can readily come apart. For fiction, not so. A putative fictional entity can be viewed or represented in an in and out way, even simultaneously, but the entity is bound by the predications and has no distinct conditions of individuation, unlike me. That is the real problem of copredication: how can we think of a single thing simultaneously in ways that conflict? The answer, I think, is to drop the idea that there is a single thing at all, as the next section will explain.

As indicated in Sect. 1, I have not the space to explore the potential rejoinders a defenders of of the traditional accounts might offer. At any rate, a general complaint might be that I have merely adduced some problems without offering promise of an alternative. In the next section, I shall make good on my promise to offer maximally simple account of fiction. The ambition here will be to show how we might accommodate in-fiction and out-fiction, and their copredication.

3 A maximally Simple Account of Fiction

3.1 Initial Motivation

The accounts of fiction discussed in the previous section all find a proverbial bump in the carpet after flattening out an initial bump. This is due to their seeking to treat fiction in a bespoke way, either by conjuring novel referents, stipulating uniform falsity, or introducing semantic machinery or attitudes, which may be adequate for one kind of fiction, but aren’t for another. Suppose, however, that our maximally simple hypothesis holds: fiction demands no special treatment at all; it can be true or false just like any other kind of statement with no additional semantic or ontological resources required (see Sect. 1). There remains an ontological difference between Donald Trump and Prince Bolkonsky, say. The difference lies in Bolkonsky being a made-up character, a denizen of a story, while Trump possesses properties that outstrip any stipulations or other cognitive activity of those who behold him. We happen to know this, but not qua knowers of English, and so it need be no part of a semantics of fiction to record the difference.

Imagine the following dialogue circa 1974:

(14)	A	Deep Throat is now famous, but no-one knows who he is.
	B	Well, Deep Throat doesn’t exist; he’s a fiction—they are making it up.

In the 1970s it was unknown whether Deep Throat was real or a fiction concocted by Woodward and Bernstein (the fact that Deep Throat wasn’t presented as a fiction makes no difference to the present point). He turned out to be a certain Mark Felt. Thus, A, in our dialogue, spoke truly, B didn’t; but there is no semantic incompetence or even semantic disagreement here; both parties are perfectly competent with the language. Indeed, Mark Felt could have gone to his grave without anyone save for Woodward and Bernstein knowing that Deep Throat did refer to him. The fiction/non-fiction distinction, we might say, is all on the side of the world, not on the side of language. The question remains how fiction, in both in and out species and involving copredication, can be variously true or false without recourse to extra ontology or any special linguistic devices. The principle move I here commend is to distinguish between what makes a sentence or claim true and its truth conditions (Collins 2017). This separation of truth conditions from truth making is inspired by Azzouni (2010, 2017; cp., Asay 2013); the general position, however, may be commended free of Azzouni’s broader metaphysical concerns, or any such concerns at all. A bare desire to account for the semantics of fiction in the most minimal way possible leads to the kind of separation I endorse.

The basic idea is that there are lots of different ways of saying or explaining what makes a claim true or what grounds the claim’s predication as holding, whereas the truth conditions of the claim are those conditions that are specified as an aspect of semantic knowledge, and are systematic and transparent to the competent speaker. Truth makers can vary greatly across tokens of a single type (see below), not being tracked by linguistic distinctions, and are mostly opaque, beyond the ken of normal competent speakers (Collins 2017). If we identify meaning with truth conditions, or at least think of meanings as being determinative of them, then the difference between truth conditions and truth making might be hard to fathom. The distinction I have in mind, however, does not seek to tear asunder the marriage of the world and truth conditions, but only to insist that the link between the two is complex and not transparently tracked by language. Still, there are clear cases. We suppose The sky is blue and The sun rises in the east to be true, and we know what it would be for these sentences to be true. Yet one need know very little, if anything, about what makes them true (think of the children). The truth conditions we know thanks to knowing the language offer no guide as to what makes the sentences true. The point here is not peculiar to physics. The same holds for all kinds of legal relations that underwrite being married, or a tenant, or an employer, etc. Similarly, consider reflections on what makes something an art work. We may readily and correctly classify certain objects as art, but are hard pressed to offer an explanation of what makes them so. This is the kind of cleavage I propose for fiction. We know what it would be for Bond to use a Walther PPK to be true; just for Bond, indeed, to use a Walther PPK. Yet what makes this true is a set of complex facts involving narration, novels, films, etc., of which the linguistically competent may be more or less unaware. Imagine Smith and Jones arguing over what sidearm Bond uses. Smith thinks it is a Walther PPK, whereas Jones thinks it is a Beretta M9. It transpires in the course of the conversation that Jones thinks that Bond is a real personage. Disabused of this error, the original disagreement remains just as it was, rather than descending into confusion or semantic disagreement. What has changed is that Jones now appreciates that what decides the dispute is not the identity of a weapon carried by a spy, but what is recorded in certain narratives.

Much could be said here, but I must rest content with spelling out how the problem of the diversity of fiction engenders such a separation and that its apparent costs are slighter than might be imagined when compared with its benefits in making sense of fiction.¹⁰

3.2 Truth Making for In-fiction

Let’s first consider in-fiction:

(15)	a	Prince Bolkonsky was married to Lisa
	b	Prince Bolokonsky was married to Natasha

Let’s assume that we want to regard (15a) as true and (15b) as false, while recognising that Prince Bolokonsky, Lisa, and Nastasha are not things that otherwise attract our ontological commitment in anyway whatsoever, either, in particular, as abstract objects or as artefactual referents. The difference between the two statements resides in how one would explain their truth value. With (15a), one may go to the text and see it written how Bolkonsky is unhappily married to Lisa, and so on. Equally, with (15b), one finds the proposed marriage of Bolkonsky and Natasha frustrated by the intrigues of Kuragin. The simpleminded but crucial move here is to distinguish between what would determine the truth values of the sentences from what a competent speaker knows about their truth conditions. As just indicated, it is Tolstoy’s text that determines the truth values of (15a, b), but knowing the meaning of (15a, b), being competent with the sentences, has nothing whatsoever to do with Tolstoy. If someone wholly ignorant of Russian literature overhears an utterance of (15a) and repeats it, they have said something true, albeit based upon ignorance about the world and what would make the sentence true.

My chief claim here, then, is that the truth conditions of the sentences of (15) are insensitive to the fact that the predications are in-fiction, and that the semantic values of the names are not bespoke ficta. Their fictional status, and the difference in truth value between the host sentences, is constituted by what makes them true or false (i.e., Tolstoy’s text), which is not in any way recorded in the knowledge a competent speaker has merely by dint of being competent in English.

I don’t want to insist that some such difference between being made true by texts and the non-textual world is always easily comprehended. In principle, however, the difference is clear: the truth of fiction is grounded in texts or stories whereas the truth of non-fiction isn’t grounded in texts or stories (cp., Donellan 1974). The semantics of the various sentences is indifferent, though, for it doesn’t record what makes sentences true or false. So, consider:

(16)	After reading his father’s autobiography, George began to model himself on his father

Suppose that this is a non-fictional claim. Its truth is still grounded in a text, however, even if the text is an autobiography. The difference between (16) and fiction is that the very grounding text—the autobiography—is spoken of in the sentence. In-fiction is grounded in texts that are not spoken of. Out-fiction can speak of texts, but out-fiction is perfectly literal; its characteristic feature is that the contribution to truth conditions a given expression makes, such as a fictive name, is grounded in a text, which is, again, unlike (16).

A complaint might be that the proposal puts too much distance between truth conditions and truth-making. Let’s suppose that in the non-fictional case knowing the meaning of a sentence is knowing its truth conditions, which fix what would make the sentence true. In the case of fiction, the present proposal appears to offer a radically different picture, whereby one may know the meaning of a sentence that does not determine a truth value (given the world being a certain way). In short, what is on the table appears to be a bespoke semantic story for fiction, and an implausible one at that.

The complaint is well-taken, but effectively describes rather than condemns the proposal. In the normal non-fictional case, as it were, one still needs to distinguish truth conditions, as what a competent speaker knows in understanding a sentence, from what makes a sentence true, which is not necessarily specifiable from the speaker’s linguistic knowledge. The connection between truth conditions and truth making here is that any circumstances that would make a sentence true be describable as realising the truth conditions of the sentence, but a competent speaker need have no access to or grasp of such circumstances beyond the truth conditions licensed by the utterance she makes. More or less all truth-apt discourse (self-reporting of an experience or stipulations might be exceptions) has its truth values determined in ways beyond our ken. We understand simple arithmetic, say, but Lord knows what makes it true (the metaphysics of mathematics is yet to be completed). We understand The table is brown, but need have no understanding of the relevant physics and psychology that makes it true. Of course, in this case, truth conditions and truth-making can coincide. One might think that what makes for the truth of the claim just is for the table to be brown (Punkt!). Yet one can’t insist that such an identification exhausts the matter, for we might still wonder what makes it that the table is brown, which is hardly answered by a primitivist realism about colours. At any rate, I can see no sense in the doctrine that appeals to truth-making must always be exhausted by what truth conditions record.

Just so for fiction. To know the meaning of the sentences in (15), say, is to know circumstances under which they would be true; yet as specified by their linguistically encoded truth conditions, such circumstances are very thin, so thin as not to distinguish between fact and fiction. Thus, Tolstoy’s text makes true or false the cases in (15), but the texts are not part of the linguistically specified conditions (recall the Bond disagreement from just above). In effect, the ‘intensional operator’ story is correct, but not as an account of the meaning or truth conditions of fictional sentences, but as a story of what makes in-fiction true. In this sense, one can know the truth conditions of a fictional sentence without knowing that the sentence is fictional, i.e., made true by a story.

A related worry is that if in-fictions are to be counted as true, with no pretence or intensional operator in view, then they must express propositions, but since I acknowledge that there are no ficta, as artefacts or otherwise, then the propositions must be gappy in some sense, lacking constituents, and so they can’t be true after all. The objection is damning if one accepts a basic realism about propositions as Russellian entities involving otherwise extant constituents. It is part and parcel of the view being commended, however, that such a view be rejected, at least as an account of linguistic meaning.

Instead of worrying about fiction-involving propositions on the assumption of a Russellean proposition picture, we might think that so long as we can specify truth conditions for sentences, as we can for fictions as much as non-fictions, then we may read propositions off such semantic properties that enter into the meanings of the relevant sentences. What makes the propositions true is not, perforce, ficta as constituents of propositions, but whatever factors are deemed relevant to the claims at hand, such as what Tolstoy or Fleming wrote, factors that do not enter into the propositions. Thus, someone who believes that Bond is a spy does not have a belief about Fleming—he might be wholly ignorant of the author—but what makes his belief true is precisely what Fleming wrote (inter alia).

3.3 Truth Making for Out-fiction

Consider (17):

(17)	a	Fleming created Bond
	b	Bond has made more money for the Broccolis than he ever did for Fleming
	c	Holmes is smarter than Dupin

What makes these sentences true or false is not the content of a single text, but texts and their outside, as it were, including the authors and other texts. So, (17a) is true, not because there is an abstract Bond object or artefact, but just because Fleming wrote some stories coining Bond, and so on. Here, assuming the relevant sense of the verb create, knowing the truth conditions is pretty close to knowing what makes the sentence true, at least to the extent that Bond is not understood to be a real person—if one were to think that Fleming was Bond’s mother, then one would have misconstrued the sentence. There is still a gap, of course, for (17a) can be made true in many different ways, with a competent speaker being ignorant of which particular way happens to make it true. A speaker who thought that Fleming directed the first Bond move would be perfectly right in uttering (17a), but for the wrong reasons.

Likewise, as a competent speaker, one may know the truth conditions for (17b) while not knowing what makes it true. In fact, what makes (17b) true is a set of complex legal arrangements of which a competent speaker need not have the least understanding. The truth conditions do not go beyond Bond being responsible for the relative incomes, where precisely who or what Bond is remains open.

(17c) patterns on much the same lines. It is up in the air whether (17c) is true. Suppose some speaker reckons it is true. She might do so competently and coherently by thinking that Holmes and Dupin were real detectives, or that only one was real, or that the detectives were created by one and the same author, or, as is the case, by two distinct authors. Again, then, no additional ontology or a departure from the fiction case is required. The gap between the truth conditions and what makes a sentence true is just the same as in the non-fiction case. The difference resides only in what makes something true between the cases, such as the content and wider social role of a text, but such matters go beyond semantic competence.

3.4 Truth Making for Copredication

Copredication cases fall into line as expected. Consider (18):

(18)	Bond kills lots of bad guys, making him one of the most popular cinema draws

(18) is true. What makes it true is a complex set of facts involving the content of various films and books and the number of people paying to see the relevant movies.

The general problem copredication poses is how to understand the relevant nominal (subject or object, say) as having two (or more) conflicting construals simultaneously. It appears as if no possible referent, real or imagined, could have the properties attributed, and we do not understand the sentences as if there were just one entity at issue, just as we don’t think that meals as foodstuffs and events, or countries as geographical regions and legal entities are one and the same things. With (18), for example, while we reckon it to be true, we hardly think that there is some one thing that is both a murderer and popular with the cinema-going public. We mean to be speaking of different things via the different predicates, and yet there is just the one argument Bond. Copredication remains an outstanding problem in the semantics literature (cp., Asher 2011; Collins 2017). Still, however copredication is best theorised, it arises for fiction in just the same terms as it arises in general: a single argument supports distinct predications that attribute properties that are not understood as being possessed by a univocal value of the argument. The extant accounts of fiction are especially beset by copredication worries precisely because they typically offer a bespoke treatment of fiction designed to explain one or other species of fiction, but not together. What copredication cases indicate is that both species of fiction must be accounted for equally. What I have proposed is a maximally simple account that does just this insofar as the differences between in-fiction and out-fiction are not encoded in the language, but are determined by how the world is such that the relevant sentences are true. Thus, there is no mystery of how (18), say, can be true, for the situation can be readily described via an account of the stories and their cinematic blockbuster status. There is nothing that corresponds to Bond univocally. What the speaker has to go on are the properties of the relevant predicates in order to discern the kind of facts that would render (18) and the like true. Yet again, though, knowing such matters goes way beyond bare semantic competence.

Still, it might be reasoned that an out-and-out literalism about fiction must be false, for fictions are not counted as really true, true unqualified, simpliciter. As Sainsbury (2010a, pp. 26–27; cp., Sainsbury 2005) puts the principal intuition: ‘[p]people don’t think that “Holmes lived on Baker Street” is really true… Even for those sentences [we might think of as true] like the Holmes sentence, we behave oddly, having much less inclination to accept the equivalent “Baker Street numbers Holmes among its inhabitants”’. Sainsbury, speaking on behalf of the folk, is right that people do not treat fiction and fact as one and the same. My simple proposal, however, does not erase the distinction between fact and fiction, but only denies that it is grounded in any linguistic fact that calls for a novel ontology, a uniform treatment of fiction, or otherwise unrequired semantico-syntactic machinery. Thus, a speaker might think that Napoleon didn’t really meet Bolkonsky, but did really meet Alexander I. It doesn’t follow, however, that the cases in (19) are semantically or even extensionally equivalent:

(19)	a	Napoleon met Bolkonsky
	b	Napoleon really met Bolkonsky

That is to say, a perfectly competent speaker might think (19a) true while denying (19b). For such a speaker, the use of really signals that the statement is out-fictional. Such a speaker might still reckon in-fictions to be true tout court; or, at any rate, there need be nothing in the literal content of (19a) that reflects the supposed metaphysical heft the adverb really introduces.

(19a), say, can be true or false, depending on whether one is talking in fiction or whether one is taking about events near Austerlitz in 1805. There is, however, no apparent ambiguity in (19a) or other distinguishing semantic properties; rather, different factors may make a sentence true or false. If one is talking of fiction, then one is taking (19a) to be made true by a text (or a third party might understand it in that way); if one is talking of the details of the Napoleonic wars, then (19a) is false. If that is so, the fiction/non-fiction distinction is perfectly robust; it is just not a linguistic distinction. One might think of it as a discursive or cognitive difference. In this light, a fictive linguistic type, such as (19a), is neither true nor false; utterance tokens acquire a truth value given a context in which what makes them true can be established, or at least agreed upon.

Likewise, Sainsbury’s out-fictional Baker Street numbers Holmes among its inhabitants appears false, because Holmes isn’t out there in West London, but in-fictionally it is perfectly innocent; after all, he doesn’t live on The Strand.¹¹ Whether we think of the fictive statement as true or not depends upon whether we construe it in-fictionally or out-fictionally. Yet since the difference is not fixed by linguistic competence alone, merely knowing the meaning of a fictive sentence type doesn’t fix how the world should be to render tokens true, for different tokens of the single type will be true or false for different reasons.

3.5 Negative Existentials

The thoughts just offered get to the heart of the problem negative existentials pose. As mentioned in Sect. 2.1, the truth of negative existentials offers a prima facie reason for a blanket falsehood view. Every theorist of fiction has a view on how to accommodate them, which I shall not dispute or adjudicate here. I only wish to register how a maximally simple view might approach them.

The puzzle of negative existentials is exemplified by the case of (20), where we want to claim that both sentences are true:

(20)	a	Bolkonsky met Napoleon
	b	Bolkonsky didn’t/doesn’t exist

On the face of it, though, the two issue in an inconsistency, if (20a) entails (21):

(21)	(∃x)[x = Bolkonsky ∧ Met(x, Napoleon)]

A ‘Meinongian’ classically resists this inference pattern (cp., Zalta 1983).¹² An alternative approach is to deny that both cases under (20) are straightforwardly true. (20b) is true, if one is seeking to enumerate the denizens of the concrete universe, and (20a) is likewise false, i.e., Bolkonsky is not among the objects to be counted any which way. If, on the other hand, one is talking of Tolstoy’s fiction, then (20b) is false (Bolkonsky is not a dream character, say), and (20a) is, of course, true. The basic point is that the evaluation of fictive negative existentials turns on whether we are talking of fiction or not. If we are, then they can perfectly well be false (or true, too). If we are not talking of fiction, then they will be uniformly true, much like if we were to append really to our predicates to indicate that we are not speaking fictively (cp., Azzouni 2017, p. 63). It remains to add, and this is just to spell out the moral from above in response to Sainsbury, that whether a speaker is using their language fictively or not is not recorded linguistically; it is not as if language provides a covert really we may silently attach to our overt predicates. In sum, just like any other statement, a negative existential will be evaluated for truth relative to a certain understanding, much as the very fiction/fact distinction operates.

3.6 Quantifying Over Ficta

Any view of fiction that seeks to eschew the reality of ficta faces the problem of how to account for quantification over fictional entities, which is perfectly acceptable in all forms:

(22)	a	Most fictional detectives are admired by Sally
	b	At least one fictional detective is smarter than any physicist
	c	Any character in Tolstoy is more interesting than any character in Dan Brown

If we assume that such sentences can be true and that quantification is ontologically committing, then there must be fictional detectives and characters; for a construction to be ontologically committing just means that its truth depends upon the existence of a domain of entities that satisfy the relevant predicates.

The presupposition of this objection is highly questionable, for it remains unclear if quantificational constructions are ontologically committing (Priest 2005; Crane 2013; Azzouni 2017; Collins 2017). Indeed, trivially, one can speak of most things or at least one thing or anything that doesn’t exist. Of course, if one thinks there are ficta, then one can take oneself quantificationally to speak about them, but the truth of (22a–c) and the like does not entail the existence of the denizens of the domain.

What makes the quantificational fictions true, given that there are no ficta, are what we may think of as truth-maker witnesses for the quantifications. So, assuming that we can count fictional detectives, if whatever makes it true that Sally admires any given fictional detective can be is replicated for more than half of the detectives, then (22a) will be true. Likewise, (22b) will be true if there is at least one truth-maker witness for ‘n is smarter than any physicist, where ‘n’ is proxy for any fictional detective name. Mutatis mutandis for (22c) and other kinds of quantification.

3.7 Intentional Identity

Geach (1967) famously posed the problem of the identity of intentional inexistents: how can people have the same thought about a particular witch, say, when there are no witches? The same issue arises with fiction. Suppose the cases in (23) are true:

(23)	a	Jack and Jill both despise Bond
	b	Sam admires Bond, but Jill despises him

If there is no Bond in any way at all, then how can Sam, Jack, and Jill have thoughts about the same thing, as is apparently demanded for (22a, b) to be true? It may seem that that some form of realism about ficta—artefactual or Meinongian—is required. At the very least, a desideratum appears to be imposed: account for sameness of construal across uses of fictional names and anaphora, with or without commitment to ficta. Many have usefully tarried over this question, and it should be stressed that it is by no means a consensus that Geach-like co-identifications entail a univocal referent, an inference that Geach (1976) himself resisted.¹³ Here I only wish to indicate that the problem is not pressing on the maximally simple view, for a very lesson of copredication is that no ontological identity is required, only non-ambiguity of the relevant lexical items.

Natural language is perfectly happy to have a failure of existential inference across anaphora:

(23)	a	Brazil was buoyed by the success of the Olympics, but it remains heavily polluted, and its election of Bolsonaro has been divisive
	b	Bill read every book in the library then burnt them

A number of different construals of Brazil are in play in (23a), which preclude a generalisation of the form, There is something such that it is buoyed, polluted, and voted in a populist. The different predicates individuate their subjects differently, as something abstract, such as an economy, a geographical area, and a population (Asher 2011; Pietroski 2003; Collins 2017). These are not the same entity, even though the predicates that support such diverse construals can be anaphorically linked. The same effect is in play with quantifiers. If Bill read every book in the library, then he needn’t have read the four copies of the Bible, but if he burns every book, then all four copies of the Bible must be burnt. Copredication involving anaphora just is a case where we resist identity, so the mere fact of anaphora deoesn’t entail identity of a referent in any independent metaphysical sense.

That said, identity is sometimes intended, such as in the classic Geach scenarios and the fiction cases depicted. The lesson of copredication transfers, however. The identity at issue need be no more than that the pronoun is read as dependent on the antecedent non-ambiguous nominal. So, (23a) can be unpacked as a conjunction with Brazil substituting for it across the clauses; ditto for (23b), mutatis mutandis. When it comes to fiction, therefore, the same moral should apply. If (22a) is true, then Jack and Jill despise the same thing, but that means no more than that Jack despises Bond and Jill despises Bond are both true, where Bond is non-ambiguous, just as Brazil is non-ambiguous. It remains possible that Jack and Jill have radically different conceptions of Bond, including Jack’s belief that he is real, say (see Sect. 3.1). If (22b) is true, then the same thing must be admired by Sam, and despised by Jill, but, again, that just means that Sam admires Bond and Jill despises Bond are both true, with Bond non-ambiguous.

In short, from a linguistic as opposed to metaphysical perspective, there is no special problem of intentional identity over fiction, for the semantics of dual predication and anaphora do not entail that there is a univocal entity; if they did, copredication would be unacceptable, which it is not. All that is demanded by the relevant phenomena is that the nominals are not ambiguous, and so any anaphora upon them are jointly construed. What makes the sentences true is another matter, for which the semantics does not dictate an answer.

Let me end on a general point that deserves emphasis. If my considerations so far are sound, then there is a blanket objection to accounts that posit ficta, whether artefactual or Meinongian, viz., such putative entities are supererogatory, fifth wheels that are unrequired to for truth conditions of fiction, the truth making of fictional statements, or the explanation of intentional co-identification. If we pair this redundancy claim with the accommodation problems that beset the alternatives (falsity, intensional operator, and pretence approaches), then we have a general argument for the maximally simple approach: extra ontology is superfluous, and there is no single semantic or cognitive attitude that distinguishes the use of fictional from non-fictional terms.

4 Conclusion

If the above is on the right lines, the maximally simple view should be the default semantic position, to be adopted in the absence of compelling reasons against it; after all, it is the null hypothesis. To the contrary, such a simple view is rarely even acknowledged as a position, let alone endorsed. That it suffers such eschewal can be explained by an unholy alliance of metaphysics and semantics, the notion that the metaphysical difference between fiction and non-fiction needs to be reflected in semantics. All alternative accounts variously subscribe to such an alliance by way of seeking a fit between semantics and metaphysics, no matter the apparent cost. There is a difference, of course, between ficta and non-ficta, for one can’t, say, bump into Holmes on Baker St. Cognisance of such differences, however, is no part of knowing a language, and so should not be enshrined in the semantics of fiction.¹⁴