The problem addressed by a theory of fictional truth[1] is that of providing a general account of the conditions under which a sentence of the form,

In fiction F, S

is true. A natural first suggestion is to identify fictional truth with what is explicitly stated in the text. There are, however, a number of familiar objections to this suggestion. First and foremost, the sentences which constitute fictional texts are not normally asserted at all. Felicitous assertion presupposes belief on the part of the speaker or writer in the truth of what is said. But authors of fiction do not (normally) believe what they say, and their speech acts are not for that reason considered defective. One might, of course, rejoin by identifying fictional truth with what is asserted by the fictional narrator rather than by the author of the fictional work. The trouble, however, is that what fictional narrators assert is neither necessary nor sufficient for fictional truth. It is not necessary because there are many fictional truths that are not stated by the narrators of the works in which they occur. Watson, for example, nowhere says that Holmes wears underpants, yet it is true in the Holmes stories that he is so clad. And it is not sufficient because fictional narrators are often unreliable: they assert or imply things that are not true – even in the story. 


Despite these difficulties, many philosophers have found the basic idea underlying this suggestion to be promising.[2] This idea is that fictional truth is correctly analyzed in terms of the speech acts or propositional attitudes of tellers of fiction works.  As we have seen, taking the relevant act/ attitude to be assertion leads to insuperable difficulties, but it remains open to analyze fictional truth in terms of other acts/ attitudes such as belief[3] or some kind of sui generis fictive illocution.[4] The views considered below differ in this respect, but also in respect of who the relevant teller is taken to be – the author, the narrator, or some third figure. As a result, it may prove fruitful to identify the following shared form of the various analyses under consideration:

‘In fiction F, S’ is true iff a V’s <p>,

where a is a teller, V is an act/ attitude, and <p> is a proposition expressed by (or otherwise suitably related to) to ‘S’.[5]


It is worth pausing for a moment to say few words about the various sorts of tellers at issue. Tellers can be usefully categorized in terms of a pair of distinctions, one tripartite and one bipartite. First, we can distinguish ontologically between those tellers that are actual and those that are non-actual, which include both fictional and non-fictional (but still non-actual) tellers. And second, we can distinguish linguistically between those tellers who engage in factual discourse, or fact-telling, and those who engage in fictional discourse, or fiction-telling. These distinctions yield six categories of tellers: (i) actual fact-tellers, such as authors of non-fiction; (ii) actual fiction-tellers, such as authors of fiction; (iii) fictional fact-tellers, such as explicit narrators of fictional works; and (iv) non-actual/ non-fictional fiction-tellers, such as the hypothetical authors invoked by Levinson[6] among others; (v) fictional fiction-tellers, such as story-telling narrators – explicit narrators who clearly indicate that the story being presented is fictional; and (vi) non-actual/ non-fictional fact-tellers, who will be discussed in section V below.


This paper consists of five sections. In the first three sections, the views of Lewis,[7] Currie,[8] and Byrne[9] are considered in turn and ultimately found to be lacking. In the fourth section, an alternative analysis in terms of the revelations of a figure I term the narrative informant is developed and shown to avoid the difficulties which beset its competitors. And in the final section, a number of potential difficulties for this analysis are addressed. In the first four sections of this paper, I will make the following simplifying assumptions. First, I will restrict my attention to works of oral fiction told or performed by their authors. And second, I will assume that the works in question have explicit narrators. Together these assumptions facilitate a model of the author of a fictional work as an actor onstage portraying a character – the explicit narrator of the work. In the final section, complications which arise when these assumptions are abandoned will be addressed.   


I: Felicitous Fictional Fact-tellers


Lewis has defended a theory which retains the idea that fictional truth is to be analyzed in terms of the assertions of a teller, where the teller is a felicitous fictional fact-teller – specifically the narrator – who tells the fictional story as known fact.[10] What is unique about Lewis’ view is that it embeds this core idea within a counterfactual analysis. Lewis in fact offers two distinct counterfactual analyses of fictional truth:

LEWIS-1: ‘In fiction F, S’ is true iff there exists a world w in which F is told as known fact and ‘S’ is true, which is closer to the actual world than any world in which F is told as known fact and ‘S’ is not true[11]

LEWIS-2: ‘In fiction F, S’ is true iff for all worlds w, if w a collective belief world of the author’s community, then there is a world w’ in which F is told as known fact and ‘S’ is true, which is closer to w than any world in which F is told as known fact and ‘S’ is not true,[12]

where the collective belief worlds of a community are those worlds in which the (overt) beliefs shared by most members of the community are true.[13] So, for example,

Holmes is the offspring of human parents

comes out true on Lewis’ analyses because there are worlds where the Holmes stories are told as known fact and Holmes is the offspring of human parents which are closer to actuality – and to the collective belief worlds of Doyle’s community – than any world in which the stories are told as known fact and Holmes had less typical origins.


Despite its virtues, Lewis’ analysis falls prey to insuperable difficulties. The first thing to note is that it is not commensurable with the possibility of an unreliable narrator. At any world in which a work is told as known fact, all the claims contained in the work will automatically be true. Hence, it trivially follows that, for any claim contained in the text, there is a world where the text is told as known fact and the claim is true which is closer than any told-as-known-fact world in which it is false. Lewis, as usual, is aware of this difficulty and suggests that works with unreliable narrators require a separate treatment:

“In these exceptional cases also, the thing to do is to consider those worlds where the act of storytelling really is whatever it purports to be – ravings, reliable transmission of a reliable source, or whatever – here at our world.”[14]

Now it may well be possible to develop this idea to cover such these sorts of cases.[15] But narrative unreliability is a more common phenomenon than Lewis acknowledges, and cannot be set aside as a mere marginal case. And, other things being equal, a univocal theory is preferable if one can be found.


One of the virtues of Lewis’ theory is that it can accommodate fictional truths that are not explicitly stated in the text. For example, despite not being explicitly stated in the Holmes stories,

Holmes wears underpants[16]

comes out true on Lewis’s analysis because worlds where the Holmes stories are told as known fact and Holmes wears underpants are closer to the actual world – and the collective belief worlds of Doyle’s community – than any world in which the Holmes stories are told as known fact and Holmes does not wear underpants. The difficulty that arises for Lewis’s view account of unstated fictional truth is that it is too permissive: it entails of any given fictional work that there are a multitude of fictional truths which intuitively are not true in that work. The trouble is that both the actual world and the collective belief worlds of an author’s community contain a lot of facts – about the distant past, about spatially remote regions of the universe, about the micro-physical constitution of things, etc. – which are almost always irrelevant to the events described in a fictional work. And a world where a text is told as known fact and this extraneous fact obtains will be closer to the actual world, or the author’s community’s collective belief worlds, than a world where it is told as known fact and the extraneous fact does not obtain (as long as it is not explicitly ruled out by the text). As a result, all this extraneous fact will be true in fiction if Lewis’s analysis is correct. But it is intuitively implausible to suppose that, for example, the fact that Julius Caesar had a certain number of hairs on his head when he died is true in the Holmes stories.[17]


Moreover, even if these extraneous fictional truths are tolerable in realistic fiction, they are not tolerable in non-realistic fiction, such as fantasy and (some) science fiction. Consider, for example, a story in which the use of magic plays a prominent role or in which the world rests on the back of a giant turtle. Even if it is not explicitly stated in the text, it may well be fictionally true that the world is non-quantum or that the events described do not occur in our universe. However, the closest told-as-known-fact world to the actual world or the author’s community’s collective belief worlds might nevertheless prove to be quantum worlds or worlds consisting of our universe. Lewis suggests that such worries can be alleviated by appeal to “inter-fictional carry-over” from other stories in the same genre:

“If Scrulch [a dragon] does breathe fire in my story, it is by inter-fictional carry-over from what is true of dragons in other stories.”[18]

But although the notion of inter-fictional carry-over does have intuitive appeal, it needs to be much more fully developed to be of use here. Minimally, an account of exactly what information gets carried over from (and to) which fictional works is required, as well as an account of how this information is to be balanced with information derived from the actual world or the collective belief worlds of the author’s community. 


Finally, Lewis’ analysis runs afoul of the problem of inconsistent or impossible fiction. If a work of fiction contains an inconsistency, then there is no possible world in which it is told as known fact. As a result, Lewis originally acknowledged that, on his analysis, “…anything whatever is vacuously true in an impossible fiction.”[19] Lewis subsequently attempted to solve problem by dividing inconsistent fictions into consistent fragments and characterizing truth in inconsistent fiction as truth in at least one of the fragments.[20] Currie has argued, however, that this strategy cannot accommodate a fictional work in which, for example, the protagonist has refuted Gödel’s theorem: since there will be no consistent fragment in which the protagonist has done so it will not prove to be true in the work that she has refuted Gödel.[21]


II: Felicitous, but Shadowy, Fictional Fact-tellers


Like Lewis, Currie analyzes fictional truth in terms of the acts/ attitudes of a felicitous fictional fact-teller. But Currie’s view departs from Lewis’ in two central respects. First, the act/ attitude he invokes is belief rather than assertion. And second, the fictional fact-teller he invokes is not the explicit narrator of a fictional work, but rather a figure he calls the “fictional author.” The fictional author is a hypothetical member of the actual author’s community who, unlike the latter, tells the story as known fact. Moreover, although distinct from the explicit narrator, the fictional author tells the story by speaking in the voice of – or, perhaps, playing the part of – the explicit narrator. And when the narrator is unreliable, the fictional author does so in a manner which indicates that what she is asserting is not to be relied upon.[22] [23]


Currie analyzes fictional truth in terms of the beliefs of this figure as follows:

CURRIE-1: “In fiction F, S” is true iff it is reasonable for the informed reader to infer that the fictional author of F believes that S.[24]

Given that “the teller – the fictional author – is a fictional construction, [who] has no private beliefs, no beliefs that could not reasonably be inferred from text plus background,”[25] this essentially equivalent to the simpler

CURRIE-2: “In fiction F, S” is true iff the fictional author of F believes that S.

So, for example, since the fictional author of the Holmes stories is a member of Doyle’s community and, as such, would presumably believe that respectable detectives wear underpants,

Holmes wears underpants

is true, on Currie’s analysis, despite its non-occurrence in the text.


In addition to accommodating fictional truths not explicitly stated in the text, Currie’s analysis avoids some of the difficulties that beset Lewis’ theory. By distinguishing between explicit narrators and fictional authors, and identifying fictional truth with the beliefs of the latter, Currie’s analysis can easily handle the difficulties posed by unreliable narrators. Although the assertions and beliefs of an explicit narrator may be a poor guide to fictional truth, the beliefs of a fictional author – “an unobtrusive narrator who, by putting words in the mouth of the explicit narrator in a certain way, signals his scepticism about what the explicit narrator says”[26] – are not. And Currie’s analysis can also accommodate inconsistent fictions. In order for a contradiction to be true in a fictional work, all that is required is that the fictional author believe it.[27] And people believe impossibilities – that Gödel’s theorem is false, e.g., – all the time.


Nevertheless, Currie’s analysis does run into serious difficulties. Although it allows inexplicit truth in fiction, it is both too inclusive and too exclusive. First, because the fictional author is a member of the actual author’s community she will presumably share beliefs prevalent among members of this community. And many of these beliefs will be simply irrelevant to the fictional story. But, as with Lewis’ second analysis, Currie’s view implies that such beliefs are true in fiction. And second, as Byrne points out,[28] typical members of, for example, Doyle’s community would have had significant gaps in their knowledge, as well as mistaken beliefs, about the relative locations of landmarks in turn of the century London. As a result, given that the fictional author of the Holmes stories is herself a (typical) member of Doyle’s community, Currie’s analysis entails that the London of the Holmes stories differs geographically from the real London in significant (but unspecified) ways.


There is also reason to balk at Currie’s refusal to identify fictional authors with (unreliable) explicit narrators. First, it presupposes an implausible account of audience (i.e., reader/ listener) engagement with fiction. Recall our model of the author of a fictional work as an actor onstage portraying the explicit narrator of the work. According to Currie, if the portrayed narrator/ character is unreliable,

“[by] his manner of uttering, the speaker indicates to us that the words he utters are really those of someone whose picture of himself and his relations to the world is distorted. In that case the audience would see the speaker at one remove from the character … even though [the character’s] words issue from the speaker’s mouth in first person form.”[29]

This would be all well and good if the speaker in question were the author/ actor.[30] But Currie denies this; there is, instead, a fictional speaker distinct from both the narrator/ character and the author/ actor, who speaks in the voice of the unreliable character and who is recognized by the discerning audience member. This, however, is exactly what discerning audiences do not recognize. They imagine there is a mismatch between the world of the play and the character’s beliefs or statements about it. And anything in the manner of speaking that they take to indicate unreliability they attribute to the actor and not some “shadowy meta-character.” As Matravers puts it,

“[if] I realize, from some story that is being related to me as known fact, that the narrator has drawn the wrong conclusion and has a number of false beliefs, it would be wrong to postulate some shadowy meta-narrator who has all the true beliefs but is choosing not to reveal himself.”[31]


Moreover, distinguishing the fictional author of a work from the explicit narrator is fraught with ontological pitfalls. These difficulties arise because the fictional author is an inhabitant of the world in which the events described in a fictional story occur. And not only is this incompatible with the phenomenon of mindless fiction, in which there is no intelligent life and, hence, no one tells the story,[32] [33] it also leads to a troubled relation between this figure and the fictional narrator, if there is one. The problem is that not only does the fictional author inhabit the same world as the narrator, she speaks in the voice of the narrator. One might be tempted to think of the fictional author as some kind of disembodied spirit with the power to possess the narrator and control, if not what he says, at least his manner of saying it. But even this rather tenuous suggestion is ruled out by the fact that the fictional author is a supposed to be a hypothetical member of the actual author’s community and, hence, an embodied human organism. The upshot is that the fictional author of the Holmes stories, for example, is an embodied member of Doyle’s community who occupies the same world as Watson and somehow has the power to speak in his voice and influence his manner of speaking – perhaps this was the secret of Moriarty’s success. And even if this is coherent, or at least acceptable in certain works of non-realistic fiction, it is hardly tolerable across the board. It is worth emphasizing, however, that the relationship between the actual author and the narrator of a work – here modelled on the actor/ character relation – is immune to such difficulties.


III:  Non-Actual/ Non-Fictional Fiction-tellers


Prompted in part by the phenomenon of mindless fiction Byrne rejects “idealist” analyses of fiction – analyses which entail that it is fictionally true that the tale is told as known fact. Instead he proposes an analysis wherein the teller is a fiction-teller rather than a fictional fact-teller. But rather than invoking an actual fiction-teller – in particular, the author – Byrne invokes a non-actual/ non-fictional fiction-teller or ideal author. Byrne’s motivation here is the thought that what a text means – roughly sentence meaning – and what an author mean by a text – speaker meaning – can come apart, and fictional truth is a function of the former.[34] The ideal author, in contrast, is a hypothetical construct whose intentions regarding the meaning of the text are by definition satisfied.


The act/ attitude in terms of which Byrne analyzes fictional truth is that of inviting readers to make-believe. This yields the following:

BYRNE-1: “In fiction F, S” is true iff the ideal reader could infer that the ideal author of F is inviting the ideal reader to make-believe that S.[35]

But given that “…the [ideal author] intends to say precisely what the ideal reader … thinks the actual author intended to say…”[36] this is effectively equivalent to the simpler,

BYRNE-2: “In fiction F, S” is true iff the ideal author of F invites the ideal reader to make-believe that S.

So, for example, the ideal author of the Holmes stories invites readers to make-believe that Holmes is a respectable nineteenth century London detective. As a result, the social conventions of respectable London society of the time, such as those governing the use of undergarments, are salient; we would expect to be told if the London of the Holmes stories differed significantly in its social conventions from nineteenth century London itself. Since the ideal author has not told us this, she has strongly contextually implied that we are to imagine actual conventions are in effect. Hence,

Holmes wears underpants

is true, on Byrne’s analysis, despite its non-occurrence in the text.


In addition to accommodating fictional truths not explicitly stated in the text, Byrne’s analysis does avoid many of the difficulties that arose for the alternatives we have considered. First, it is specifically designed to avoid the problem of mindless fictions. Since the ideal author is not a fictional fact-teller, Byrne’s account does not entail that in every fictional work it is fictionally true that the story is told as known fact and, hence, that mindless fictions are inconsistent. Second, since nothing prevents an author from inviting a reader to make-believe a contradiction, Byrne’s analysis has no difficulty handling inconsistent fictions. And third, since the ideal author of a fictional work is distinct from the explicit narrator, narrative unreliability poses no difficulty for Byrne’s view, at least as long as there is no reason to identify the reliable invitations of the former with the unreliable assertions of the latter.


There are, as might be expected, a number of reasons to resist Byrne’s analysis. First, like Currie’s fictional author, Byrne’s ideal author is a shadowy figure unfamiliar to most appreciators of fiction. Consider again an actor on the stage portraying an unreliable character. A discerning audience member would, of course, refrain from taking the assertions of this character to be fictionally true. And she might, following Byrne, resist taking the intentions of the actor (or even the director) to be definitive of fictional truth.  After all, in ordinary conversation, when meaning and intention come apart, one’s primary focus normally involves discerning the intentions of one’s interlocutor; but when the meaning of a fictional text deviates from the intentions of its author, the former is of independent (and arguably primary) interest. But she is no more likely to identify fictional truth with the intentions of a shadowy meta-actor than she is to identify it with the beliefs of a shadowy meta-character.


More serious difficulties arise when we consider the relation that must hold between a set of intentional contents – putatively of an ideal author’s invitations to make-believe – and the sentences or utterances which constitute a fictional text in order for the former to serve as fictional truths embodied (for lack of a better term) by the latter.[37] Byrne models this relation on the relation that holds between a non-fictional text and the propositions asserted in it (or, perhaps, by means of it). Propositions asserted explicitly by a text are recoverable from the information linguistically encoded in the text (and features of the context, in the case of ambiguity). And propositions asserted implicitly are recoverable by pragmatic inference involving explicit assertions, tacit conversational maxims, and the like.[38] Where propositions are asserted by (means of) non-fictional texts, invitations to make-believe propositions are made by (means of) fictional texts. And as with assertion, invitations to make believe can be either explicit or implicit:

“Now what the Author [=ideal author] invites the Reader [=informed reader] to make-believe may not be explicitly stated in the text. But just as implicit assertions in non-fiction can be recovered by pragmatic inference, so can implicit invitations to make-believe in fiction.”[39]

Byrne’s view, then, is that set of fictional truths embodied by a fictional text consists of those propositions (informed) readers are explicitly invited by the text to make-believe and those propositions readers implicitly invited by the text to make-believe.


Byrne’s suggestion that the utterances which constitute fictional texts are explicit invitations to readers to make-believe whose contents are fictional truths is problematic on two grounds. First, it presupposes that fiction-telling involves the performance of uniquely fictive illocutionary actions, distinguished from other sorts of illocutionary acts in terms of the intended effects upon listeners.[40] But this idea is highly controversial. Searle has argued, for example, that since (i) “[in] general, the illocutionary act (or acts) performed in the utterance of the sentence is a function of the meaning of the sentence…”[41] and (ii) the very same sentences are used both in fictive and non-fictional discourse, if there were a uniquely fictive speech act, the sentences of our language would have to be ambiguous between their use in fiction and their use in non-fictional discourse.[42] And second, this suggestion runs afoul of unreliable narratives. Even if an ideal author who voices a suspect assertion by an unreliable narrator can be taken to invite readers to make-believe the asserted proposition, this proposition cannot be taken to be a fictional truth; rather it should be taken to be a fictional falsehood.[43]


Now Byrne might respond to this second criticism by suggesting that the contents of explicit invitations to make-believe are only provisionally included among the fictional truths. When an explicit invitation violates some conversation maxim or other, readers can infer that the ideal author is not really inviting them to make-believe the proposition in question. This would be all well and good if communicative maxims governed invitations to make-believe. But the familiar conversational maxims do not do so. Grice’s maxim of quality, for example,

G-QUALITY: do not say what you believe to be false.[44]

is simply orthogonal to any invitations to make-believe an author might make. An author who invites a reader to make-believe what she does not in fact believe does not violate this maxim; and, hence, no pragmatic inferences can be made by appeal to it. Byrne might, of course, invoke communicative maxims specifically governing fictional discourse such as,

F-QUALITY: do not invite readers to make-believe what you do not intend to be fictionally true,

or, perhaps, very general maxims from which both F-QUALITY and G-QUALITY follow. But in lieu of an independent argument for the existence of such maxims, this strategy has more than a hint of ad hoc-ness to it.


IV: Infelicitous Non-Actual Fact-tellers


There does, however, remain open to Byrne an alternative abductive strategy: the invitations to make-believe whose contents are the fictional truths embodied by a fictional text are those which best explain the text.[45] Now I have no objection to this strategy; in fact, in my view it is on the right track. But what I want to insist on is that any adequate explanation of the text as an act of fiction-telling will have to be mediated by an explanation of it as an act of fact-telling. And this renders Byrne’s appeal to an ideal author superfluous. Let me elaborate.


The ideal author is a text-dependent entity; the only evidence readers have of her invitations to make-believe, and her psychological states more generally, is the text.[46] But, as we have seen, the phenomenon of unreliable narration prevents any kind of straightforward inference from the text to the invitations of an ideal author. The alternative on the table is that readers need to proceed by determining abductively how things are fictionally on the basis of what the narrator says, evidence as to the narrator’s reliability, and whatever relevant background information one might have.[47] At this stage, however, the fictional truths have already been discerned. One could, if one chose, go on to attribute to a Byrnean ideal author the act of inviting readers to make-believe these propositions – or belief in them to a Curriean fictional author, for that matter. But there would be little point: all the heavy lifting has already been done.


The task of the remainder of this section is two-fold: the abductive strategy to which I have heretofore merely hinted needs to be developed; and I need to show how this strategy yields an analysis of fictional truth in terms of the speech acts or attitudes of a teller of some kind. The guiding idea underlying both of these undertakings is that the narrator in a fictional story is, in effect, the reader’s informant regarding the fictional world described or embodied by the text.[48] Consider, first, informants outside of fiction. Such persons are charged with providing information about events or circumstances of various kinds – paradigmatically, criminal acts – and pursuant to this charge make assertions (and perform other familiar kinds of speech acts) about these circumstances. But although it is often true that informants reveal desired information by means of their assertions, what they reveal cannot simply be identified with the contents of their assertions. After all, the assertions of informants are frequently defective: they regularly say things which they believe to be false, or for which they lack adequate evidence.[49] Nevertheless by means of such defective utterances they regularly reveal correct information. Let me elaborate.


When interviewing informants, interrogators – let’s call them – do not come to the table empty-handed. Rather they come to the table with various kinds of background information, including both specific information about the circumstances under discussion – perhaps gleaned from forensic evidence or revealed by other informants – and more general information about the world writ large. Moreover, interrogators do not simply passively accept what they are told, but rather actively test an informant’s assertions against this body of information. In addition, in light of the results of these tests – as well as her manner of speaking/ or writing – a picture of the informant’s character and competence emerges. Ultimately an interrogator develops a theory of the events or circumstances at issue which, in his judgment, best fits what the informant has said and implied, the evidence concerning her sincerity and reliability, and both general and specific background information. If he thinks the informant is reliable and sincere, the interrogator will accept as true her account of the circumstances she has described, perhaps filling it in with specific background information and by means of pragmatic inferences from the things she has said. But if he judges her to be unreliable or insincere, some of what she has said will have to be rejected and, if her distortions are judged to be systematic, replaced in the interrogator’s theory by inferences drawn from it. Note: if the interrogator judges the informant to be sincere and reliable, but some of what she says clashes with background information, some of the background information itself may have to be revised. The theory which best fits the evidence will, of course, often depart significantly from what the informant has said; nevertheless, in an important sense this theory has been revealed by her report. 


The relation between reader/listeners and fictional narrators can, in my view, be fruitfully modelled on the relationship between interrogators and informants, subject to the following qualification. Unlike interrogators and informants, reader/ listeners and narrators are capable only one-sided communication. That is, while reader/ listeners can read or listen to the assertions (and other illocutionary acts) of narrators, they cannot make assertions to or ask questions of narrators. Otherwise things are more or less the same. Readers come to the table with both specific background information about the events or circumstances the narrator is describing and general information about the (type of) world in which these events occurred; they test the narrator’s assertions against this information; and in light of the outcome of these tests they develop a picture of her character and competence. Ultimately a reader develops a theory of the events or circumstances at issue which, in his judgment, best fits what the narrator has said and implied, the evidence concerning her sincerity and reliability, and the aforementioned general and specific background information. And the theory which in fact best fits the evidence corresponds to what the narrator has revealed by means of her report.  


When dealing with actual informants, it is important to note that a theory which best fits the evidence may nevertheless prove to be false. How things are – the facts, if you like –serves as an independent measure of theories concerning actual events and circumstances. When it comes to the reports of fictional informants (i.e., narrators), however, there is no corresponding measure of fictional truth; that is, there are no independently existing events or circumstances against which a theory which best fits a narrator’s assertions, etc., can be evaluated. Even if one endorses an ontology of fictional worlds,[50] one does not determine what the correct theory of a fictional text is by appeal to the world that is independently associated with the text; rather one determines which world corresponds to text by appeal to the theory which best fits it. As a result, fictional truth can be directly analyzed in terms of such theories as follows:

AUTHOR-1: “In fiction F, S” is true iff the theory which best fits the narrator of F’s reports, the evidence for her reliability, and relevant background information entails that S,

or, more simply,

AUTHOR-2: “In fiction F, S” is true iff the narrator of F reveals that S.[51]

Consider again the sentence,

Holmes wears underpants

The narrator of the Holmes stories – Dr. Watson – by mentioning dates and places, pragmatically implies that the events he is describing occurred in nineteenth century England. He gives every indication of being a reliable informant about these events except, perhaps, exaggerating his own contribution to them. Since the relevant background information includes the fact that respectable detectives in nineteenth century England wore underpants, and Watson has neither said nor implied anything to the contrary regarding Holmes, the theory which best fits the facts entails that Holmes was so clad. Hence, this is something that Watson reveals. 


V: Loose Ends


There remain a number of residual issues which are worth addressing at this point, the most pressing of which is the problem of mindless fiction. As it stands, it looks as though my analysis can only be applied to works with explicit narrators. But as Byrne and others rightly point out, many works lack narrators.[52] And to insist, as Currie does, that any such work is contradictory because “it is told as known fact and … there is no one there to tell it” seems to require unpalatable bullet biting.[53] My strategy is to follow Currie and accept that in every fictional work there is a non-actual fact-teller – let’s call this figure a “narrative informant.” The motivation for doing is twofold: first, determining the theory which best fits a text involves, in part, discerning what the teller pragmatically implies by her utterances, but, as above, Gricean implicative mechanisms presuppose the teller is engaged in fact-telling discourse; and second, a fictional story’s actual teller – the author – engages only in fiction-telling when telling the story. In a work with an explicit narrator, the narrative informant just is the narrator. But contra Currie, in a mindless fiction the narrative informant is not an inhabitant of the fictional world she is describing; rather, she is, rather a non-actual/ non-fictional fact-teller. As a result, there are no fictional truths about her – fictional truth is wholly matter of what occurs in the fictional world – and so the threat of contradiction is forestalled.


Two sorts of worries about this manoeuvre do jump mind, however. First, one might balk at the intelligibility of someone fact-telling about a world she does not inhabit. What I want to point out, however, is that such activity is commonplace. Talk of Heaven or the afterlife is a familiar example. Second, one might argue that a narrative informant in mindless fiction is as shadowy as I have charged the figures Currie and Byrne invoke with being. This charge is, in my view, ill-founded. I have been modelling an author engaged in oral storytelling on an actor portraying a character onstage. The charge I made against Currie’s fictional author and Byrne’s ideal author was they were distinct from both the character and the actor portraying her – the former is instead a meta-character and the latter a meta-actor. On my view, in contrast, the character an author of mindless fiction portrays or pretends to be just is the narrative informant – a non-actual fact-teller describing a world she does not inhabit.


A second potential worry for the analysis developed here is that the theatrical model of the author/ narrator relation breaks down in the case of written, as opposed to oral, works of fiction. Fictional composition is a distinct kind of activity from story-telling performance; in particular, the former does not require that the agent portray the character she is creating or engage in any kind of pretense whatsoever.[54] And authors of written fiction need never engage in any public performances (e.g., readings) of their works.   More generally, in composing fiction, authors need not engage in any acts of fiction-telling at all, regardless of whether fiction-telling is best understood as engaging in assertive pretense[55] or the performance of sui generis fictional illocutionary acts;[56] and in composing non-fiction, authors need not make assertions or perform any of the other illocutionary acts characteristic of serious or fact-telling discourse.[57]


Although I do think that some such distinction between composition and performance is in order, it does not pose serious difficulties for the analysis of fictional truth at issue. Although authors of written fiction need not pretend to be the narrative informants of their works, appropriate engagement with written fiction arguably requires that readers imagine of authors of such fiction that they in fact are authors of written non-fiction.[58] And by disseminating their written works, authors of non-fiction can reveal information about events and circumstances much in the same way that informants can by means of their assertions. As a result, by broadening the notion of a narrative informant to include both the fact-tellers that authors of oral fiction pretend to be and the non-fiction writers that readers of fiction imagine authors of written fiction to be, the analysis of fictional truth in terms of the revelations of narrative informants can be retained without revision. 


A third potential difficulty for this analysis is presented by the phenomenon of story-telling narrators.[59] A story-telling narrator, recall, is an explicit narrator who clearly indicates that the story being presented is fictional and, hence, that she is a fiction-teller rather than a fact-teller. But a narrative informant is a fact-teller and, so, insofar as an author pretends (or is imagined) to be this latter figure, she is not pretending (or imagined) to be the narrator of her work. Moreover, it will not do to suppose such works are exceptions wherein authors pretend (or are imagined) to be fiction-tellers; after all fiction-tellers do not reveal anything by means of their utterances.


The best way to handle this phenomenon is to take fictional works with story-telling narrators to generate stories within stories. The outer (and typically much thinner) story concerns the narrator’s act of fiction-telling and any surrounding events; and the inner story is the story the narrator relates by means of this act. Now, on my view, when she tells the outer story, the narrator engages in fact-telling, making assertions about who she is and what she is doing (namely, fiction-telling) and the like. And when she tells the inner story, the narrator pretends to be a fact-teller – in particular a non-actual (from her perspective)/ non-fictional fact-teller. The actual author pretends (or is imagined) to be the story-telling narrator when he tells the outer story and pretends to be this fact-teller when he tells the inner story. Finally, fictional truth in the outer story can be identified with the revelations of the story-telling narrator, whereas fictional-truth in the inner story can be identified with the revelations of this other figure.


The final potential worry that will be addressed is that it remains unclear how the analysis defended here avoids some of the difficulties – in particular, the problems of inconsistent fiction and irrelevant truths – that beset the alternatives. First, consider, again, a work in which it is true that the protagonist has refuted Gödel’s theorem. In order for the revelation analysis to accommodate this possibility, it will have to turn out that the theory which best fits the evidence entails that this mathematical breakthrough has occurred – and that the narrative informant who reports it is both sincere and reliable. But given the impossibility of this occurrence, it seems that likely that a theory which entails that Gödel’s theorem was not refuted, and that the narrative informant, if sincere, was certainly unreliable, would at least equally well, if not better, fit the evidence. And second, given that the Holmes stories, for example, occur in a world very much like ours, it seems plausible to suppose that a theory which entails that Julius Caesar had the number of hairs on his head when he died that he in fact had would again fit the evidence at least as well as a theory which does not entail this.


Both of these problems can be resolved, however, by appeal to background information. If the background information in the Holmes stories fails to include propositions about the numbers of hairs on Caesar’s head, then it is unlikely that the theory which best explains this information, together with Watson’s assertions and implications and the evidence as to his reliability, will entail some such proposition about Caesar. And if the background information in a work about the putative refutation of Gödel includes the proposition that actual mathematics fails to hold in the world being described, then the theory which best fits the evidence will likely entail that the narrator is reliable and that the protagonist refuted Gödel’s theorem. But if it alternatively includes the proposition actual mathematics holds, it is more likely that the theory which best fits the evidence will entail that the narrator is unreliable and that Gödel’s theorem remains unrefuted.[60]


The obvious rejoinder is that this presupposes principled grounds for determining what the relevant background information is for any given work of fiction. And it is far from clear that any such procedure is forthcoming. The three central sources of background information include (a) facts about the actual world, (b) inter-fictional carry-over – information from other works of fiction – and (c) information derived from the authorial and critical discussions of the story. But it remains controversial what and how much, if any, information in each category is relevant in any given case. In some circles, the role of information about or provided by the author remains especially controversial. One might attempt to proceed here by something akin to a process of reflective equilibrium, such as the following: (i) identify some relatively uncontentious fictional truths; (ii) working backwards, discern what background information is required to yield such truths; (iii) inductively develop general principles relating background information to fictional works; and (iv) test these principles by applying them to further cases. Whether or not such a procedure could ultimately prove successful is, I take it, an open question. But even if it were to fail, it is not clear that this would undermine the revelation analysis of fictional truth. The mark of a good theory is not that it yields no indeterminacies but rather that it yields indeterminacies only in the right places. To the extent that the relevance of various sorts of background information to the interpretation of fictional texts is indeterminate, questions of fictional truth are themselves indeterminate.





Byrne, Alex (1993), “Truth in Fiction: The Story Continued,” Australasian Journal of Philosophy 71: 24-35.


Currie, Greg (1986), “Fictional Truth,” Philosophical Studies 50: 195-212.


---- (1990), The Nature of Fiction, Cambridge: Cambridge University Press.


---- (1995), “Unreliability Refigured: Narrative in Literature and Film,” Journal of Aesthetics and Art Criticism 53: 19-29.


Grice, Paul, (1975), “Logic and Conversation,” in Syntax and Semantics, vol. 3: Speech Acts, P. Cole and J.L. Morgan (eds), New York: Academic Press.


Hoffman, Sarah (2004), “Fiction as Action: Currie and Searle on Speech Act Theory and the Nature of Fiction,” Philosophia 31: 513-29.


Kania, Andrew (2005), “Against the Ubiquity of Fictional Narrators,” Journal of Aesthetics and Art Criticism 63: 47-54.


Krasner, Daniel (2002), “Semantics and Fiction,” Erkenntnis 57: 259-275.


Levinson, Jerrold (1996), “Intention in Interpretation and Literature,” in The Pleasures of Aesthetics: Philosophical Essays, Ithaca, NY: Cornell University Press, pp. 175-213


Lewis, David (1978), “Truth in Fiction,” American Philosophical Quarterly 15:37-46.


---- (1983), “Postscript to ‘Truth in Fiction,’” in Philosophical Papers, Vol. 1, New York: Oxford University Press, pp. 277-280.


Matravers, Derek (1995), “Beliefs and Fictional Narrators,” Analysis 55: 121-2.


Searle, John (1975), “The Logical Status of Fictional Discourse,” New literary History 6: 319-332.



[1] The concern is with truth in fiction as opposed to truth through fiction.

[2] See, e.g., Lewis (1978), Currie (1986, 1990), Byrne (1993), and Matravers (1995).

[3] Currie (1990).

[4] Byrne’s (1993) invites this interpretation.

[5] Note: Lewis’ (1978) analysis only roughly fits into this format.

[6] Levinson, 1996.

[7] Lewis, 1978.

[8] Currie, 1986, 1990.

[9] Byrne, 1993.

[10] Lewis, 1978.

[11] Lewis (1878) p. 42.

[12] Lewis (1978) p. 45.

[13] Lewis (1978), p. 44.

[14] Lewis (1978), p. 40, n. 7.

[15] But see Krasner (2002), p. 265.

[16] Lewis (1978), p. 41.

[17] Currie (1986), p. 197.

[18] Lewis (1978), p. 45.

[19] Lewis (1978), p. 45.

[20] Lewis (1983), p. 277.

[21] Currie (1990), p. 69.

[22] Currie does at times seem content to identify fictional authors with reliable explicit narrators. Given the simplifying assumptions we have been making, however, this possibility need not detain us.

[23] Currie (1990), p. 125.

[24] Currie (1990), p. 80. An informed reader is someone who “knows the relevant facts about the community in which the work was written” (1990, p. 79).

[25] Currie (1990), p. 80.

[26] Currie (1990), p. 124.

[27] Of course, the fictional author of an inconsistent fiction could not tell it as known fact.

[28] Byrne (1993), p. 28.

[29] Currie (1990), p. 125.

[30] Of course, more subtle actors and storytellers might indicate unreliability other than by their manner of speaking. And the script or the story by itself might also suffice without the need for any ham acting.

[31] Matravers (1995), p. 122.

[32] Byrne (1995), p. 30.

[33] Currie’s view entails that mindless fictions – or, more generally, fictions which exclude a teller – are contradictory. Currie himself simply embraces this consequence:

“[on] my theory mindless fiction generates a game of make-believe in which we are called upon to make believe contradictory things: that it is told as known fact and that there is no one there to tell it.” (1990, pp. 125-6).

[34] Byrne (1993), p. 31.

[35] Byrne (1993), p. 33.

[36] Byrne (1993) p. 31.

[37] Note: the considerations raised here pose a difficulty for an analysis of fictional truth in terms of invitations to make-believe made by the actual author as well as Byrne’s ideal author..

[38] Byrne (1993), p. 30.

[39] Byrne (1993), p. 32.

[40] Byrne (1993, p. 32, n. 31) endorses Currie’s (1990) account of sui generis fictive illocutions in his discussion of explicit invitations to make-believe.

[41] Searle (1975) p. 324.

[42] Currie (1990, pp. 14-5) does respond to Searle’s argument. But Hoffman (2004) has argued that Currie’s defense of his uniquely fictive illocution is ultimately unsatisfactory.

[43] Or, perhaps, a fictional indeterminacy.

[44] Grice, 1975.

[45] The central example Byrne works through (1993, pp. 33-4) does have something of an abductive flavour to it despite his official adherence to the explicit/ implicit invitation strategy.

[46] Certain kinds of background information including inter-fictional carry-over may also be relevant. More on this below.

[47] What counts as relevant background information will be addressed below.

[48] The fictional world embodied by a text is just the collection of fictional truths embodied by it.

[49] Searle (1975), p. 332.

[50] Although I am committed to the utility of the “fictional worlds” idiom, I do not take it to have ontological import.

[51] It is, of course, theoretically possible that more than one theory will equally (and best) fit the evidence. In such circumstances, what a narrator reveals will have to be identified with what is entailed by all such best theories. 

[52] See Byrne (1993) and Kania (2005).

[53] Currie (1990), pp. 125-6.

[54] An author may, of course, engage in imaginative activities of various kinds as part of the creative process but there is no requirement that she do so – failing to do so undercut her claim to be engaging in composition. 

[55] Searle, 1975.

[56] Currie, 1990.

[57] See Author’s Article for a more developed version of this argument.

[58] Author’s Article. I would argue that engagement with oral fiction requires similar imaginative activity on the part of listeners.

[59] See Kania (2005), p. 52.

[60] If the background information contains neither of these propositions – and there is simply no grounds for thinking either theory better fits the remaining evidence – there is no fact of the matter whether or not it is fictionally true that Gödel’s theorem has been refuted.