Navigation – Plan du site
Penser la fiction

Psychologizing the Semantics of Fiction

Psychologiser la sémantique de la fiction
John Woods et Jillian Isenberg


Les théoriciens sémantistes de la fiction cherchent typiquement à expliquer nos relations sémantiques au fictionnel dans le contexte plus général des théories de la référence, privilégiant une explication de la sémantique sur le psychologique. Dans cet article, nous défendons une dépendance inverse. Par l’éclaircissement de nos relations psychologiques au fictionnel, nous trouverons un guide pour savoir comment développer une sémantique de la fiction. S’ensuivra une esquisse de la sémantique.

Haut de page

Texte intégral

1. Fictional relations

  • 1 Not to overlook the dead of Guernica. See below, section 7.

1Philosophical theories of fiction pivot on issues that are dominantly relational. Semantic theorists are interested in semantic relations. Do we refer to fictional beings and events? Do fictional sentences relate to things in ways that assign them truth values? Do fictional sentences stand in inferential relations? Do we ourselves bear relations to these sentences in ways that qualify as belief?  Philosophers of art also have a stake in the relational. What is the relation between a painting and what it represents? What is its relation, if any, to the intentions of the artist? In what emotional, affective, and other psychological relations do we stand, not only to the fictional, but also to (say) the dead of Guernica?1. Semanticists want an account of reference, truth and inference in fictional accounts. Aestheticians want an account of more generally representational matters, and of the psychological features of art. Both camps share an interest in doxastic relations – in the conditions (if any) under which the fictional can be objects of belief, and the conditions (if any) under which such beliefs might pass for knowledge.

  • 2  [Frege, 2001, p. 202-203; Sainsbury, 2005, p. 202-209]

2Semantic theorists of fiction are faced with three framework issues. One is whether to honour Parmenides’ Rule, according to which, since quantification is existentially loaded, there is nothing that doesn’t exist. The second is whether to accept the Non-Existence Postulate, according to which the objects of fiction don’t exist. The third framework question is whether to go along with Frege’s Dismissal, according to which, since fiction has no scientific importance, it should be accommodated, if at all, in a semantic theory that does justice to our scientific concerns – making of fiction a sort of poor cousin, sleeping below-stairs2. If this means that finding a place for fiction is an ad hoc tight squeeze for standard semantics, so much the worse for fiction; it is all that is owed to fiction by the philosophy of language.

  • 3 For recent writings to the same effect, see Woods (ed.) 2010c, Fictions and Models: New Essays, Mun (...)
  • 4 Not even the relation of denying their existence.
  • 5  Standard semantic theories take the position that for a name to refer, it must have a referent (th (...)

3It will help clarify our objectives in this paper to simply declare ourselves without ado on these framework questions. We accept Parmenides’ Rule, we accept the Non-Existence Postulate and we reject Frege’s Dismissal. Accordingly, we hold that a good theory must preserve the root idea that fiction leaves no metaphysical footprints, that no object is an object of fiction. We also hold that no general purpose theory of reference, truth and inference will do justice to the fictional. A good semantic theory of fiction must be purpose-built3. Let us be clear about what is required. Parmenides and Non-Existence together entail the impossibility of our bearing relations of any kind to the fictional  semantic, psychological or otherwise4. The rejection of Frege’s Dismissal entails the unavailability of established ways of accommodating reference-failure5. Fresh thinking will be required. In the pages to follow, we will have abundant occasion to speak of “our relations” to fictional objects. Since there aren’t any, this is a bit awkward. Let that pass for now: “our relations to fictional objects” is a façon de parler. (we might even say that they are fictions.)

  • 6 The Postulate says that all fictional objects are non-existent. But “all” is a quantifier, and the (...)
  • 7 See footnote 35 at the end of the paper for some indication of the problems we see in alternative a (...)

4It is well to note early in the proceedings that, if the grammar of English is anything to go on, Parmenides’ Rule is deeply counterintuitive – so counterintuitive that it is scarcely possible to state the Non-Existence Postulate without violating it6. We don’t want to shrink from this; nor do we refuse to consider that a theory of fiction might do better if the Rule and the Postulate were dropped. This is not the place to explore that option in any detail, but it gives us occasion to lay down a marker, to be redeemed elsewhere. While the alternative approaches are well-worth working up, our present view is that the ensuing theory of fiction will not do conspicuously better on the score of intuitiveness than the theory that will occupy us here7. This leads us to a sobered-up appreciation of the theory of fiction. It is that fiction holds out only slight promise for intuitively satisfying theoretical workings-up. If this is so, not only is it advisable to know it going in – for it allows us to tailor our expectations accordingly – but it also gives us a valuable constraint on a theory’s adequacy.

Not making too much of intuitiveness

  • 8 In other contexts, the present admonition is rather small beer, indeed hardly worth mentioning. Thi (...)

When it comes to a theory of fictionality, counterintuitiveness is an unreliable reason to reject it8.

5Let us also pause to give preliminary notice of how our own semantic proposals for fiction will unfold. We think that the fictional semantics under the constraints of the Rule and the Postulate will considerably overlap the theory produced with those constraints removed. Roughly speaking, the semantics for fictional reference will be different, but the semantics for fictional truth and inference will be the same. This is something we will develop a bit more in section 6 et passim. But for now, there is a point of particular importance to which we should call attention. It is widely assumed that a purported referent’s ontological status will be consequential for its semantics. In our view, this is only partly true. We say that the principle holds for real objects, but not for fictional objects. Whether or not Holmes exists can be quantified over is a real constraint on a theory of reference for Holmes. But it is a matter of indifference for the concomitant theories of truth and inference. Consider an example. Whether or not Holmes exists, whether or not Holmes can be quantified over, it remains a fact that sentences rawly in the form “Holmes lived in Baker Street” are made true by the story and made untrue by the world. One of the main jobs of a truth theory for Holmes is to make this an abidable inconsistency  indeed, preferably no inconsistency at all. Similarly, irrespective of whether Holmes exists or is a value of a variable bound by quantification, the sentence “Holmes probably had quite a high I.Q.” is inferable from the sentences of the text and is true and false in the same senses in which “Holmes lived in Baker Street” is true and false. As we say, full details are better left for another time, but not before remarking the further and related oddity that if our approach is sound, it turns out that the Rule and the Postulate are deep truths for the ontology of fiction but not deep truths for two-thirds of the semantics of fiction.

  • 9 A brief historic note: Parmenides’ Rule is very old, possibly originating with Parmenides himself. (...)

6Let us also note a certain tactical weaknessinvolved in our assumptions. For if, as with “the present king of France”, “Zeus” and “Vulcan”, there is nothing that “Sherlock Holmes” denotes and if, as with “The present king of France is bald”, “Zeus lived on Olympus” and “Vulcan was said to explain the orbital perturbations of Mercury”, there is nothing for “Holmes lived in Baker Street” to be true of, why would we not semanticize the latter in one or other of the standard and well-understood ways in which we semanticize the former? And, if so, wouldn’t we have a good enough reason to embrace, rather than reject, Frege’s Dismissal? If all theories of fiction are destined for a certain counterintuitiveness, why not mitigate the attendant disappointment by selecting one of the best-understood and formally precise semantics for “empty” terms, and adapt it to fictional contexts, counterintuitiveness and all? At least this way we won’t have risked the development of an untried new semantics. If both approaches give intuitively unsatisfactory accounts of the fictional, why not stick with the semantically tried-and-true? Our simple answer to these good questions is that there is counterintuitiveness and there is counterintuitiveness. We will try here to motivate the basic features of a purpose-built theory for fiction, a theory that is designed in such a way as to make Frege’s Dismissal an unattractive assumption for fiction. Our hope is that overall there will be more to like about this kind of stand-alone account than any of the mainline treatments of reference-failure9.

  • 10 Among recent contributions, see [Sainsbury, 2005].
  • 11 Among recent contributions, see [Coplan, 2004; Hjort and Laver eds., 1997; Meskin and Weinberg, 200 (...)

7In addition to these framework issues, semanticists of fiction also face a priority-question. In a first approximation, this is the question of which relations to the fictional are theoretically fundamental, and which are best analyzed in the light of what the theory says about the fundamental ones. Philosophers of language tend to privilege semantic relations, and deal, if at all, with psychological relations in a more derivative way10. Although aestheticians are as a rule not quite so priority-minded, there is some evidence of a priority ranking in the other direction11.

8We are philosophers of language fully at home with the semantic agenda for theories of fiction. But we reject the semanticists’ standard prioritization. Our view is that part of what makes fiction a peculiar challenge to theories of reference is precisely that it resists the priority ranking favoured by philosophers of language. So as our first point of departure, we advance :

The Psychology first principle

  • 12 We are not unaware of the historical provenance of the Psychology First Principle. To the best of o (...)

The best way to produce a good theory of semantics for fiction is to deal first with the issue of our psychological relations to the fictional, and to develop the semantic account under the guidance of the psychological account12.

9Why would one be drawn to the Psychology First Principle? It is because psychological data are both richer and epistemically more secure than the rather more intuitive data on which semantic theories are based. Consider an example. It is a datum for a semantic theory of fiction that beings like us refer to fictional beings and fictional goings-on. True, as the theory itself evolves, it might turn out that the theory overturns this datum. (Indeed if it subscribed to Parmenides and Non-Existence, it would have to overturn it.) But there is no doubting its initial function in the theory and its initial importance. If upon reflection it seems best to discard it, then it must fall to the theory that does so to explain the intuition away. On the other hand, it is a datum for a psychological theory of fiction that we have psychological responses to fiction, that we cry at the death of Nell, and that we are seized with apprehension about whether Karla will actually surrender to Smiley’s team. Given the generally problematic character of fiction, there is more that’s fishy about the reference datum than there is about the crying or the anxiety datum. It is easier to call into question whether we did in fact refer than whether we did in fact cry or seize up – for were there not tears on our cheek, and didn't our pulse-rate spike upwards?

10We can, if we like, reduce the above observation to a slogan. That we refer to Nell is an intuition.That we cry upon reading those passages in Dickens is a fact. So the slogan is:

Facts before intuitions

  • 13 Another way of making the same point is that, to the extent that its subject matter permits it, we (...)

In constructing a theory whose motivation includes intuitions and facts, give pride of place, albeit defeasibly, to the facts13.

  • 14 See here [Woods, 2006].

11It bears repeating that the psychological facts with which we propose to launch our semantic account of fiction are not free-on-board. They are compromised by our subscription to Parmenides’ Rule and the Non-Existence Postulate. If we cry at Nell’s suffering and if our heart races for Karla to follow through with his defection, the crying and the racing are informed by the knowledge that these things never happened, that there is no death of Nell, and there is no defection of Karla. There is no Nell. There is no Karla. These are paradigmatic examples of what we might call the “double-aspectness” of our responses to fiction.14 We react to what the story recounts knowing that there is nothing whatever to what is recounted.  So a question for the psychological account is

The possibility of affective response?

How is it possible to cry for Nell and nervously await Karla, knowing that nothing is Nell and nothing is Karla?

12The double-aspectness of our relations to the fictional is for us the primary datum for a psychological theory of the fictional. And given our subscription to the Psychology First Principle, it is also desirable that double-aspectness, or some clear counterpart of it, figure prominently in the theory’s adaptation to semantic concerns. In a word, if double-aspectness is fundamental to a satisfactory analysis of our psychological relations to thefiction, it or something like it will likewise be fundamental to a satisfactory analysis of the semantics of the fiction.

  • 15 Here, and in what follows, we adapt with revisions from passages in [Woods, 2009] and [Woods, 2010a (...)

13On the face of it, the sentence, “I am crying for Nell, and, of course, there is no such thing as Nell”, is dissonant (some would say paradoxical). People tend to respond badly to paradox if there is no clear way of disarming it. People experience a kind of cognitive dissonance about inconsistencies they are unable to resolve, that is, to analyze away. What is puzzling about the present case is the utter lack of cognitive dissonance in people who have affective responses to fiction. A further complication is that they seem, in the general case, not to know how to resolve the inconsistency without, in so doing, rendering the psychological fact simply inexplicable. Another way of saying this is that there is no pretheoretic way to resolve the inconsistencies engendered by these affective responses. It is different with semantic inconsistencies. We refer to Holmes knowing that there is no Holmes to refer to. People have little difficulty in proposing ways out. Holmes exists in the story, but not in reality; there was no one we referred to, just a character in a story; and so on. Think of these measures what we might, it is a much harder sell that our tears didn’t spill and our hearts didn’t race15.

  • 16 The exertions of dialetheic logicians notwithstanding.

14It is instructive to compare inconsistencies that don’t bother us with inconsistencies that do. When Russell reported to Frege the inconsistency that destroyed intuitive set theory, there was much gnashing of teeth. It was an inconsistency that simply couldn’t be abided16. Abided inconsistencies are an interesting phenomenon. They call for explanations. Perhaps the most common such explanation is that, properly understood, the abided inconsistency is not actually inconsistent, and that the fact of its abideness is tacit recognition of its underlying consistency. This certainly is the position of pretense theorists of fiction. Our sadness at Nell’s non-existent death wasn’t sadness after all. It was pretend-sadness, in light of which the inconsistency dissolves. It is an interesting speculation. The trouble with it is that there isn’t the slightest empirical evidence to suggest that it is true.

  • 17  See section 18 below for predictions about the sort of empirical date that will prove useful in re (...)

15There is an interesting asymmetry here. Fictional practice – the writing, reading and discussion of it – embodies this double-aspectness without a murmur of cognitive turbulence. But theories of this practice are pretty much guaranteed to be counterintuitive, thus producing a dissonance of its own17.

2. Texts

16No one doubts that our tears are real. But isn’t it possible, even so, that they were occasioned by a mental state which is not in fact the relational state of being sad about poor Nell or being shattered by her death? This is not a possibility to be dismissed out of hand. Indeed, it takes little reflection to see that, under the present assumptions, something like this pretty well has to be the case. Even so, if we embrace it, there are some tricky issues that press for attention. One is how to characterize the state that appears to be sadness at Nell’s death. The other is how to explain what it is about this state that triggers those tears, given that the appearance of our standing in an emotional relation to Nell is notone that takes us in.

  • 18   [Walton, 1990; Saatela, 1994; Yanal, 1999]

17A suggestion similar to the pretend-emotion thesis of Kendall Walton is that while we do not, as some theorists say, stand in emotional relations to fictional persons or events, we are nevertheless capable of “quasi-emotions”18. Walton’s quasi-emotion is an affective state, whose possession “makes it fictional of ourselves” that we experience the emotion. We demur from this, mainly ongrounds that, if true, the making-fictional-of-ourselves of our psychological states leaves no empirical trace. “Quasi” is Latin for “as if” or “not quite”. Its use by philosophers is suspect. Consider a case. There is something x which you want to account for. One of the things you can say about it is that it isn’t F, although it resembles F-things; or it is partly F. Well, we say, that’s what it isn’t and that’s what it’s like in part. Now tell us what it is. “Quasi-F-thing” is not an answer to this question. “Quasi-F-thing” lacks descriptive import. You say “quasi” when you don’t know what else to say. “Quasi”-talk is talk faute de mieux. Better that “What is x?” be answered substantivally, and that “quasi-F-thing” be applied afterwards simply for referential ease. So let us recur to the point that the distinctive feature about our affective responses to fictional goings-on is that they engage us – and often grip us or shake us with laughter – notwithstanding our certain knowledge that there is nothing there to respond to. What, then, should we call these double-aspected responses? Let us call them “double-aspected responses” (“DA-responses”, for short).

18When we read the relevant passages of Dickens’ text, we fall into the DA-emotions of shock and sadness. These states are not as they appear. When we are seized by DA-sadness over Little Nell’s death, there is indeed a relation in which we stand to something. But that something is neither Nell nor Nell’s death; there areno such things. Rather the relation in which we stand when we have that quasi-sadness is to Dickens’ text (which, perversely, pretense theorists call a “prop”). This we hold to be a defining feature of DA-emotions. Suppose that the DA-emotion view had legs. Then one of its virtues would be that it makes intelligible a motivation for the pretense or make-believe theory. For we could now conjecture that what pretense-theorists were trying to get at was a theory of a type in which our relations to the fictional are DA-relations in much the way that our sadness about Nell is DA-sadness. Of course, we are now left with the task of explaining what it is about our connection to Dickens’ text that could possibly trigger our tears. In any event, pretense, make-believe and quasi-emotion theories embed a common assumption and a valuable insight. It is that “Nell-sadness” and real sadness have some of the same causal effects without having the same kinds of causal antecedents. They are, so to say, “causally askew” from the real thing.

  • 19 They are, as David Lewis puts it, “told as known fact” ([Lewis, 1983, p. 270]).

19We are proposing that a suitably vigorous account of our semantical relations to fiction should await convincing solutions to the problems generated by our affective and other psychological relations to it. In this regard, a further point must be taken note of. It is that in the passage in which Dickens chronicles the death of Nell, there is the appearance of its being reported, and that in the passage in which Le Carré describes Karla’s defection, there is likewise the appearance of its beingreported. Of course, no one seriously supposes that these sentences do in fact report the events they appear to describe. Even so, they do have the appearance of reports. Reports are rendered in the assertive mode. They narrate the states of affairs they report. They are fact-stating modes of speech19. The syntactic heart and soul of reports is the indicative sentence. The pragmatic heart and soul of reports is the declarative utterance. What our present speculations help us to see is that when a declarative sentence is used in non-reportative ways, it can – and typically does – retain the appearance of reportedness. It retains the appearance of fact-stating. In so saying, we have a fundamental fact about literary texts:


Stories carry the non-delusive appearance of histories.

20The qualification “non-delusive” is important. It provides that readers of fictional texts respond to them as reports knowing full well that there is nothing to report. This is double-aspectness again. Reading a story is standing to a text in a certain relation. Reading a story with understanding, that is, knowing it to be just a story, is a different relation to the text. Reading a story is not a DA-relation to it, but reading it with understanding assuredly is. In so saying, we have occasion to suggest a principled connection between the apparency of apparent reports and the DA-ness of our reading them with understanding.

  • 20  Against this it might be objected that we also cry at Juliet’s despair in Prokofiev’s ballet. We d (...)

21We would do well to emphasize the kind of apparentness we are ascribing to literary texts. The sentences of Smiley’s People have the appearance of reports and the appearance of fact-staters. These appearances are neither illusions nor pretendings. The illusion-interpretation is ruled out by the fact that the appearance of fact-stating does not at all disguise the non-fact-stating character that Le Carré’s sentences actually possess. Likewise, the pretend-interpretation of their apparently reportative character is ruled out by the fact that our merely pretending that these sentences are reportative renders unintelligible the empirical reality of our reactions to them, still less that we are unconsciously playing at their being reportative20.

3. Causal powers

22Assertions have causal powers. The same is true of apparent assertions. One of their standing powers – albeit a defeasible one – is to push the psychological buttons of those to whom they are directed. If someone reports to us the death of the little girl next door, it may make us cry. If someone tells us of the final gripping minutes in the match between Italy and Spain, our heart-rate may spike. What the fictional cases suggest is that it is not a condition on the button-pushing causal efficacy of these narratives that they be considered true. On the contrary, what they suggest is (1) that psychological or psychophysical reactions to reports turn on the appearances they carry, and (2) that knowing that an appearance is false does not preclude a report’s carrying it. (So, again, appearances aren’t illusions and aren’t pretendings.) Sooner or later, this is a fact of which we will need to take semantic note.

23Assertions are kinds of telling. Being told things is a central feature of our cognitive lives. Virtually everything that we know or think we know we have been toldor depends on something that someone or other has been told. The economies of such arrangements speak for themselves, and do so in a way that suggests an adaptive advantage for them. This further suggests that the knowledge-via-testimony modality is a deeply primitive arrangement. If this is right, then we would have it that

Made to believe

It is typical of a report that φ that φ be believed on receipt of it.

The connection between being told that φ and believing that φ is defeasibly causal.

24Here, too, confusions should be avoided. It is not at all essential to the point at hand that the beliefs induced by this modality be epistemically respectable (for example, that they be true à la mode de Tarski or Kripke). It suffices for present purposes that it be an empirically supported defeasible fact about reportative utterance that it causally induces psychological acceptance of its message. Except when there is particular reason not to, beings like us believe what we are told, and do so in consequence of the belief-inducing powers of reportative utterance.

25The empirical record also advances a further characteristic of what it is to be told things. It is that we tend to remember what we’ve been told. Much of it goes into short-term memory and much of it goes into long-term memory. It is a peculiarity of these reminiscential transactions that what is told as the fiction of so-and-so’s having done such-and-such is remembered as so-and-so’s having done such-and-such. A further peculiarity is that it is not at all a condition of the first peculiarity that it be forgotten that so-and-so’s doing such-and-such was told you as a fiction. Double-aspectness again.

  • 21 [Audi, 2004; Coady, 1992; Adler, 2002; Lackey, 2008; Woods, 2010b]  For present purposes, it is eno (...)

26The causal powers thesis is not an undisputed one in the rapidly expanding literature on testimony21. In as much as it is the central part of the present suggestion about how attestation is related to belief and to the emotions, this is a question which theorists have a stake in closing, one way or the other. But as long as the causal powers thesis remains an open possibility, there is a further possibility to consider. What the present hypothesis suggests is that

Causal mixtures

Double-aspected responses are produced by mixed causal signals. It suggests in particular that for the kinds of case under review, these competing causal signals aren’t (respectively) strong and weak enough for the one to neutralize the causal efficacy of the other.

  • 22 We don’t want to minimize the complexity of the causal mixtures thesis. There are cases not well-ca (...)

27This contrasts markedly with standard situations of belief-revision. If up to now you believe that φ and now come to believe that ψ, and also that ψ is inconsistent with φ, then for a great range of cases your belief-forming mechanisms will erase your belief that φ. This is precisely what doesn’t happen in double-aspected situations. Not only does  your belief that no one is Holmes not erase your belief that Holmes lived in Baker Street, not only does your belief that no one is Holmes not erase your memory that Holmes solved the case of the speckled band, but it is not something that you are unaware of or troubled by22.

4. Belief and DA-belief

28Believing that Barbara lived at 57 Berczy Street is standing in a certain kind of doxastic relation to Barbara; it is standing in the relation of belief to the proposition that that is where she lived. Believing that Holmes lived at 221B Baker Street has a like appearance. Believing this about Holmes feels like believing that about Barbara. It feels like this notwithstanding our knowing full well that there was no place whatever that Holmes lived. Our believing that Holmes lived at 221B Baker Street occurs only if we are in the appropriate mental state. Believing that he lived there knowing full well that he didn’t demands a certain kind of mental state. Believing that he did, knowing that he didn’t, suggests a kind of inconsistency. But if it is an abided inconsistency, the state of mind that abides it must be a quite particular mental state, a state in which the inconsistency is retained and persisted with or a state in which the inconsistency is somehow dissolved.

29Our belief about Holmes resembles, but is different from, our belief about Barbara. Our belief about Holmes is a DA-state, and the doxastic relation in which we stand when that state is activated is a DA-relation. The similarity between DA-belief and DA-sadness speaks for itself. But there is a difference. Belief strides the divide between the psychological and the semantic. Believing that is believing that is true. Believing that Holmes lived in Baker Street may well involve being in a DA-psychological state, but it also involves, or purports, a connection with truth. The inconsistency question again bestirs itself. Believing that Holmes lived in Baker Street is believing true what we know to be untrue. Why isn’t this an inconsistency – an unabidable inconsistency – in the manner of the set that is both a member of itself and not?

30What are we to say of these double-aspected situations? However we may be minded to answer this question, there is a fact of some importance that should be taken into account. It is that your belief that Holmes lived in Baker Street, unlike your belief that Barbara lives in Berczy Street, is attended by significantly different possibilities of action. You could in principle visit Barbara in Berczy Street, but you couldn’t visit Holmes in Baker Street. You could visit Baker Street, but not visit Holmes. This raises a critical question about the identity conditions for beliefs. If apparently different tokens of some same belief type turn out to produce radically different types of action possibilities, can it be that they really are beliefs of that same type? A theory of fiction will have to say something about this. We lack the space to consider here in any detail how such a logic would go. But before bringing this section to a close, we might briefly consider in a quite general way one possible line of development.

31Suppose, then, that we took seriously the suggestion that


The belief that Barbara lives in Berczy Street and the belief that Holmes lives in Baker Street are not tokens of the type “the belief that x R y.”

  • 23   In the words of [Schroeder and Matheson, 2006].

32Suppose that, upon reflection, we decided that the Berczy Street case and Baker Street case carry different ontological commitments, with the first requiring that the believer stand in some relation to Barbara (and perhaps to Berczy Street), and the second requiring only that the believer stand in some relation to  let us say it briefly  some causally efficacious text. If this – or something like it – were so, then the relations imputed in these two cases could not plausibly be instances of the same relation. It would be, in the fictional case, a “distinct cognitive attitude”23.

5. A taxonomy of fictional sentences

33As we have developed it here, a fact of first importance for an account of psychological relations to the fictional is that the right-hand relatum is a text, not a person or a situation in which a person finds himself or a fact about a person. Again, our sadness at Nell’s death is a DA-emotion triggered not by the death as reported in a certain way to us, but by a text apparently reporting a death which the understanding reader knows never happened. Part of what we have been supposing is that distinctive features of the psychological account of such states and relations will also appear in the account of the semantic states and relations we feel ourselves to bear to the fictional. In particular, for it to be true that fictional texts cause DA-responses – both affective and doxastic – there should be some strictly semantic feature (or features) of those texts in the absence of which the psychological account would lose some of its footing, if not all of it. More particularly still, a semantic theory of the fictional should specify some feature (or features) of fictional texts which afford some natural explanation as to why DA-states aren’t states of unabided inconsistency.

34On present assumptions, the oddness of our DA-affective and DA-doxastic states must also find a home in the semantics for fictional truth. Fictional sentences fall into five groupings.

  • 1. Sentences occurring expressly in a fictional text. (“Holmes waved our strange visitor into a chair”).

  • 2. Sentences having an implicit occurrence in the text. (“Holmes probably had quite a high I.Q.”).

  • 3. Sentences expressing the observations and speculations made by readers of a fictional text. (“Othello is not the main character in Othello”).

  • 4. Sentences reporting relations in which we ourselves stand to the objects and events reported by sentences of the first two classes. (“Agatha Christie admired Holmes more than any other detective”).

  • 5. Sentences registering cross-story comparisons. (“Holmes was certainly more intelligent than Li’l Abner.”)

  • 24  [Currie, 1990] classifies sentences of and about fiction as either fictive, metafictive or transfi (...)

35For ease of reference, call these sentences “explicit”, “implicit”, “intensional”, “external” and “cross-over”24.

36Let us not be the last to say that these distinctions are far from crisp. (For one thing, whatever it is in fine, the distinction between groups three and four doesn’t imply the emptiness of their intersection.) Jointly, the sentences in group one and a large proper subset of group two constitute the full story (FS). The sentences in groups three and four stand markedly apart. Group three sentences owe their reference to the prior provisions of groups one and two, but unlike group two sentences are not part of the story. Group four sentences state facts or apparent facts about the world, but do so under a rather singular constraint. Although they are not sentences of the story, they couldn’t be true except for them. For example, if, as is said, Agatha Christie admired Holmes more than any other detective, what makes for her admiration are facts about Christie and what makes for the putative object of it are facts about the Holmes stories that are fixed by the sentences of the FS, that is to say, by the stories’ explicit and implicit sentences. Similarly, although group five sentences aren’t true of any given story, they couldn’t be true without the truth of group one and group two sentences of different stories.

37We said just above that the FS of a story is constituted by its explicit sentences and a “large proper subset” of its group two sentences. What explains the qualification? This, too, is a tricky point whose full development will have to be dealt with elsewhere. For present purposes it suffices to make brief mention of an example. It is an example governed by a principle which we will simply state. It is the Anti-Closed World Assumption (ACWA), according to which, except (recursively) for contra-indicating sentences in group one and/or group two, stories inherit the world. Everything true in the actual world at the time of the story is true in the world of the story, save for those deviations wrought by the author. So, then, there is an intuitive difference, which ACWA captures, between what the story makes true of Holmes and his ilk and what’s true of Holmes’ world. What the story makes true of Holmes is that he lives in London; and what’s true of the world inhabited by Holmes is that London is thousands of miles east of Moose Jaw. All the things made true of Holmes and the other objects introduced by Doyle’s sayso constitute the full story. But many of the things true of Holmes’ world aren’t part of it. Holmes’ residency in London is part of the story. London’s easterliness of Moose Jaw is not.

  • 25  The distinction between the story and the world of the story is adumbrated in the distinction betw (...)

38Let us call this larger subset, that is, the union of the sentences constituting the FS and the sentences true of the world in which the whole story occurs, the maximal account (MA). In so doing, we appropriate a term introduced by Parsons ([Parsons, 1980]), whose intended extension is what we are calling the full story. In our borrowing, the term’s extension enlarges so as to include the full story and the world of the story.25 As we now see, our taxonomy is incomplete. A sixth category is required.

  • 6. Sentences true of the world of the story but not part of the story itself. In other words, sentences in the set MA-FS. (“London is thousands of miles east of Moose Jaw.”)

39Let us call these “world sentences”.

6. A glimpse of semantics

  • 26 Quine’s doctrine shows a reverse dependency. A theory is ontologically committed to an object x or (...)

40A semantics for a language or language-fragment L is customarily an ordered triple (R, T, I) of a theory of reference, a theory of truth and a theory of inference. In a great many approaches – compositional theories are a notable example – the elements of the triple are ordered by a partial dependency relation26. Thus the theory of truth for L depends in part on its theory of reference, and its theory of inference depends in part on its theory of truth. The dependency relation is transitive. In our treatment of it, a semantics of fiction is different. It, too, is representable by a triple ordered by a partial dependency relation, but its elements occur in reverse order, (I, T, R). Accordingly, the triple for the semantics of fiction is the converse of the triple of those of the mainstream, and the dependencies reverse accordingly. Inference precedes truth, and truth precedes reference.

41A further common feature of a semantics for L is that the meaning of sentences φ is furnished by the truth conditions for φ. When L is the language of fiction, sentential meanings are not the product of truth conditions. The sentences of, say, The Hound of the Baskervilles obtain their meanings in ways independent of the conditions which make them true. In what follows, we shall briefly discuss the basic form of our semantics, beginning with the theory of inference, followed by the theory of truth and the theory of reference.

42Inference. The whole job of a theory of inference for fiction is, to determine the membership of the text’s MA. The bulk of that assignment is already delivered by the Anti-Closed World Assumption. Not only is Holmes’ MA stacked with all the sentences true of the real world at the time of the Holmes’ stories, subject only to defections occasioned by explicit sentences, but it also provides as a corollary what we might call

The Type-Licence

Except where contra-indicated by an author’s explicit sentences, an MA contains all sentences of fictional objects of type K true of K-things in the real world.

43So, for example, like all humans, Holmes had a mother, a birthday and an oesophagus.

44Taken together, the Anti-Closed World Assumption and its Type License corollary yield two important and related facts about the fictional.

The determinacy principle

  • 27  To say that the story is determinate is to make the ontological claim that, for all ways, the worl (...)

Except where expressly provided-for by an author’s explicit sentences, a story’s world is as determinate as the actual world is. In other words, world-inheritance is determinacy-preserving27.

The completeness principle

  • 28 Full disclosure: The two authors of this paper are agreed that the Completeness Principle is a plau (...)

Except where expressly provided by an author’s explicit sentences, fictional objects are as complete as are objects of the same type in the actual world. In other words, type-inheritance is completeness-preserving28.

45Our two inheritance assumptions make for wholesale additions to the sentences that hold in an MA given its explicit sentences. But there are also inference-questions of more singular forms, such as whether the deductive consequences of MA-sentences can be added to the MA. More generally, are the rules of (say) classical first order logic rules of inference for fictional contexts. The short answer is that not all are. Consider again the fundamental idea of this paper. Our apparent relations with the putative objects of fiction are, upon analysis, DA-states causally triggered by the requisite sets of sentences. It is useful to compare the fictional case with the reportorial case. If we read in Morley’s The Life of Gladstone that Gladstone was born in 1809, the belief it induces admits of de re formulationas,

46x(x = Gladstone Belreader(born 1809, (x))).

47Similarly, when that belief is reported (“Of course, Gladstone was born in 1809”), the embedded reference is likewise expressibly de re:

48x ((refers (“Gladstone”, x)).

49It is conspicuously otherwise with DA-states. A reader’s admiration of, belief that, and reference to Holmes – still assuming Parmenides’ Rule and the Non-Existence Postulate – can only take de dicto form. Accordingly, whatever its details, there must be nothing in the semantic account of the sentences of the Holmes stories to contradict or obscure this fact. Accordingly, we have an important constraint on inferences in fictional contexts:


  • 29 Let us note that No Reification appears to solve the Asymmetry Problem of [Woods, 1969]. Consider a (...)

Picking up on the present point, a semantics for fiction that honours the assumptions of this paper must withhold fictional sentences from positions de re. In particular, it must suppress the existential generalization rule in fictional contexts29.

50Truth. The first task of a theory of truth for fiction is to define a truth predicate for the sentences of a story’s MA. There follows the business of explaining the truth of sentences in the remaining groupings in our taxonomy of fictional sentences. Here we concern ourselves with MA-truth, listing the main features of the account. Truth for non-MA sentences is also deferred to another occasion.

  • 30 The idea that “F” is a modal operator has been around for a long time. (See, e.g., [Woods, 1974, ch (...)

511. Ambiguity. Ordinary discourse about the fictional is systematically ambiguous. “Holmes waved our strange visitor into a chair” is ambiguous as between “A (Holmes waved our strange visitor into a chair)” and “F(Holmes waved our strange visitor into a chair)”, where “A” abbreviates some world indicating adjective such as “actually” or “in actuality”, and “F” is a dummy expression for adverbs indicating stories, hence abbreviates in the present case “in The Hound of the Baskervilles”. The ambiguity of the unmodified sentence is constituted by the falsity of the A-sentence and the truth of the F-sentence30. All sentences in a story’s MA carry the “F” operator tacitly.

522. In and out. The dominant task of a semantics for fiction is to furnish an implicit definition of “F”. This is done by providing truth conditions for sentences F(φ)¬whose scopes, φ, occur under tacit modification of the A-modifier. A further task for a semantics for F(φ)¬is to provide a semantics for its scope. Why should this be so? Because it strikes us as reasonable to suppose that, although the theory of reference is independent of the theory of truth, the double-aspectness of our believing or knowing that Holmes waved the strange visitor into a chair should be reflected in an underlying semantic double-aspectness. Accordingly, in disambiguating “Holmes waved our strange visitor into a chair”, our semantics must catch the following distinction: It must tell us what such sentences mean, and it must do this in ways that harmonize with the fact that they are true notwithstanding that there is nothing of which that they are true. This will require a semantics for F(φ)¬and a semantics for , its scope; hence a semantics for “out” and a semantics for “in”. The in-semantics will provide for the falsity (or untruth) of φ. The out-semantics will provide truth conditions for F(φ)¬What it does not do is give the meaning of F(φ)¬.Hence the semantics for F(φ)¬is of a radically different order from the semantics of φ. The measure of that radicalness is reflected in the difference between (3) and (4) immediately below.

  • 31 Since a story’s MA includes sentences that are not part of the story, why would we affix the F-oper (...)

533. Syntactizing F. The kernel idea is that F(φ)¬is true (that is, is A-true) if and only if φ occurs in the maximal account and in its occurrence there carries the “A”-modifier tacitly. Thus the truth of F()¬ consists of the syntactic fact that , as described, is a member of MA31.

  • 32 We note that there is nothing in our rejection of Frege’s Dismissal that precludes our selecting a (...)

544. Semanticizing φ. As before let be the scope of a true F(φ)¬. Then the semantics of φ should tell us what means. This is achieved by specifying truth conditions for φ, taking into account that φ will be A-true or A-false by virtue of how the world provides for the purported objects of φ that is to say, in the ordinary way or not at all. Again, whatever the details, the semantics will provide for the A-intended “Holmes waved our strange visitor into a chair” as it does for “Gladstone was disturbed about prostitution among the poor of London”32.

55Reference. DA states are essential to our account of affective and doxastic relations to the fictional. The Psychology First principle proposes that we would get a better account of our semantic relations to the fictional if it were constructed under the guidance of a working understanding of these psychological relations. Needless to say, for this to be so it is not necessary that the semantic account agree with the psychological account in all details essential to it. But if Psychology First is to be given any sway here, we should expect that the semantics of our reference to the fictional will pivot in some ineliminable way on double-aspectness. Has this expectation been met here?

56Consider a case. Barbara utters “Holmes lived in Baker Street.” In so saying, Barbara may feel herself as having referred to Holmes, even knowing full well that no one is Holmes. There are accounts galore according to which Barbara merely thought she was referring to Holmes even though she wasn’t actually referring to Holmes, that she wasn’t referring at all. But these aren’t the accounts we want. The account we want must provide grounds for saying that reference to the fictional can be likened to beliefs about the fictional, that is, that reference to the fictional can be made out to be DA-reference. What is it, then, that we think is going on in Barbara’s case?

57Barbara’s situation is something like this. Since “Holmes lived in Baker Street” is in The Hound of the BaskervillesFS, “‘Holmes’ refers to Holmes” is also in it. Indeed, it is part of the story that Holmes’ name is “Holmes”. Just as “Holmes lived in Baker Street” causally induces the DA-belief that Holmes lived in Baker Street, so in a similar way “Holmes lived in Baker Street” induces the DA-belief that ‘Holmes’ refers to Holmes. There is a sharper way of saying these things:

581. It is true that F (‘Holmes’ refers to Holmes).

592. Given that Barbara’s reading of The Hound of the Baskervilles was a reading with understanding, then Barbara knows that F (‘Holmes’ refers to Holmes).

603. Barbara knows that in uttering “Holmes lived in Baker Street”, she refers to what “F (‘Holmes’ refers to Holmes)” refers to.

614. She also knows, even so, that ‘Holmes’ refers to nothing.

62We said, just above, that we didn’t want to identify Barbara’s DA-reference to Holmes with the merely mistaken belief that she referred to Holmes. DA-reference is something more stand-alone. So let us make it official:


In uttering “Holmes lived in Baker Street”, Barbara DA-referred to Holmes if and only if Barbara knows that she’s referred to what the true sentence “F (‘Holmes’ refers to Holmes)” refers to, knowing as well that ‘Holmes’ refers to nothing.

63In a rough and ready way, in uttering “Holmes lived in Baker Street”, Barbara refers to what Watson refers to when F (“‘I worry about his melancholy’, I said sadly”) is true. This is as close to DA-reference as we are likely to get. The question is, is it close enough?

64Before quitting this section, this would be a good place to take note of the fact that in seeking for a DA-semantics for reference to the fictional, we do not – in the first instance, at least – seek for information that allows us to complete a biconditional in the form.

65“N” refers to N if and only if.

66One of the reasons for this is that where “N” is the name of a fictional entity, “N” refers to nothing. If the desired similarity with DA-sadness and DA-belief is to be achieved, what is needed is DA-reference. But in no straightforward sense is DA-referability definable for linguistic items such as names. As we saw, DA-sadness and DA-belief are states of mind carrying the appearance of a relation between an affected subject and the object of her affective state, and between a believing subject and some proposition. If referring is to be relevantly similar to these, it must in the first instance be a relevantly similar mental state. In other words, person-reference must precede linguistic reference. This is what we have here. It is, in the first instance, Barbara who DA-refers to Holmes and, for that to happen, Barbara must be in a requisitely double-aspected mental state. This raises an obvious question. Is DA-reference now definable for referring expressions. The answer is Yes. Consider a case:

Nominal reference

  • 33 We note in passing a similarity between our account of nominal reference and the quasi-pragmatic ap (...)

 ‘Holmes’ DA-refers to Holmes just in case it is true by default in the general case that users of ‘Holmes’ act on their knowledge that F (‘Holmes’ refers to Holmes)33.

7. Trouble from the past

67In pressing for a stand-alone theory for the sentences of fiction, we may have left the impression that the theory sketched here, if good for anything at all, is good only for the sentences of fiction. On reflection, it is easy to see that this is not in fact the case. Consider the following list of sentences, each consigned in at least one lexical place to the bin of reference-failure and the scrap-heap of untruth if Parmenides and Non-Existence are retained:

  • 1. Holmes solved the case of the speckled band.

  • 2. Zeus was worshipped by the ancient Greeks.

  • 3. Vulcan is the planet hypothesized by LaVerrier as explaining the orbital perturbations of Mercury.

  • 4. Caesar’s defeat of Vercingetorix was strategically brilliant.

  • 5. The present king of France is bald.

  • 34 Concerning (4), the Caesar-sentence, [Kripke, 1981] is a purported exception. The difference betwee (...)

68Of those that honour Parmenides and Non-Existence, mainline semantic theories – that is, the ones that also honour Frege’s Dismissal – tend to make their treatment of (5), “The present king of France is bald”, canonical for the others34. On the face of it, this is highly unsatisfactory. Part of our motivation in not treating (1) in that way is that in some non-dismissible sense we know who Holmes is. But, no more and no less, the same is true of Zeus, Vulcan and Caesar. Though none of them exists, and although none can be quantified over or made to stand in identity relations to anything, we know who or what they are. And since texts play the lead role in explaining (away) this knowledge of Holmes, it is tempting to suppose that a like treatment would work for the others – but not “for” the present king of France. If this idea could be made to stand up, a point of more general significance would present itself. It is that the standard semantic accounts of reference-failure do not work for the majority of types of it.

69Of course, Caesar, and all the others that used to be, are very bad news, not only for the present view of fiction but also for favoured general approaches to the semantics of reference. On August 5th the papers report the death of Gerald Cohen, Chichele Professor Emeritus at All Souls, Oxford. Professor Cohen had a rich and valuable life and was noted for his many accomplishments, whether as the pre-eminent exponent of neo-Marxism or for his riotous turns at stand-up comedy (we think the two are not unrelated.). Now if the view we have been forwarding here has legs enough to stand on, none of this is true. It is not true that Cohen died on August 5th. It is not true that he was born on April 14th, 1941. It is not true that he was called “Jerry”. It is not true that he was born in Montreal. Neither is it true that we know who Cohen was  not even his colleagues at All Souls know who he was  for there is no one of whom any Cohen-individuating fact is true. This is irritating. It carries all the marks of a theory spoiler. It is the counterexample from hell. Can we make it go away?

70It is not that rescue-packages haven’t been proferred – whether temporal or tense logics, tense-free quantification, or whatever else. Interesting as these systems assuredly are – though, it must be said, less for their conceptual elucidations than their mathematical virtuosities – they are all felled by any form of the Existential Generalization Rule – from Fa to infer xFx  that obeys Parmenides and Non-Existence. So it is proposed that Caesar defeated Vercingetorix. Of whom is this true? Where is the x such that, in that far off day, x defeated Vercingetorix? There is no such x but Caesar. But nothing there is is he. Suppose, then, that we stipulated that for the purposes of logic something qualifies as the value of a bound variable of quantification, and is eligible to be the bearer of a name, just in case it exists now or that it is now the case that existed in the past. This would let Caesar and Cohen in, but keep Sherlock, Zeus and Vulcan where they belong, namely, out.

71Plausible as this may seem, it won’t work. Counting against temporal indexing of existence is the question of Caesar’s existence now. If there is nothing now that used to be Caesar, then nothing now is that particular thing of which “used to exist” is true. True, there was something then of which “used to exist” is now true, but nothing of which “was something then of which ‘used to exist’ is now true” is now true. Parmenides and Non-Existence conspire to ill effect against both the long-since and recently departed. Parmenides says that the cost of quantification over objects lost to us through the passage of time is that they trigger the existence assumptions that quantification has been loaded with. And Non-Existence (adjusted to the departed) confronts the blunt fact that time has extinguished their existence. Jointly, at anytime at which Caesar is dead it can’t be the case that there is something that used to be Caesar.

  • 35 Full disclosure: On this point there is also a difference between the two of us, rather more on the (...)

72We don’t mind saying that we are disturbed by this result. The Parmenides-Non-Existence alliance does well for fiction – and for Zeus and Vulcan as well. But this same approach is wrenchingly counterintuitive for the no-longer-existing, making for a straightforward standard semantics for their existent interludes and a very different DA-semantics for thereafter. Our present view is that unless we can think of a principled way to rid ourselves of the Caesar problem, then it is back to the drawing board for the semantics of fiction. But, then, it will likewise be back to the drawing board for a whole class of mainstream approaches to reference35.

  • 36 Let us pause to give some brief indication of these problems. The accounts of Parsons and Jacquette (...)

73This brings us to a concluding observation about a point we made in section 1. It is possible to get a theory of fiction without the encumbrance of the Rule or the Postulate. [Parsons, 1980] and [Jacquette, 1996] are admirable examples in which both these assumptions are over-turned. But they are far from problem-free. They are not problem-free for the fictional. They are not problem-free for the dead36. As we said, at the end of the day it may well be the case that an intuitive theory of the fictional is simply too much to ask for.

8. Future work

  • 37 See footnote 13, regarding the Facts Before Intuitions Principle.

74At the outset of this paper, we took it as one of our working assumptions that a correct theory of fiction should be sensitive to the relevant disclosures of empirical theory37. Especially relevant to our purposes here are some recent (and pending) developments in neuroscience. Following [Schroeder and Matheson, 2006], we would consider the evidence produced by such research in developing our own account of DA-responses, much as they do in arguing for the role of distinct cognitive attitudes (DCAs). We think that the evidence they cite for DCAs is more suggestive than conclusive, since it gets applied to fiction via the mediating concept of imagination. We ourselves are of the view that there is nothing intrinsically imaginative about writing fiction or reading it with understanding. Even so, the neuroscientific evidence certainly suggests that the neurological correlates underlying such responses as sadness and DA-sadness are remarkably similar. We find this encouraging for a number of reasons. It is empirical basis for resisting both the illusion- and pretend-interpretations. And, it comports with the physical fact that we cry real tears, whether they are induced by real states or DA-ones. What is particularly interesting about DA-states is that not only do we feel anxious about Karla’s anticipated defection, notwithstanding our knowledge that nothing is Karla, but that we are also capable of being terribly pleased about the fact that LeCarré has crafted this episode so well. That is, whereas in the normal case, our anxiety blocks out the agreeable affect, in the fictional case unpleasant DA-anxiety does not block out real pleasure. We are optimists about these developments and are both hopeful and expectant that current and future neuroscientific research will be a profitable source of empirical data for theories of fiction.

  • 38 An early version of this paper was presented to the Symposium on Fiction at the Congress of the Can (...)

75Finally, we have put down our markers for a number of issues raised but not dealt with in this paper. They will have to be redeemed if our account here has any chance of developing into a comprehensive theory38.

Haut de page


Adler, Jonathan. 2002. Belief’s Own Ethics, Cambridge, MA: MIT Press.

Audi, Robert. 2004 .“The a priori authority of testimony”, Philosophical Issues, 14 (2004), 18-34.

Bach, Kent. 2002 . “Georgione was so called because of his name”, Philosophical Perspectives: Language and Mind, 16 (2002) 73-103.

Burge, T. 1974. “Truth and singular terms”, Noûs, 8 (1974), 309-325.

Coady, C. A. J. 1992. Testimony: A Philosophical Inquiry, Oxford: Oxford University Press.

Coplan, Amy. 2004. “Empathic engagement with narrative fictions”, Journal of Aesthetics and Art Criticism, 62 (2004), 333-342.

Currie, Gregory. 1990. The Nature of Fiction, Cambridge, Cambridge University Press.

Frege, Gottlob. 1952. “On sense and reference”. In Peter T. Geach and Max Black editors, Translations from the Writings of Gottlob Frege, Oxford: Blackwell, 56-78,.

Hjort, Mette and Sue Laver (eds.). 1997. Emotion and the Arts, New York, Oxford University Press.

Kripke, Saul. 1981. Naming and Necessity, Oxford, Blackwell.

Lackey, Jennifer. 2008. Learning from Words: Testimony as Source of Knowledge. Oxford: Oxford University Press.

Lewis, David. 1983. Philosophical Papers, volume 1, Oxford, Oxford University Press.

Meskin, Aaron, and Jonathan Weinberg. 2003. "Emotions, fiction, and cognitive architecture", British Journal of Aesthetics 43, (2003), 18-34.

Morley, John. 1903. The Life of Gladstone, London, MacMillan.

Rahman, Shahid et al. “Fiction, creation and their dynamics: First explorations”, forthcoming.

Saatela, Simo. 1994. “Fiction, make-believe and quasi emotions”, British Journal of Aesthetics, 34 (1994), 25-34.

Sainsbury, R. M. 2005. Reference Without Referents, New York, Oxford University Press.

Schroeder, Timothy and Carl Matheson. 2006. “Imagination and emotion”. In Shaun Nichols (ed.), The Architecture of the Imagination: New Essays on Pretense, Possibility and Fiction, New York, Oxford University Press, 19-39.

Thomasson Amie, 1998. Fiction and Metaphysics, Cambridge, Cambridge University Press.

van Inwagen, Peter. 1977. “Creatures of fiction”, American Philosophical Quarterly, 14 (1977) 299-308.

Walton, Kendall. 1990. Mimesis as Make-Believe: On the Foundations of the Representational Arts, Cambridge MA, Harvard University Press.

Woods, John. 2009. The Logic of Fiction: A Philosophical Sounding of Deviant Logic, The Hague and Paris: Mouton, 1974. Second edition, with a new Foreword by Nicholas Griffin, London, College Publications.

Woods, John. 2006. “Fictions and their logic”. In Dale Jacquette (ed.), Philosophy of Logic, volume 5 of Dov M. Gabbay, Paul Thagard and John Woods (eds.), The Handbook of the Philosophy of Science, Amsterdam, North-Holland, 835-900.

Woods, John. 2009. “Preface” to Fiction: logiques, langages, mondes, Brian Hill et al. (eds.). Forthcoming in the series Les Cahiers de Logique et d’Epistemologie from College Publications, London.

Woods, John. 2010a. Sherlock’s Member: New Perspectives on the Semantics of Fiction.

Woods, John. 2010b. Seductions and Shortcuts: Error in the Cognitive Economy.

Woods, John (ed.). 2010c. Fictions and Models: New Essays. Munich: Philosophia Verlag.

Yanal, R. J. 1999. Paradoxes of Emotion and Fiction, Philadelphia, Pennsylvania State University Press.

Haut de page


1 Not to overlook the dead of Guernica. See below, section 7.

2  [Frege, 2001, p. 202-203; Sainsbury, 2005, p. 202-209]

3 For recent writings to the same effect, see Woods (ed.) 2010c, Fictions and Models: New Essays, Munich: Philosophia Verlag, with contributions by Robert Howell, Amie Thomasson, Mark Balaguer, Otávio Bueno, Mauricio Suarez, Roman Frigg, Jody Azzouni, Alexis Burgess, John Woods and Alirio Rosales and Giovanni Tuzet.

4 Not even the relation of denying their existence.

5  Standard semantic theories take the position that for a name to refer, it must have a referent (that is, there is something in the world which the name picks out). Since there is nothing in the world that is the referent of a fictional name, it must fail to refer. Similarly, for a sentence to be true, there must be something in the world of which they are true. Accordingly sentence embedding fictional names can’t be true and therefore are either false or meaningless.

6 The Postulate says that all fictional objects are non-existent. But “all” is a quantifier, and the Rule says that quantifiers are existentially loaded.

7 See footnote 35 at the end of the paper for some indication of the problems we see in alternative approaches.

8 In other contexts, the present admonition is rather small beer, indeed hardly worth mentioning. Think here of the Routley-star semantics for first-degree entailment and post-paradox set theory. But, historically, theorists of fiction have laid considerably greater emphasis on the role of the relevant linguistic intuitions.

9 A brief historic note: Parmenides’ Rule is very old, possibly originating with Parmenides himself. Non-existence, formulated as a claimabout fictional objects, is much newer (it undergirds pretense theories, for example.) But more general versions have been held or intimated also since antiquity (as witness materialism). Frege’s Dismissal, while given an original formulation by Frege, is also implicit in the uniformities reflected in the treatment of negative existentials from earliest times.

10 Among recent contributions, see [Sainsbury, 2005].

11 Among recent contributions, see [Coplan, 2004; Hjort and Laver eds., 1997; Meskin and Weinberg, 2003; Yanal, 1999]

12 We are not unaware of the historical provenance of the Psychology First Principle. To the best of our knowledge, ours is the first express statement of it; but its influence certainly pre-dates that formulation. All four of our assumptions – the Rule, the Postulate, the Dismissal and Psychology First – are explicitly or implicitly present in the pretense account, as well as explicitly present in our own. But, as will become apparent, our account is non-trivially at odds with the pretense theory – indeed incompatible with it. This suggests – and we agree – that sound though those assumptions may be, they significantly underdetermine what it takes to produce a tenable (or superior) theory of fiction.

13 Another way of making the same point is that, to the extent that its subject matter permits it, we prefer a naturalized semantics for fiction; that is, a theory that is sensitive to relevant disclosures of observation and empirical theory. See below, section 8.

14 See here [Woods, 2006].

15 Here, and in what follows, we adapt with revisions from passages in [Woods, 2009] and [Woods, 2010a].

16 The exertions of dialetheic logicians notwithstanding.

17  See section 18 below for predictions about the sort of empirical date that will prove useful in resolving this issue.

18   [Walton, 1990; Saatela, 1994; Yanal, 1999]

19 They are, as David Lewis puts it, “told as known fact” ([Lewis, 1983, p. 270]).

20  Against this it might be objected that we also cry at Juliet’s despair in Prokofiev’s ballet. We do so knowing that the performance is play-acting. Yes, so it is. But it hardly follows that, in our sadness over Juliet’s wretchedness, we ourselves are playing at being sad.

21 [Audi, 2004; Coady, 1992; Adler, 2002; Lackey, 2008; Woods, 2010b]  For present purposes, it is enough that everyone agrees that we come to believe things as a result of being told them. However, central to the literature is a debate about whether testimony is a basic justification for knowledge claims. Reductionists argue that it is not and anti-reductionists argue that it is. However, concerning the particular tie between fiction and telling, it would appear that one could not be a reductionist with respect to apparent telling, since there can be no further justification on which to ground our claims. Consider: “Holmes lived in Baker Street?” “Yes”. “Who says?” “Well, it’s right there in Doyle’s story.” “Ah-ah ! Why should we believe Doyle?”

22 We don’t want to minimize the complexity of the causal mixtures thesis. There are cases not well-caught by the present version of it. We say that DA-sadness is produced by causal relations between us and the text in ways that resemble our real sadness at reading an account of some real tragedy. But consider the case of the man who slipped on a banana peel and knocked himself silly. If the account were a piece of history – say it had happened to J.F.K.  the causal response would have been one of sympathy and regret. “How awful !”, we would say. But if the text conveys a farce, the normal responses would be glee and mirth. Our thesis needs to be adjusted for such cases. Intuitively, when we laugh our heads off, we are responding to the story’s art. Perhaps we could accommodate this fact in our requirement that responses to fiction arise from having read the requisite texts “with understanding.” We are grateful to Christopher Mole for conversation on this point.

23   In the words of [Schroeder and Matheson, 2006].

24  [Currie, 1990] classifies sentences of and about fiction as either fictive, metafictive or transfictive. Not only are these last two names not especially illuminating, but Currie’s classification appears to miss two further nuanced details. On Currie’s account, it is unclear how we are to classify implicit sentences. While Currie states that metafictive statements are about the fiction, he also states that fictive sentences are just those uttered by the author in the act of producing a work. The category of implicit sentences is one that does not seem to fit comfortably into either class. Similarly, it is unclear how Currie would have us classify (at least some) external sentences. He says that transfictive statements compare fictional characters with both the characters of other fictions (cross-over), as well as with actual individuals. However, as we have seen, not all of the relations that we can express using external sentences are comparisons. So while some external sentences will be counted as transfictive, many will not be.

25  The distinction between the story and the world of the story is adumbrated in the distinction between what’s true of the history of, say, the Battle of Britain and what’s true of the world at the time of its occurrence. The RAF’s technical superiority to the Luftwaffe is part of that history. The fact that Evan Smith’s mum lived in Moose Jaw is part of the world but not part of the Battle of Britain.

26 Quine’s doctrine shows a reverse dependency. A theory is ontologically committed to an object x or to objects of kind K iff that object or those objects are values of variables bound by quantifiers in the theory’s true sentences. Thus the reference associated with ontological commitment is parasitic on truth.

27  To say that the story is determinate is to make the ontological claim that, for all ways, the world is one way or another. For example, it must be that Holmes had a mole on his back or didn’t. Against this, ontological indeterminatists hold that if the author doesn’t tell us whether Holmes has a mole on his back, then he neither has it nor lacks it. The Determinacy Principle does not commit us to the view that readers can always know all the facts about Holmes. Just as there are many facts about the actual world to which we lack epistemic access, so too do we lack it to some of the facts about the story’s world.

28 Full disclosure: The two authors of this paper are agreed that the Completeness Principle is a plausible thing to gather from the Anti-Closed World Assumption. But one of them (JI) is a good deal less sure than the other (JW) that the principle is actually true.

29 Let us note that No Reification appears to solve the Asymmetry Problem of [Woods, 1969]. Consider a case. Doyle might have made it the case that Holmes once visited Prime Minister Gladstone and had tea with him.  Suppose he did so. Then it is true of Holmes that he had tea with Gladstone, but it is not true of Gladstone that he had tea with Holmes. This is problematic. Having-tea-with is a symmetrical relation. Given the meaning of “had tea with” it is a necessary truth that if x had tea with y then y had tea with x. So what happened on the day of that visit could not possibly have happened. If No Reification is the right way to go, we have a solution to the asymmetry problem. The sentence “It is not true of Gladstone that he had tea with Holmes”, has “Gladstone” in de re position, and “It is true of Holmes that he had tea with Gladstone” does the same for “Holmes”. The former sentence is not disallowed by the No Reification Rule, but the latter is ruled out by it entirely. So it is not true of Holmes that he had tea with Gladstone. See the section on truth immediately below.

30 The idea that “F” is a modal operator has been around for a long time. (See, e.g., [Woods, 1974, chapter 5].) We ourselves are not now so sure. Consider the standard semantics for modalities “necessarily” and “possibly”. Then the necessity of a sentence φ is its truth in the universal quantification of an abstract set-theoretic object, and the possibility of a sentence φ is its truth in the existential quantification of that same structure. It does not strike us that the semantics of  F(φ)¬is anything like these cases. The truth of “In The Hound of the Baskervilles, Holmes waved the strange visitor into the chair” is not a matter of the truth anywhere of “Holmes waved the strange visitor into a chair”, hence is not a matter of that sentence’s truth in a quantification of any kind of set-theoretic structure. See just below.

31 Since a story’s MA includes sentences that are not part of the story, why would we affix the F-operator to them? The answer is that since it is part of the story that Holmes is genidentical to everything else in the physical world, so everything true of Holmes’ world will be linked to Holmes. That is reason enough to permit the F-operator to drift across the intuitive divide between FS and MA-FS, that is, to give the operator homogenous application in MA.

32 We note that there is nothing in our rejection of Frege’s Dismissal that precludes our selecting a standard semantics for scopes. In particular, it is not ruled out that with respect to the of  ⌐F(φ)¬an acceptable semantics could turn out to be, say, a negative free logic (NFL) in the manner of [Burge, 1974]. For present purposes we leave this question open. But what is not in the least open is whether a NFL will work for F()¬. Not a chance. Further explorations of the free logic option for fiction may be found in work in progress by Shahid Rahman and his colleagues “Fiction, Creation and their Dynamics: First explorations”. We regret the lack of space to review that contribution here; but see [Woods, 2010a].

33 We note in passing a similarity between our account of nominal reference and the quasi-pragmatic approach of [Bach, 2002].

34 Concerning (4), the Caesar-sentence, [Kripke, 1981] is a purported exception. The difference between Caesar and Holmes, Zeus and Vulcan is that in the first instance, but in none of the others, we ourselves are genidentical to the act of bestowing on Caesar Caesar’s name, which in turn is part of a causalized version of Russell’s acquaintance condition on denotation ([Russell, 1905]). Indeed we are members of that chain, but only if the extant texts are accurate. This turns out to be a vital constraint. Suppose God were aware that all those texts are clever frauds. Then according to Kripke, we never referred to Caesar; indeed it is always open that when we think we refer we really don’t. Since texts play an indispensable role in preserving the causal-historical account there is a certain economy in shifting the emphasis from the baptisms the texts attest to the texts themselves. But – see the text just above – there is another problem for the genidentity account. If Parmenides’ Rule is true and if death implies non-existence, then it is not true that we are genidentical with the event of Caesar’s baptism.

35 Full disclosure: On this point there is also a difference between the two of us, rather more on the optative side than the alethic. We both are inclined to think that the Caesar problem is decisive against a Rule + Postulate treatment of the dead. JI is of the view that, if so, this does no damage to the Rule + Postulate treatment of fiction. JW is not so sure.

36 Let us pause to give some brief indication of these problems. The accounts of Parsons and Jacquette are tied to a Meinongean theory of objects. On this view, objects are wholly constituted by their properties and, furthermore, existence is a property. Lying at the heart of these claims is a distinction between nuclear and extranuclear properties. For present purposes a little circularity won’t hurt. Nuclear properties are object-constituting. Extranuclear properties are true of their respective objects but not constitutive of them. Red-breastedness is a nuclear property of Robin Red-Breast. Being thought of by Terence Parsons is an extranuclear property of it. Consider now the case of Caesar. Prior to his death, Caesar had the property of existence, and afterwards it was a property he lacked. Since Caesar alive has a property that Caesar dead lacks, Caesar alive cannot be the same object as Caesar dead. Meinongeans attempt to block this objection by insisting that existence, while a property, is not a nuclear or object-constituting property. No one – not even they – would deny that existence is a metaphysically momentous property. So perhaps we will be forgiven our scepticism when we say that the decision to “go nuclear” on existence has nothing to recommend it but tendentiousness. Let us turn now to the fictional. Since any combination of nuclear properties constitutes the object having them, then all the properties ascribed to Holmes in the full story constitute Holmes independently of their ascription there. This means that Sherlock Holmes pre-dates Doyle’s “creation” of him, and that the real Holmes stands to the Holmes stories as the real London stands to them. Not only is this deeply counterintuitive, but it would seem that, upon reflection, the Holmes who pre-dates Doyle’s Holmes cannot be the same Holmes as he. Why? Because it is object-constituting of Doyle’s Holmes that he was created by Doyle. Here, again, Meinongeans would plead the extranuclearity of the property of being constituted by Doyle. But, as before, our own view is that saying so is too ad hoc by half. So there are problems with theories that reject Parmenides’ Rule and the Non-Existence Postulate. Other theories that reject these assumptions include theoretical entity accounts in the manner of [van Inwagen, 1977] and artefactual accounts in the manner of [Thomassen, 1998]. These, too, call for careful attention, but not here.

37 See footnote 13, regarding the Facts Before Intuitions Principle.

38 An early version of this paper was presented to the Symposium on Fiction at the Congress of the Canadian Philosophical Association, Carleton University, in June, 2009. We are indebted to our co-symposiats Peter Alward, and Sarah Hoffman for helpful criticism, as well as to those in attendance, of whom the following are known to us by name: Robert Stainton and Gary Foster. A considerably altered version was presented to UBC’s excellent Fiction Group, in August, 2009, whose participants provided welcome stimulus for improvement. These are Christopher Mole, Bryan Renne, Alirio Rosales and Jim Westin. Subsequent comments by Derek Matravers and Jaakko Hintikka were also very helpful. We also thank this journal’s anonymous reviewer for pressing us to be clearer about some of the paper’s main points.

Haut de page

Pour citer cet article

Référence électronique

John Woods et Jillian Isenberg, « Psychologizing the Semantics of Fiction », Methodos [En ligne], 10 | 2010, mis en ligne le 07 avril 2010, consulté le 25 mars 2017. URL : ; DOI : 10.4000/methodos.2387

Haut de page


John Woods

Department of Philosophy, University of British Columbia, Group on Logic and Computation, King’s College London

Jillian Isenberg

Department of Philosophy, University of British Columbia

Haut de page

Droits d’auteur

Licence Creative Commons
Les contenus de la revue Methodos sont mis à disposition selon les termes de la Licence Creative Commons Attribution - Pas d'Utilisation Commerciale - Pas de Modification 4.0 International.

Haut de page
  • Logo DOAJ - Directory of Open Access Journals
  • Logo UMR Savoirs, Textes, Langage
  • Logo CNRS - INSHS
  • Logo Université de Lille