Philosophical Studies

, Volume 153, Issue 2, pp 183–211

The great beetle debate: a study in imagining with names


    • Department of Philosophy, Heythrop CollegeUniversity of London

DOI: 10.1007/s11098-009-9485-4

Cite this article as:
Friend, S. Philos Stud (2011) 153: 183. doi:10.1007/s11098-009-9485-4


Statements about fictional characters, such as “Gregor Samsa has been changed into a beetle,” pose the problem of how we can say something true (or false) using empty names. I propose an original solution to this problem that construes such utterances as reports of the “prescriptions to imagine” generated by works of fiction. In particular, I argue that we should construe these utterances as specifying, not what we are supposed to imagine—the propositional object of the imagining—but how we are supposed to imagine. Most other theories of thought and discourse about fictional characters either fail to capture the intentionality of our imaginings, or else obscure the differences between imaginings directed toward fictional characters and those directed toward real individuals. I argue that once we have an account of prescriptions to imagine about real individuals, we can adapt the same framework to specify the contents of prescriptions to imagine about fictional characters, and thereby to account for the truth (or falsity) of statements about fictional characters.


Fictional referenceEmpty namesImaginationPrescriptions to imagineKendall WaltonJohn PerryFictional charactersTruth in fictionNotion networks
In his essay on Kafka’s The Metamorphosis, Vladimir Nabokov writes:

Scene I: Gregor wakes up. He is alone. He has already been changed into a beetle, but his human impressions still mingle with his new insect instincts. (Nabokov 1980, p. 260)

This description, which concludes an argument for Nabokov’s interpretation, is controversial: the text says only that Gregor Samsa has been turned into a gigantic “vermin” (Ungeziefer). Nabokov agrees with a number of critics that Gregor has become an insect, but the text never specifies which kind. Here is part of Nabokov’s argument that Gregor becomes a beetle:

Next question: what insect? Commentators say cockroach, which of course does not make sense. A cockroach is a vermin that is flat in shape with large legs, and Gregor is anything but flat: he is convex on both sides, belly and back, and his legs are small. He approaches a cockroach in only one respect: his coloration is brown. That is all. Apart from this he has a tremendous convex belly divided into segments and a hard rounded back suggestive of wing cases. In beetles these cases conceal flimsy little wings that can be expanded and then may carry the beetle for miles and miles in a blundering flight. Curiously enough, Gregor the beetle never found out that he had wings under the hard covering of his back. (Nabokov 1980, pp. 259–260)

These passages raise certain familiar puzzles. Nabokov appears to be talking about an individual, Gregor Samsa, of whom he asserts, in opposition to other commentators, that Gregor has been changed into a beetle. Yet there has never been such an individual, and Nabokov knows that. So what would make his claims right or wrong, true or false?

In this paper I argue that if we construe utterances reporting fictive content, such as
  1. (1)

    Gregor Samsa has been changed into a beetle,


as statements of how a work of fiction prescribes imagining, we can explain how such statements can be true even though they contain empty (non-referring) names.1 In particular, I argue that we should construe utterances like (1) as specifying, not what we are supposed to imagine—the propositional object of the imagining—but how we are supposed to imagine. Certain ways of imagining count as imagining that Gregor changed into a beetle, while other ways of imagining, such as imagining that Gregor changed into a praying mantis or that Hamlet changed into a beetle, do not satisfy the prescription. In particular, certain ways of imagining count as imagining about Gregor, while others do not (where the term about does not indicate ontological commitment). What is required is a theory that captures this distinction: that is, a theory that provides a framework for specifying appropriate ways of imagining.

After explaining the problem posed by (1) in more detail, I criticize standard theories of fictional discourse, arguing that most fail to capture the intentionality of our thoughts and utterances about fictional characters. In the rest of the paper, I argue that once we have an account of the ways in which fictions prescribe imaginings about real individuals, we can use the same framework to specify the contents of prescriptions to imagine about fictional characters, and thus to explain what is going on in Nabokov’s argument.

Before turning to this argument, a clarification is in order. In what follows I treat utterances such as (1) as potentially true or false and therefore as amenable to truth-conditional analysis. According to some views, however, statements such as (1) should be understood as mere “pseudo-assertions,” which can be correct or incorrect but not true or false (see especially Eagle 2007, p. 138ff). I do not take a stand on this issue here. Those who take the relevant utterances to be pseudo-assertions can replace my talk of truth and truth conditions with talk of correctness and correctness conditions.2 My proposal for specifying the contents of prescriptions to imagine about fictional characters will be unaffected by this change.

1 Prescriptions to imagine

It might seem just obvious that statements like (1) can be true, if we take them to be implicitly elliptical for statements containing a fictionality operator such as ‘According to the fiction’. Suppose that (1) should be understood this way. Even so, the complacency is unjustified, especially when we consider the most popular account of proper names. According to referentialism, utterances containing proper names express singular propositions and thus have singular truth conditions: the semantic contribution of a proper name to the proposition is the individual designated, rather than some mode of presentation of that individual. Because an empty name designates no individual, it would appear to make no semantic contribution, so that utterances containing the name either fail to express any propositions, or express incomplete (“gappy”) propositions. On this view, the complex statement formed by prefixing (1) with a fictionality operator cannot express a complete proposition if (1) does not express a complete proposition. If it does not express a complete proposition, it seems that it cannot be judged either true or false.3

That empty names pose a problem for referentialism is a standard objection to the theory; but statements like (1) pose a problem for every account that has been proposed to handle them. The reason is simple: even supposing (1) to be implicitly prefixed, we require an account of what, exactly, is the case according to the fiction, and no alternative account answers this question satisfactorily (or so I will argue). Given independent reasons to accept referentialism about names, my proposal will be consistent with a version of the theory.

Let us now consider the prefix, ‘According to the fiction’. Intuitively, for any sentence S, ‘According to the fiction’S tells us that the fiction says what S says. But this is too underspecified. For example, neither S itself, nor any sentences that logically entail S, need occur in the text for the fiction to “say” what S says in the relevant sense. Further, if I report that according to the Encyclopaedia Britannica, Vulcan does not exist, I have told you something about what the encyclopedia says is true; but “According to the story, Gregor Samsa has been changed into a beetle” is a report, not of what anyone thinks is the case in the actual world, but instead of what is the case in the fictional world—the world of Kafka’s story. Saying this does not get us very far, however, without an explanation of what a fictional world is. For a philosopher the temptation is to identify fictional worlds with sets of metaphysically possible worlds, as Lewis (1983) did. But this proposal faces serious difficulties. Consider a story according to which Houdini squared the circle. The apparently true report, “According to the story, Houdini squared the circle,” would, on one construal of this proposal, turn out to be false, since it is not true in any possible world (Howell 1979). On a different construal, anything whatever would be vacuously true in this fiction (Lewis 1983), including “According to the story, Houdini did not square the circle.” As a result, every contradictory story would generate exactly the same fictional truths; but surely different fictions generate distinct fictional worlds. Although there are replies the Lewisian can make to these worries,4 a more promising approach dispenses with the appeal to fictional or possible worlds altogether.

In my view, the best account of what it means for something to be the case “according to the fiction” is to be found in Walton’s concept of prescriptions to imagine. The basic idea—which can be understood independently of his invocation of pretense or games of make-believe—is that works of fiction prescribe imaginings about their content: we are supposed to imagine, for example, that certain people did certain things.5 When we imagine that such-and-such is the case according to a work of fiction, the way we imagine is guided, even constrained, by the interpretation of the work. Of course, I am not prevented from imagining anything I want; if I so desire, I may imagine that Raskolnikov has wings. There is a perfectly clear sense, however, in which this would be an unauthorized response to Crime and Punishment. And the sense in which this is an unauthorized response is reflected in the assessment that it is not true-in-the-fiction that Raskolnikov has wings, and therefore that “Raskolnikov has wings” is a false claim about Dostoevsky’s novel.

Given this background, utterances such as (1) can be interpreted in two ways. (All examples are utterances—concrete events in particular contexts—rather than sentences.) On the first interpretation, we take the utterance to be spoken from a perspective “external” to the fiction, as involving an elided ‘according to the fiction’ operator. Alternatively, the utterance could be taken as a natural continuation of our imaginative engagement with the fiction, from an “internal” perspective. On Walton’s analysis, for example, such unprefixed statements constitute participation in authorized games of make-believe, indicating that it is appropriate to speak or pretend in the relevant manner (Walton 1990, pp. 397–399). The latter approach has a number of advantages over the operator view (see, e.g., Eagle 2007), including the fact that it can be extended to other types of fictional discourse. The account I offer is compatible with either way of construing (1).6

The central claim of this account is that an utterance ‘S’, occurring in the context of a discussion of a fiction F, is true if and only if F prescribes imagining that S. This will be either because ‘S’ is to be understood as meaning, or getting across, something like ‘According to F, S’, or because it is to be understood as a continuation of the imaginative engagement authorized by F. Therefore (1), occurring in the context of Nabokov’s discussion of The Metamorphosis, is true if and only if Kafka’s story prescribes imagining that Gregor has been changed into a beetle.

What does it mean to say that the story prescribes this sort of imagining? I have suggested that the answer to this question would specify the way of imagining that counts as imagining that Gregor changed into a beetle. But it seems that we cannot specify this by appeal to a propositional content to be imagined, since there is no (complete) proposition such that Kafka’s story prescribes imagining it. It is for this reason that Walton, as a referentialist, does not appeal to prescriptions to imagine in accounting for empty names. Indeed the standard approach to empty names is to reject referentialism and identify a general proposition that is true-in-the-fiction: that is, on the present interpretation, a proposition to be imagined. A brief overview of the various proposals in the literature will show that this approach is misguided.

2 Standard accounts

Consider the following debate, based on Nabokov’s essay:

Nabokov says,
  1. (1)

    Gregor Samsa has been changed into a beetle.

To which another critic, Smith, replies,
  1. (2)

    No, Gregor Samsa has been changed into a cockroach.


Let’s assume that Nabokov is right about his interpretation of the story, and thus about what the story prescribes imagining.7 Then Nabokov has said something true while Smith has said something false. More importantly, Nabokov and Smith have disagreed: they are making incompatible claims about the same thing, even though there is no thing they are talking about. Nabokov claims that Kafka’s story prescribes imagining that Gregor has been changed into one kind of insect, Smith that Gregor—the very same character—has been changed into a different kind of insect. We want a theory to explain the possibility of intersubjective identification: the phenomenon of thinking and talking about the same thing, even when there is no such thing. Notice that such identification goes through even if imaginers do not recognize it. I might take my imaginings “about Odysseus” to be directed at a different character than my imaginings “about Ulysses,” but I am wrong.8 Although the names ‘Odysseus’ and ‘Ulysses’ do not co-refer (because they do not refer at all), they co-identify.

Several analyses of (1) and (2) fail to explain these features of imagining. A sententialist analysis, according to which (1) is true so long as The Metamorphosis prescribes imagining that the sentence ‘Gregor Samsa has been changed into a beetle’ is true, faces numerous difficulties. It implausibly takes our imagining to be about a sentence, especially problematic if the sentence is in English while the story was in German. It also fails to distinguish between imagining that a sentence is true and imagining that what the sentence says is true. Most importantly, the analysis does not have sufficient internal structure to explain the sense in which (1) and (2) are both about the same thing. A different proposal focuses on the proper name, and says that (1) is true so long as we are supposed to imagine that there is an individual named ‘Gregor Samsa’ who has been changed into a beetle. But this approach will not do, since the same fictional character can have different names (‘Odysseus’ and ‘Ulysses’) and distinct fictional characters can have the same name (Austen’s Emma and Flaubert’s Emma). In short, meta-linguistic accounts fail to capture basic features of our imaginings.

Descriptivist and quantificational theories, which interpret imagining that Gregor Samsa has been changed into a beetle as, roughly, imagining that there is a bearer of certain properties that has been changed into a beetle, are flawed for similar reasons.9 These accounts run into trouble with counter-fictional imagining. For instance, I might imagine what the Samsa family’s life would have been like had Gregor never changed into a vermin. Even though I imagine contrary to what Kafka’s story prescribes—thinking of Gregor in ways contrary to the fictional descriptions—I continue to imagine about the same character. And that is how it should be: considering the question of what would happen if Gregor were not transformed is central to understanding Kafka’s story; the ending strongly implies that in such a circumstance his parents and sister would not have learned how to rely on themselves. Without engaging in this kind of counter-fictional imagining, we would be unable to comprehend the significance of events in characters’ lives. We also would not appreciate fictions that import fictional characters from other works. In Marc Estrin’s novel Insect Dreams, for example, Gregor Samsa survives his transformation and becomes the main attraction at a circus. Failure to recognize that this is the same character as the one in Kafka’s story would be failure to understand the point of Estrin’s novel.

Co-identification and counter-fictional imagining are two features that indicate the intentionality of imaginings about fictional characters: the sense in which they are object-directed, even though there is no object they are about. Theories on which fictions prescribe that we imagine non-singular propositions fail to capture the intentionality of prescribed imaginings. So let us consider alternative approaches.

Realists claim that fictional characters are real, though nonexistent or abstract, entities, and thereby seem to account for the intentionality of our imaginings straightforwardly. The purported advantage of this approach is that we can give exactly the same kind of explanation of how we are supposed to imagine when a work prescribes imaginings about fictional characters like Gregor, as when it prescribes imaginings about (uncontroversially) real individuals like London or Napoleon. The realist could specify the content of (1) with a singular proposition. I do not have the space to argue against the various forms of realism, which I believe are untenable.10 But it is worth noting that realism of this sort is not easily restricted to imagining, since other propositional attitudes, such as belief, exhibit exactly the same features of intentionality. Suppose that in Michigan, Daniel believes (as he would put it) that “Santa Claus is coming,” while in Surrey, Kaitlin believes (as she would put it) that “Father Christmas is coming.” There is some sense in which Daniel and Kaitlin believe the same thing, even if there is no individual their beliefs are about (cf. Everett 2000). Thus their beliefs support de re-style reports, like “Kaitlin believes that Santa Claus is coming, but she doesn’t know he’s called ‘Santa Claus’.” Realists would explain these facts by saying that the names ‘Santa Claus’ and ‘Father Christmas’ refer to the same real individual. Surely, however, it is more plausible to say that there is no Santa Claus, than to postulate entities any time there is intersubjective identification.

Referentialists who reject realism assume either that utterances like (1) and (2) express no propositions, or that they express incomplete propositions. Walton (1990) assumes the former, and aims to account for the intentionality of our imaginings by appeal to kinds of pretense: some kinds of pretense are “Gregor-directed” while others are not. On Walton’s view, in using the name ‘Gregor Samsa’, Nabokov and Smith pretend to refer to an individual and pretend to assert something about him. The question is: what makes their shared pretense a pretense about Gregor? For utterances that contain referring names, Walton specifies the kind of pretense by specifying the proposition fictionally asserted by the utterance. But the pretense exemplified by (1) and (2) cannot be specified this way, because on Walton’s referentialist view, there are no propositions such that Nabokov and Smith fictionally assert them (see esp. 390ff.). Walton notes that we cannot appeal to the name ‘Gregor Samsa’, since Nabokov could have made the same fictional assertion by uttering “Greta’s brother has been changed into a beetle.” Because the name is insufficient to individuate the pretense, Walton concludes that there is no way to specify it apart from pointing to instances related in certain ways to the fiction. This conclusion is not satisfactory, though, especially as applied to (1) and (2), because it looks as though Nabokov and Smith each claim that a different kind of pretense is appropriate to The Metamorphosis. What makes it the case that they are in some other sense engaged in the same kind of pretense, namely, the Gregor-directed kind? To ask this question is simply to ask how Walton accounts for the intentionality of our imaginings, and at this point we have no answer.11 Other pretense accounts face the same problem. Invocation of a “Meinongian pretense” (Walton 1990; Recanati 2000; Everett, ms), according to which there are both existent and nonexistent objects, still requires a way to determine which pretend-references are about which nonexistent objects.

A number of referentialists have rejected the view that utterances containing empty names express no propositions, and instead argue that they express gappy propositions: propositions structured like singular propositions, but with a gap where an object ought to go.12 Adopting the standard representation, the debate between Nabokov and Smith looks like this:
  • (N)  <_____, having-been-changed-into-a-beetle>

  • (S)  <_____, having-been-changed-into-a-cockroach>

The gappy proposition view has the virtue of representing the structural similarity between imaginings about real things and imaginings about fictional characters without taking the latter to be real. Yet an obvious drawback to this approach is that it cannot, by itself, distinguish between the gappy proposition expressed by (1) and the one expressed by the utterance,
  1. (3)

    Hamlet has been changed into a beetle.


Since (1) is, by hypothesis, true, whereas (3) is false, any theory must be able to distinguish between them.

Advocates of gappy propositions recognize this requirement. They maintain that although the semantic content of (1) and (3) is the same, namely (N), they can be distinguished by their different pragmatic implications, which generally appeal to the different descriptions associated with the names ‘Gregor Samsa’ and ‘Hamlet’. Now, while the distinction between what is semantically expressed and what is pragmatically communicated may matter for some purposes, it is unimportant for the present inquiry. Our goal is to specify the way in which The Metamorphosis prescribes imagining, and it is plain that (1) and (3) indicate different sorts of imagining. If (1) and (3) express exactly the same semantic content—in the sense that referentialists usually intend—semantic content in this sense does not suffice to specify how we are supposed to imagine. The distinction between kinds of imagining must be explained in a different way.

So let us take a look at the pragmatic portion of the gappy proposition view. According to some proponents of the view, the two utterances communicate, via various pragmatic processes, distinct non-singular propositions related to the gappy proposition, propositions about properties rather than about objects.13 How does this approach explain prescriptions to imagine? Since imagining the gappy proposition by itself cannot satisfy the prescription, either readers are supposed to imagine both the gappy proposition and the related non-singular proposition; or else they are just supposed to imagine the non-singular proposition. I am not sure how to make sense of the former option, since it seems implausible that a single episode of imagining has two distinct propositional objects. The latter option suggests that works of fiction, in expressing the gappy propositions, trigger pragmatic processes that get readers to imagine non-singular propositions. In other words, this approach invokes, at a pragmatic level, exactly the same type of general content as descriptivist theories invoke at the semantic level. As a result, it faces precisely the same problems in capturing the intentionality of prescribed imaginings as these other theories.

A different version of the gappy proposition theory is more plausible. On this approach, we would distinguish between (1) and (3) by appeal, not to different propositions, but instead to the different ways in which the same gappy proposition is to be imagined. Braun (2005) suggests that there would be a difference between imagining in a “Gregor-ish” way or a “Hamlet-ish” way. Of course the problem is to explain which type of imagining counts as “Gregor-ish” and which type counts as “Hamlet-ish.” A promising method is to appeal to the history of the uses of a name: uses of the name ‘Gregor’ are connected to each other and back to Kafka’s text, but not to uses of the name ‘Hamlet’ or Shakespeare’s text.14 This is the approach I will advocate. But before I can develop it, it is useful to look at a case that involves a referring name.

3 Singular fancies

Suppose that Frank, discussing Orwell’s 1984, says,
  1. (4)

    London is the chief city of a fascist state.


According to the account I have proposed, (4)—that is, Frank’s particular utterance in its specific context—is true so long as Orwell’s novel prescribes imagining that London is the chief city of a fascist state. Assuming that Orwell’s novel is about the real city London, to imagine as prescribed does not require readers to think of London in exactly the same way. Say that Frank, reading the novel in San Francisco, imagines that London is the chief city of a fascist state. Meanwhile his cousin Lawrence in London, also reading the novel, imagines (what he would express as) “this is the chief city of a fascist state.” Alternatively, suppose that Kripke’s Pierre, who never realizes that the city called ‘Londres’ in French is identical to the city called ‘London’ in English—and thinks of each in entirely different ways—reads 1984 in translation while in France, imagining that “Londres est la ville principale d’un état fasciste,” and then reads 1984 in the original English while in London, imagining that “London is the chief city of a fascist state.” We want to capture the sense in which Frank, Lawrence and Pierre (at both times) have all imagined as prescribed, even though they think of London under distinct modes of presentation.

A straightforward method is to say that they imagine the same singular proposition as a result of reading the novel: that is, a proposition that contains London itself, rather than a particular way of thinking of London, as a constituent. Let’s suppose for the sake of simplicity that being-the-chief-city-of-a-fascist-state is a property. Then the proposition can be represented as an ordered pair of an object and a property15:
  • (L)  <London, being-the-chief-city-of-a-fascist-state>

To say that the place London and the property of being the chief city of a fascist state are constituents of the proposition is just to say that a thought or utterance whose content is (L) will be true so long as the place has the property, regardless of how the place or property is represented. Now, I have said that an utterance reporting the content of a fiction is true so long as it states how the fiction prescribes imagining. If that is right, and imagining (L) is imagining as prescribed by 1984, we can say that (4) is true so long as 1984 prescribes imagining (L).

Others have argued, however, that the prescribed content cannot be a singular proposition. “No-reference” theories of fiction maintain that fictional contexts suspend reference to real individuals, so that the “London of 1984” is a fictional place rather than the real location (see, e.g., Wellek and Warren 1956; Riffaterre 1990). But this view is false. The fact that in the novel London is portrayed as having properties it never really possessed should not make a difference, so long as the correct interpretation of the fiction indicates that reference has been made to the real city.16 And it does. Apart from using the name ‘London,’ Orwell sprinkles clues throughout the novel to ensure that readers get the reference. For example, he writes that before the Fifties, “Airstrip one, for instance, had not been so called in those days: it had been called England or Britain, though London, [Winston] felt fairly certain, had always been called London” (Orwell 1962, p. 30). At the same time, various aspects of the dystopia are drawn directly from post-war Britain. Insofar as the city is fictionalized, the distortion is crucial to the power of Orwell’s vision: readers are supposed to imagine what London would be like if England had pursued a certain path. We should conclude that 1984 prescribes imagining about London.

A different objection is that (L) is insufficient to state how 1984 prescribes imagining. On one descriptivist view, to satisfy the prescription generated by 1984, the reader’s imagining must include a mode of presentation determined by the way London is represented in the fiction (Lamarque and Olsen 1994, p. 127). The idea is that a singular proposition like (L) places too few constraints on how we are supposed to imagine, because certain ways of thinking of London are authorized by the novel, while others are not. This approach is misguided, however, because it confuses modes of presentation—the ways in which thinkers access an individual—and what we are supposed to imagine about that individual.17 Lawrence’s thinking of London as the place where he lives can only increase the vividness of his engagement with 1984, and similarly for the many tourists who go to Baker Street in London to see Holmes’s famous address. Reading novels would be a less interesting experience for most of us if we did not make this sort of connection to our own lives and the world around us. In these cases readers imagine, of the place where they are, that such-and-such fictional events transpired there. They would fail to fulfill the prescription only if they imagined, of the place where they are (London), that it is the place where they are. Thus were Lawrence to utter, in a discussion of 1984 taking place in London, “This is the chief city of a fascist state,” his utterance would be true for just the same reason as (4) is true.

Yet even if (L) specifies the content we are supposed to imagine, this does not completely answer the question of how we are supposed to imagine in response to 1984.

What does it take to satisfy the prescription? First, satisfying the prescription requires more than imagining (L); it requires imagining (L) because so imagining is prescribed by 1984.18 I will return to this issue. Second, satisfying the prescription requires the reader to be in a particular mental state, or more specifically to have a particular propositional attitude. I assume that a person counts as imagining that London is the chief city of a fascist state so long as she stands in the relation constitutive of imagining, to a mental representation that means that London is the chief city of a fascist state.19 Because English fails to provide a noun for the mental representation that corresponds to imagining (though it is generous with belief, desire, wish, etc.), I adopt the term fancy for this purpose.20 Just as one might say that a believer has a belief with such-and-such content, I will say that an imaginer has a fancy with such-and-such content.

To say that Frank’s fancy has the content (L) is to assume that it exhibits roughly the same sort of syntactic and semantic properties as we attribute to the sentence ‘London is the chief city of a fascist state’. So we can say that Frank’s fancy has (at least) these two components: a representation of a thing, London, and a representation of a property, being-the-chief-city-of-a-fascist-state. I’ll use the term notion for a mental representation of an individual, and the term idea for a mental representation of a property or relation.21 On this way of speaking, Frank’s fancy is composed of a London-notion combined with an idea of being the chief city of a fascist state. It can be represented this way:
  • London-notion, idea-of-being-the-chief-city-of-a-fascist-state

It is common to construe notions as being associated with mental dossiers or file folders containing the information and misinformation we correlate with the individuals our notions are of.22 Let’s say that in Frank’s mental file on London are collected such ideas as being the capital of the UK, having double-decker buses and containing Arsenal fans. The name ‘London’ plays a role in Frank’s cognition as a kind of file-folder label, so that when he perceives or uses the name the relevant file is accessed.23 The next question is: what makes Frank’s London-notion a notion of London? I will return to this question.

Although 1984 prescribes imagining a singular proposition, it would be a mistake to conclude that all prescriptions to imagine about real individuals can be specified this way. In the next section I consider cases that require a different analysis.

4 Ways of imagining

Italo Calvino’s novel If on a winter’s night a traveler begins as follows:

You are about to begin reading Italo Calvino’s new novel, If on a winter’s night a traveler. Relax. Concentrate. Dispel every other thought. Let the world around you fade. (Calvino 1993, p. 3)

As one continues reading, it becomes clear that the opening passage refers to a fictive novel that you, the reader, have begun to read, and which also provides the title for the real novel. You actually get to read the first chapter of the fictive novel, describing a mysterious man in a railway station. Then, according to the same narrator, you discover that a bookbinding error has left you with only the first chapter, repeated over and over again. Naturally you are distressed:

The next day, as soon as you have a free moment, you run to the bookshop, you enter, holding the book already opened, pointing your finger at a page, as if that were enough to make clear the general disarray. “You know what you sold me? … Look here. … Just when it was getting interesting …” (Calvino 1993, p. 26)

The entire (real) novel consists of chapters about your attempts to track down the rest of the (fictive) novel, interspersed with the first chapters of different fictive novels—the chapters you find instead of finding the rest of the fictive If on a winter’s night a traveler.

Calvino’s real novel may be called an ‘interactive novel’, like those stories one reads as a child—“If you want to enter the dragon’s lair, go to p. 36; if you prefer to follow the winding path, go to p. 78”—or the video games more children play nowadays. Despite the differences, the important point is that these types of story prescribe what I shall call indexical imaginings. Each reader is supposed to imagine, of him or herself, that s/he is searching for a complete novel. Whether or not a reader can succeed at so imagining is a different question. For instance, at a certain point in the novel, it becomes clear that the reader is (fictionally) male, and this might hinder a female reader’s ability to go along with the story; nonetheless, she could still recognize that she was supposed to imagine about herself.

To the extent that I succeed in imagining as prescribed, we could specify the propositional content of my imagining with a singular proposition:
  • (F)  <Friend, searching-for-the-rest-of-Calvino’s-new-novel>

It will be immediately obvious, however, that (F) could not be used to specify the prescription to imagine generated by If on a winter’s night a traveler. If Walton, for example, were to read the novel, he would imagine a different singular proposition:
  • (W)  <Walton, searching-for-the-rest-of-Calvino’s-new-novel>

In short, when a fiction prescribes that the reader imagine something about him- or herself, a singular proposition containing the reader as a constituent does not enable us to specify the prescription.

The problem here is familiar from the literature on indexicals and demonstratives (such as I, you, here, now, this, that, this pear, that porcupine). To capture the role played by certain beliefs in action, we must appeal, not to the propositional content of those beliefs, but instead to something like Kaplan’s character. The belief that one would express as “That man’s trousers are on fire” would lead to rather different actions than the belief that one would express as “My trousers are on fire,” even if both beliefs have the same propositional content—as occurs in Kaplan’s example, where the believer does not at first realize that he is looking at a mirror (Kaplan 1989). The character of an indexical or demonstrative is, according to Kaplan, the linguistic rule that determines content from context: thus I designates the speaker, you designates the hearer, and so on. The character also corresponds to a psychological role: the thoughts that one would express in the first person are thoughts that involve what Perry calls one’s self-notion (Perry 1990), the essentially indexical way one thinks of oneself.24 Significantly, this psychological role can be specified intersubjectively: although we think of ourselves under diverse descriptions, we can all follow the prescription to imagine in a self-directed way.

From this perspective, (F) is incomplete as a representation of how I imagine, for I could imagine (F) without realizing that I am the object of the imagining, just as the amnesiac Castañeda could read about someone named ‘Castañeda’ without realizing that he himself is the subject of the book (cf. Castañeda 1966). If I imagined (F) without doing so in a self-directed way, I would not satisfy the prescription. Thus a better way of representing my imagining is as follows, where the subscript indicates the character under which I imagine the proposition:
  • (F’)  <Friend(self), searching-for-the-rest-of-Calvino’s-new-novel>

Since Calvino’s novel does not prescribe imagining about any particular reader, this approach suggests the following way of specifying the prescription to imagine generated by If on a winter’s night a traveler:
  • (C)  <_____(self), searching-for-the-rest-of-Calvino’s-new-novel>

That is, readers are supposed to imagine, of themselves, that they are searching for the rest of Calvino’s new novel. The italicized subscript is not part of the propositional content to be imagined, but instead specifies the way in which we are supposed to imagine that content, by specifying the type of notion to be deployed in so imagining. Any reader can satisfy the prescription, so long as s/he imagines this content under the character specified by (C): that is, by imagining in a way that involves a self-notion. By satisfying the prescription, the reader automatically “fills in” the blank in (C).

In the case of Calvino’s novel, the prescription is to imagine in a particular way, rather than (or in addition to) imagining a particular (gappy) proposition. In this respect it seems to contrast with the prescription generated by 1984, which was to imagine a singular proposition, namely (L). I suggest, however, that the prescription generated by 1984 should also be construed as indicating a way of imagining, one which happens to be specified in terms of the object of imagining. There are multiple ways to imagine, only some of which involve imagining singular propositions.

Still, because it does not require imagining a particular singular proposition, the prescription for If on a winter’s night a traveler provides a model for how we might approach the problem of (1). Readers satisfy the prescription generated by Calvino’s novel by imagining about themselves under a certain character. Because proper names are not associated with linguistic characters, the obvious question is this: how must readers imagine in order to satisfy prescriptions involving names? To answer this question, let’s consider a different type of case.

Before the rediscovery of Troy by Heinrich Schliemann in the nineteenth century, it was believed for centuries that Homer’s Iliad was about a fictional place and a fictional war. Suppose Samuel Johnson, talking about Homer’s epic in the eighteenth century, said,
  1. (5)

    Troy was burnt to the ground.

Because (5) is an accurate report of what happens in the epic, we can say that it is true if and only if readers of The Iliad are supposed to imagine, of Troy, that it was burnt to the ground. In other words, The Iliad prescribes imagining:
  • (T)  <Troy, having-been-burnt-to-the-ground>

Did eighteenth century readers of Homer’s epic satisfy this prescription, in spite of their mistaken view that there was no Troy? Yes. Although they were not aware of it, their imaginings were about a real city, the same city that my imaginings are about when I imagine as prescribed by The Iliad. While Johnson would have thought of Troy as a fictional city, whereas I think of Troy as a real city, it is the fact that we both imagine (T) that accounts for our satisfying the prescription. My fancy, like Johnson’s, has the following structure:
  • Troy-notion, idea-of-having-been-burnt-to-the-ground

Yet there is a difference between me and Johnson: Johnson’s understanding of the epic is only partial. I would argue that when a work of fiction refers to a real thing, full understanding requires recognizing the reference. Given the structure of his fancy, though, we can represent Johnson’s merely partial understanding as follows:
  • (T’)  <_____(Troy-notion), having-been-burnt-to-the-ground>

Johnson thinks that The Iliad prescribes imagining in the way indicated by (T’). But because, in fact, The Iliad is about a real place, the blank is filled in by the city. One need not know that to imagine as prescribed.

With this in mind, it seems that we have at least the beginning of an account of the debate about Gregor Samsa. Remember, Nabokov says
  1. (1)

    Gregor Samsa has been changed into a beetle.

And Smith replies,
  1. (2)

    Gregor Samsa has been changed into a cockroach.

Each is saying that The Metamorphosis prescribes having a different fancy, though these fancies involve the same Gregor-notion:
  • Gregor-notion, idea-of-having-been-changed-into-a-beetle

  • Gregor-notion, idea-of-having-been-changed-into-a-cockroach

We can now construe these statements as saying that The Metamorphosis prescribes imagining in the following ways:
  • (N’)  <_____(Gregor-notion), having-been-changed-into-a-beetle>

  • (S’)  <_____(Gregor-notion), having-been-changed-into-a-cockroach>

Of course, simply saying this does not actually solve the problem, without an answer to the question: what makes a Gregor-notion, a notion of Gregor? To answer this question, I propose first to address the question: what makes a London-notion, a notion of London?

5 Notions and their objects

What makes Frank’s London-notion a notion of London? This is a question about how the reference of singular thoughts is determined. Recall that Frank’s London-notion is associated with a dossier or mental file containing various bits of information associated with London. Let’s say that Frank in the US opened the file after hearing about his parents’ trip to Europe and that he has updated it through information (or misinformation) gleaned through reading the London Times, watching the BBC news and visits to his English cousin in Clapham. Many people have notions of London that do not share these features. Jeanne, who formed her notion of London when she first saw Mary Poppins, might never have been to London, might never watch the BBC or read the Times and so on. By contrast, George Orwell’s notion of London was no doubt formed at an early age, hearing about it from his parents in India, and he would have updated it directly by living in London for much of his life. In spite of the dissimilarities, Frank, Jeanne and Orwell all have notions of London.

Glossing over numerous subtleties, there are three competing explanations of this fact. According to the first, a given mental representation counts as a notion of London so long as London satisfies the majority of descriptions contained in the relevant dossier. On this descriptivist view, the reference of a notion is secured by fit. Assuming that Frank’s sources have been reliable, London is likely the object that fits Frank’s information. According to the second explanation, a notion is of London so long as it is embedded in a causal chain of reference-preserving links that originates in London. On this approach, the reference of a notion is secured by origin (cf. Kripke 1980). London is certainly the origin of Frank’s notion: he formed it based on his parents’ reports of their perceptions of London. The transmission of (mis)information from London to Frank’s parents to Frank is but one chain in a larger network of information flow that is rooted in the city; Jeanne’s and Orwell’s notions of London are also embedded in this network. According to the third explanation, a mental representation is a notion of London so long as London is causally responsible for most of the relevantly important information contained in the dossier (Evans 1973). On this view, the reference of a notion is secured by dominant source. London is the dominant source of information associated with Frank’s London-notion: when he updates his file he incorporates information that is itself about the city. The approach in terms of dominant source must also assume a network of (mis)information flow. The immediate source of most of Frank’s information about London might be the London Times, but the Times is not the object of Frank’s thoughts about London. Rather, the newspaper is a link in a network through which information about London is transmitted. I will call the networks posited by both the second and third approaches “intersubjective networks of notions,” or notion networks for short (Perry 2001, p. 128ff.).

As I have set up the case, Frank counts as thinking and talking about London on all three criteria. The debate over which criterion is correct turns on our intuitions in cases where they come apart, as in Kripke’s Gödel-Schmidt case or Evans’ Madagascar example. Given the objections to descriptivism raised by Kripke, Evans and others, most philosophers accept that reference for singular thoughts or for utterances containing proper names is not determined by fit. In what follows I assume that this is the right conclusion. Evans argues that the second explanation, in terms of origin, will not suffice either. Though I find Evans’ arguments persuasive, it is not my present aim to defend one or the other of these accounts of how the reference of singular thoughts is determined.

I will describe the difference between the two non-descriptivist views as a disagreement about whether the dominant source or the origin of a notion network is the root of the network.25 The term ‘root’ is just a placeholder for the thing that has the right relation (whatever that is) to the notion network to be the referent of a name.26 I take it that however the details are worked out, and whatever the label—“history of the uses of a name” (Donellan 1974), “referential frameworks” (Everett 2000), “mechanisms of co-reference” (Taylor 2000), “name-using practices” (Sainsbury 2005), or some alternative—any non-descriptivist theory of how the reference of singular thoughts is determined is committed to something like notion networks.27 It is fair to say that non-descriptivists have not worked out all, or even most, of the details (though some elaborations can be found in Everett 2000; Perry 2001; Sainsbury 2005). But they must assume that these details can be worked out, if we are to have an alternative to descriptivism. In what follows I argue that, given the assumption that notion networks secure reference, we have just what we need to explain the object-directedness of imaginings about fictional characters.

The key to this explanation is the concept of participation in a notion network. Let’s return to Frank’s utterance (4). To fulfill his intention to refer to London, Frank need not have used the name ‘London’; Pierre refers to the same city with ‘Londres’ because the notion network supporting both naming conventions is rooted in London. Co-reference is secured, not by the use of the same name or naming convention, but by the fact that the notions guiding the uses of the names are embedded in a notion network rooted in London.28 The same holds true for their imaginings: because their fancies involve notions connected, via this notion network, to London itself, their imaginings are about the city. I will say that it is in virtue of their participation in the London-network (which is not to be confused with their thinking about this network) that Frank’s and Pierre’s thoughts and utterances are about London.

Therefore, Frank imagines (L), because he stands in the relation constitutive of imagining, to a mental representation that means that London is the chief city of a fascist state; and this fancy refers to London because Frank’s London-notion is embedded in the appropriate notion network, rooted in London.

It looks as if we can give essentially the same account for fictional characters. In uttering (1), Nabokov identifies Kafka’s character by using the name ‘Gregor Samsa’. How is this possible, given that there is no Gregor? Assume that Nabokov formed his Gregor-notion in reading The Metamorphosis, opening a new mental file labeled with the name in which he collected ideas gleaned from the story. It makes sense to attribute a Gregor-notion to Nabokov because he can entertain a variety of different thoughts and make a variety of different claims about Gregor, such as that he has a sister and that he is a fictional character. Because Nabokov’s objective in writing the essay is to discuss his own interpretation of Kafka’s story, we can assume that this Gregor-notion guides his use of the name in (1). The connection between Kafka’s use of the name ‘Gregor Samsa’ and Nabokov’s thus has the same basic structure as the connection between Frank’s use of the name ‘London’ and Orwell’s; and similarly for the connection between Johnson’s use of the name ‘Troy’ and Homer’s use of the Greek original (though in this last case the connection is more indirect). It makes sense to say, therefore, that Nabokov participates in a notion network that relates his thoughts and utterances about Gregor to Kafka’s writing the story—in different terms, his use of the name defers to Kafka’s—though not through that story to a root. That there is some practice of using the name ‘Gregor Samsa’ to identify Kafka’s character, and that this practice traces to Kafka’s writing of the story, should be indisputable. Because the apparatus of notion networks is already needed to explain how we think and talk about real individuals, it is a useful way to understand this practice.

Once we recognize that Nabokov’s use of the name ‘Gregor Samsa’ constitutes participation in a notion network, it seems that solving the puzzle posed by Nabokov’s utterance is straightforward. For we can say that (1) is true if and only if (iff) The Metamorphosis prescribes imagining in the following way:
  • (N’)  <_____(Gregor-notion), having-been-changed-into-a-beetle>

And the reference to the Gregor-notion is now cashed out as the notion embedded in the Gregor-network. So if Nabokov is right about his interpretation of the story, imagining as prescribed means standing in the relation constitutive of imagining, to a mental representation that means that Gregor Samsa has been changed into a beetle; and this fancy is about Gregor so long as one’s Gregor-notion is embedded in the appropriate notion network.

This claim prompts a question: how do we know that we are “imagining about Gregor”? That is, how do we know that our notions are embedded in the appropriate network? The simple answer is that we do not. If this answer worries you, ask yourself how you know which real individuals you are thinking about. There is nothing in the explicitly entertained content of a thought like “London is in England” that guarantees that your thought is directed in the right way at London itself. You are presumably aware of the descriptions and perceptions you associate with London—the ideas collected in your mental file—but if descriptivism is wrong, your thought is not about London in virtue of satisfying these. What matters is that the London-notion associated with these ideas is embedded in a notion network rooted in the city. Most of the facts that determine singular reference for thought and discourse are simply not “in the head.” Nonetheless, some of these facts can be used to specify prescriptions to imagine, even about fictional characters.

6 Identifying networks

I have said that 1984 prescribes imagining a certain singular proposition via a London-notion, where this means a notion embedded in the London-network; and The Metamorphosis prescribes imagining a certain gappy proposition via a Gregor-notion, where this means a notion embedded in the Gregor-network. This account presupposes some way of identifying notion networks. How can this be done?

One immediately dismissible option is to individuate networks by their roots, to say, for instance, that the London-network is just the network that can be traced back to London. Not only do some notion networks, like the Gregor-network, lack roots, different notion networks might have the same root. Consider Frege’s (1997) example of isolated tribes living on opposite sides of a mountain, so that there are two distinct practices of referring to it using the names ‘Aphla’ and ‘Ateb’—two distinct notion networks rooted in the mountain. Here is a different scenario. Suppose that, contrary to fact, the belief in Troy’s fictionality persisted, so that when Schliemann unearthed the ancient city in present-day Turkey, no one identified the excavation with the city of Homer’s Iliad. Still, noting a number of similarities to the fictional place, Schliemann called his discovery ‘Troy’ or ‘Ilium’—using these names interchangeably—in honor of his favorite epic poem. It is plausible to say that this scenario involves two distinct notion networks, both of which are rooted in the actual city. These examples show that co-reference does not require sameness of network. By contrast, because the identification of fictional characters depends on the thoughts, and usually the writings, of persons, co-identification of characters does seem to require sameness of notion network, even if, as in the case of Santa Claus and Father Christmas, the networks branch in different directions (cf. Everett 2000). Until we have criteria for individuating notion networks, we do not have a complete explanation of co-identification.

One way to reply to this worry is just to point out that it is a problem for any standard non-descriptivist theory of singular reference. The accounts proposed by Kripke, Evans, Donnellan and others all assume that we can identify particular practices extended over time, connecting uses of a proper name to each other and back to the name’s referent. But this is not a very satisfactory reply. So I would like to sketch a proposal that will answer the question just to the extent necessary for my account.

I propose that notion networks can be identified deictically, by reference to embedded representations. Even if you do not know whether or not the name ‘Troy’ refers, you do know that you are using the term ‘Troy’ to identify the same place as Homer’s poem. So the Troy-network just is the practice that connects you to The Iliad.

This way of construing our identification of notion networks fits nicely with a further constraint on prescriptions to imagine that I mentioned above. When a work of fiction prescribes imagining about a real thing such as London, imagining as prescribed requires imagining the relevant singular proposition. Thus 1984 prescribes imagining (L), and for this reason (4) is true. But imagining (L) is insufficient for satisfying the prescription generated by 1984, because someone could imagine (L) without having any awareness of Orwell’s novel. A plausible additional requirement, then, is that to satisfy the prescription, one must imagine (L) because doing so is prescribed by 1984. Explaining the causal connection suggested by the “because” is not easy, however; we have to rule out, for example, cases where the imagining is prompted in some accidental way by the novel—for instance if the mention of the title provokes imaginings that exaggerate the horrors of Thatcher’s reign.

We are supposed to imagine about London because London is referred to in the text. The name London plus other cues prompt the imagining, causing those of us who recognize the reference to open our dossiers on London and to begin associating our pre-existing London-notions with new fictional information. But as the case of Johnson and Troy shows, we may not recognize the reference. In this case, we open a new dossier to collect information from the fiction. The same occurs with the names of fictional characters. In each of these cases, imagining as prescribed requires not merely imagining a particular proposition, but imagining with a notion embedded in the right notion network or part of a notion network.

Return to the hypothetical scenario where Schliemann performs a new dubbing of Troy, without recognizing that this is the same place as in The Iliad. Suppose that Sam, an archaeologist on the site, reads TheIliad. Among other things, he imagines that Troy was burnt to the ground. In doing so, he imagines (T). Though Sam does not believe the city he studies to be Homer’s setting, his reading then prompts an episode of daydreaming, in which he imagines the various events of the epic occurring in the excavated city. Once again he imagines (T). Yet there is an important sense in which Sam’s daydreaming does not satisfy the prescription generated by The Iliad. One way of explaining this is to say that Sam’s daydreaming involves a notion of Troy embedded in a distinct notion network from the one transmitted through the epic. By contrast, in the case of someone imagining (L) about Thatcher’s time as prime minister, the same notion network is involved, but the imagining involves a different branch or segment.29 Either way, it seems that imagining as prescribed requires imagining with a notion embedded in a network segment downstream from the fiction. So it makes sense to identify the relevant notion network as the one that includes the fiction.

Of course the content of the imagining need not make reference to the fiction. In imagining that Troy has been burnt to the ground, Johnson need not imagine that “the place The Iliad is about” or “the place The Iliad designates with ‘Troy’” has been burnt to the ground. Apart from wanting to preserve the singularity of the fancy, we also want to say that it is possible to identify the notion network indirectly. If you had never heard of The Iliad, and I told you one of the stories from The Iliad as if I had just made it up, you could still imagine as prescribed. Your Troy-notion would still be downstream from the fiction.

Although I think that this resolves the problem about identifying notion networks, there is still a further difficulty. The network supporting Frank’s use of the name ‘London’ can be traced through 1984 to London, since Orwell was writing about the real place. If, on the other hand, we trace back through the Gregor-network, we will not find an object that plays the right role to be the referent; instead, we will find that the network ends in what Donellan (1974) calls a block. Kafka’s making up the character in writing The Metamorphosis constitutes the event that blocks our finding a referent—there is no Gregor “beyond” or “behind” the novel to whom Kafka was referring. I have said that Frank’s fancy is about London because London is the root of the London-network. But if the Gregor-network ends in a block, a parallel explanation for the Gregor-directedness of the imaginings prescribed by The Metamorphosis is unavailable—there is no object that plays the right role to be the referent. Worse, if being the root of a notion network means being the origin of the notion network or the dominant source of information about Gregor, it looks as though the Gregor-network does have a root: Kafka’s writing The Metamorphosis.

In writing the story, Kafka institutes a new naming convention, and this act originates a new notion network because it is what allows multiple readers to identify Gregor.30 The fact that the ideas Kafka associates with Gregor determine whether or not statements like Nabokov’s count as true or false suggests that Kafka’s descriptions of Gregor constitute the dominant source of the information transmitted through the network. But neither the story, nor Kafka’s act of writing the story, is the referent of our use of the name ‘Gregor Samsa’, just as the Times is not the referent of Frank’s use of the name ‘London’. Our theory is incomplete without an account of the relation an individual must bear to a network to count as its root.

Again I am inclined to say that this is a more general problem. The accounts proposed by non-descriptivists all assume that the referent of a name is the individual who bears the right relation—whatever that is—to the relevant practice. If Kafka’s writing The Metamorphosis constituted both the origin of the Gregor-network and the dominant source of information about Gregor, neither theory would be satisfactory.

Consideration of our imagining about fictional characters should therefore be construed as a test case for theories of intentionality. The question is how to construe the reference-determining relationship. The solution must provide a way to rule out Kafka’s (writing of the) story as the root of the Gregor-network, and thereby to explain why it is blocked. Explaining why the story does not bear the right relationship to the use of the name ‘Gregor Samsa’ to be its referent, without falling into circularity, poses a challenge, especially because the solution must also exclude from the role of referent any real individual who might have been the model for a fictional character. (Compare the need to rule out the real Saint Nicholas as the referent of our use of the name ‘Santa Claus’.) In addition to excluding these options, the solution must explain how a blocked network can secure object-directedness when it has no root: why are certain thoughts about Gregor, rather than some other character or the Samsas’ apartment or anything else in the story? The right explanation will provide a more detailed account of what it takes to qualify as the origin or dominant source of information for a notion network, as well as what it takes to “create” a fictional character.

I do not have the space to develop such an account, though I suggest that it will appeal to the psychological mechanisms engaged in forming notions of individuals. Readers systematically form Gregor-notions in reading The Metamorphosis, notions distinct from their notions of the novel, of Gregor’s sister, of the Samsas’ apartment, etc. Psychologists studying reading comprehension and object-directed mental representations aim to tell a causal story about how this happens, and we should be able to appeal to this story in explaining the Gregor-directedness of our imaginings. But this is a project for another paper. For the present, I would like to suggest a different way to use the apparatus of notion networks to specify truth conditions for statements like Nabokov’s (1).

7 Contents and truth conditions

I have argued that (1) is true so long as The Metamorphosis prescribes imagining in a particular way, with a notion embedded in a particular notion network, the Gregor-network. Given that Nabokov’s use of the name ‘Gregor Samsa’ in (1) constitutes participation in the same network, we should be able to offer truth conditions for utterances like (1) directly in terms of the networks that support the uses of names. To develop this proposal, I borrow a framework for contents and truth conditions developed by John Perry (2001).

On Perry’s view, contents are just propositions used by philosophers to classify utterances and cognitive states by their truth conditions for various purposes. On this approach, utterances like (4) about London or (5) about Troy not only express singular propositions—in Perry’s terminology referential contents—but are also systematically associated with other contents that specify how the referential content is presented. Any given thought or utterance is associated with multiple contents and truth conditions. Rather than focusing on the truth conditions associated with the proposition expressed (the referential content), Perry argues that we should think of the truth conditions of an utterance or thought given various facts about it (Perry 2001, p. 80ff). Then we ask: given that such-and-such is the case, what else has to be true for the utterance or thought to be true? Different answers give us different propositional contents to specify the same thought or utterance.31

To illustrate the general approach, take my utterance of
  1. (6)

    I am in Paris.

The referential content of this utterance is a singular proposition. Given that ‘I’ designates Friend (and using a subscript to indicate the kind of content), (6) is true iff:
  • (6R)  <Friend, being-in-Paris>

Referential content represents the conditions on the subject matter of an utterance: the things in the world the utterance is about. Suppose, however, that you overhear (6) and do not know who uttered it. You still have a good idea of what would make the utterance true, since you know the meaning of the first-person pronoun in English. Given the meaning of ‘I’ but not the identity of the speaker, (6) is true iff:
  • (6I)  <The speaker of (6), being-in-Paris>

Perry calls this the indexical content of (6). (The italics indicate that the identifying condition, the speaker of (6), rather than the individual who satisfies the condition, is a constituent of the proposition.) Indexical content is reflexive because it refers to the utterance itself: it is the utterance that generates this level of content. Indexical content should be familiar from Kaplan’s discussion of character, though Kaplan does not count it as a type of content because it does not capture what is intuitively said by an utterance of (6). If I utter (6), I do not say something about the rule for using ‘I’; I say something about myself. Nonetheless, (6I) tells us something about the way (6R) is presented, and there can be little doubt that (6I) represents a condition under which (6) is true. Indeed, (6R) just is (6I) once we “load in” the fact that ‘I’ picks out Friend. On Perry’s approach, to say that an utterance has multiple contents is just to say that the truth conditions of the utterance can be specified in a variety of ways.

We can apply this account to proper names. Consider again Johnson’s utterance,
  1. (5)

    Troy has been burnt to the ground.

I have said that (5) is true iff The Iliad prescribes imagining in a particular way, namely one that involves imagining the following proposition:
  • (T)  <Troy, having-been-burnt-to-the-ground>

(T) is the referential content of (5): it takes as given the fact that the name ‘Troy’ in (5) refers to Troy. Because on Perry’s scheme an utterance like (5) has multiple kinds of content, it will be helpful to reformulate the prescription. Let us say that fictions prescribe having fancies with various kinds of content. Then we can say that The Iliad prescribes having a fancy with the referential content given by (T). Using a subscript to indicate the kind of content, that is:
  • (5R)  <Troy, having-been-burnt-to-the-ground>

Because Johnson and his contemporaries do not know is that Troy is a real place, they do not know that (5) is true if and only if The Iliad prescribes imagining (5R). But their ignorance of this truth condition on the subject matter of the prescription does not mean that they have no idea what would make the utterance true. In particular, Johnson knows that ‘Troy’ is a proper name that derives (through translation) from The Iliad. What he doesn’t know is that the name refers.

This suggests that one option for representing the eighteenth century understanding of the prescription to imagine is by reference to the name itself, so that imagining as prescribed means imagining in a way that invokes the name ‘Troy’. But appeal to the name will not help in the present case. The mere fact of translation suffices to make this point: the original prescription to imagine did not use the name ‘Troy,’ and we do not want to say that French readers who think of Troy as named ‘Troie’ cannot satisfy the same prescription. Furthermore, even if we restrict ourselves to The Iliad in English translation, Troy is designated in a variety of other ways, for example as “Ilium” and as “Priam’s city.” Johnson and his contemporaries know that there are multiple ways to think of Troy. As a result, they know that Johnson would have made the same claim about what is fictionally the case in The Iliad had he said,
  1. (7)

    Ilium was burnt to the ground.


We need a way to specify the prescription that takes as given that the various ways of designating Troy are tied to each other and to Homer’s text, though not that they refer to a real place.

On the analysis I have offered, this means that Johnson’s use of the name ‘Troy’ takes for granted a notion network—the Troy-network—that leads back (at least) to The Iliad, one that also supports the convention for using ‘Ilium’. Johnson presupposes that his use of ‘Troy’ in (5) connects appropriately to Homer’s epic; it is because his use of the name defers to the fiction that he can use it to talk “about the same thing” as Homer does. He also knows that he can use the name ‘Ilium’ (and other terms) to talk about that thing. Although Johnson does not realize that Troy is a real place, he does know that according to the fiction it is; or in different terms, that within the fiction the Troy-network has a root. If this is so, then it seems we can give truth conditions for (5) and (7) by reference to the network itself, using what Perry calls network content. Given that the use of ‘Troy’ in (5) and the use of ‘Ilium’ in (7) are supported by the Troy-network, both utterances are true iff The Iliad prescribes having a fancy with the network content:
  • (5N)  <the x: x is the root of the Troy-network, having-been-burnt-to-the-ground>

It is crucial to recognize that imagining this way does not mean imagining that the x such that x is the root of the Troy-network has been burnt to the ground (much as understanding (6) does not require entertaining the thought that the speaker is in Paris). Just as (5R) provides a way for us to specify prescribed fancies via their connections to real objects, (5N) provides a way for us to classify them via their connections to notion networks. We privilege propositions like (5R) because they reflect the subject matter of our thought and discourse, what they are about. Network contents, by contrast, tell us something about the way a given referential content is entertained, by specifying which notion network secures the reference of the thought (if any). Thus (5N) does not say that to imagine as prescribed, a reader must have a fancy about the Troy-network (a fancy one might specify using a referential content), but instead that she must imagine in a way that involves participation in that notion network. When I imagine as prescribed by The Iliad, my imagining is directed at Troy itself—but its directedness is secured by participation in the Troy-network. As a result, the contents are systematically related: (5R) just is (5N) once we load in the fact that Troy itself satisfies the condition, the x such that x is the root of the Troy-network. If readers of The Iliad imagine (5R) in virtue of participation in the Troy-network, their fancies thereby have the network content (5N).

If this is right, we can give the same analysis of Nabokov’s utterance,
  1. (1)

    Gregor Samsa has been changed into a beetle.

(1) does not express a complete proposition. At the level of referential content it expresses at most a gappy proposition. Nonetheless, because Nabokov’s use of ‘Gregor Samsa’ constitutes participation in the Gregor-network, we can provide truth conditions for (1) by appeal to that network. Then (1) is true iff The Metamorphosis prescribes having a fancy with the network content:
  • (1N)  <the x: x is the root of the Gregor-network, having-been-changed-into-a-beetle>

Again, (1N) tells us, not what we are supposed to imagine in response to Kafka’s story, but how: by participating in the Gregor-network.

Even if we keep these distinctions in mind, there is a serious objection to the present approach. Network content seems to flout a plausible constraint on theories of content: that content attributions be constrained by the concepts thinkers possess.32 While thinkers plausibly have concepts of what their thoughts are about, it is implausible to suppose that they possess the concept of a notion network. How can we take Johnson’s utterance (5) to have the content (5N) if he has no concepts of networks or roots? How can Kafka’s fiction prescribe imagining in the way indicated by (5N) if he too lacks those concepts? Notice that the same problem does not arise for indexical content. When you overhear my utterance of (6) without knowing to whom the first-person pronoun refers, you are aware of the linguistic meaning of ‘I’, as am I. So while I do not say that the speaker of (6) is in Paris, it is reasonable to attribute to you knowledge of the indexical content of my utterance purely on the basis of your linguistic competence with ‘I’; and the same applies to me. A similar analysis is problematic in the case of network content.

There are two replies to this objection. The first is to repeat that on the present approach, contents are just tools used by philosophers to classify utterances and cognitive states for various purposes. Network content need not be accessible to speakers and hearers or thinkers themselves, so long as they correctly capture the truth conditions of the relevant utterances. For example, if we adopt some version of Davidson’s paratactic account of propositional attitude ascriptions (Davidson 1968), we could say that in (5) Johnson is merely demonstrating, or offering a sample of, the sort of content one is supposed to imagine.33 To do this does not require having the concepts of roots or networks, but only the ability to recognize that “Troy has been burnt to the ground” is similar in relevant respects to what the fiction says: that it “samesays” with Homer. We, as theoreticians, may then interpret (5) as offering a sample of referential content for some purposes and of network content for other purposes.

A different reply is to reject the assumption that to possess a concept, a thinker must be able to articulate it. Perhaps having the concept of a notion network manifests itself simply as the ability to speak and think in ways that obey certain rules—for example, to use the name ‘Troy’ as systematically co-identifying with other expressions—even if one cannot verbalize those rules.34 In fact I think that we have more of a grasp of the concepts of networks and roots than this. As discussed in the previous section, when we imagine in response to a fiction we are supposed to imagine about whatever the fiction is about—London, Troy, Gregor Samsa, etc.—and to do so precisely because that is what the fiction is about. Similarly, when we make claims about the content of a fiction, we take ourselves to be talking (or writing) about whatever the fiction is about, and to do so because that is what the fiction is about. Johnson is aware that his use of ‘Troy’ in (5) is supposed to identify the place in the epic, variously called ‘Troy’, ‘Ilium’, ‘Priam’s city’ and so on. Furthermore, he is aware that the epic prescribes imagining that these terms co-refer: that they all lead back to a particular real place. Johnson thus knows that The Iliad prescribes imagining in such a way as to take various terms in The Iliad to refer to a single, real place. Given the definitions of ‘notion network’ and ‘root’, this amounts to imagining in such a way as to take the Troy-network to have a root. So it does not appear far-fetched to say that Johnson knows that the poem prescribes imagining in a way that is captured by network content, even if he cannot articulate this way of putting the point.

I believe that Perry’s framework, including his concept of network content, provides a useful way of thinking about truth conditions; however, my brief discussion is insufficient to give it a full defense. Still, the main points I have made in this chapter do not depend on accepting this framework. Those who doubt the explanatory value of recognizing network content should still accept that fictions prescribe imagining in certain ways, and that these ways can be specified by notions embedded in notion networks. It remains the case that to imagine as prescribed requires participating in the relevant network.

8 Conclusion

Starting from the assumption that statements like Nabokov’s (1) should be construed as reports about the ways in which fictions prescribe imagining, I have developed two ways of representing the truth conditions such utterances. Both permit us to account for the sense in which Nabokov and Smith are arguing “about the same thing” even though there is no thing they are arguing about. On the first approach, their disagreement is about whether The Metamorphosis prescribes imagining in the way specified by (N’) or (S’):
  • (N’)  <_____(Gregor-notion), having-been-changed-into-a-beetle>

  • (S’)  <_____(Gregor-notion), having-been-changed-into-a-cockroach>

On the second, their disagreement is about whether The Metamorphosis prescribes imagining in the way specified by (1N) or (2N):
  • (1N)  <the x: x is the root of the Gregor-network, having-been-changed-into-a-beetle>

  • (2N)  <the x: x is the root of the Gregor-network, having-been-changed-into-a-cockroach>

In either case we have an explanation of both the structural similarity and the metaphysical difference between fancies about real individuals and fancies about fictional characters. Because both can be specified by appeal to notion-types and notion networks, we can account for the intersubjectivity of discourse about fictional characters as well as cases where reference is not recognized. Because only fancies about real individuals can be specified by a singular proposition or referential content, we can account for an important difference in what it takes to understand a work of fiction, a difference that is obscured by realist theories.

I said in discussing 1984 that failure to recognize that London is designated by Orwell’s novel indicates that one’s understanding of the work is only partial. This is because the network supporting Frank’s use of the name ‘London’ can be traced through 1984 to London, given that Orwell was writing about the real place. When the correct interpretation of a work determines that it prescribes imaginings about a real individual, readers are usually expected to have some background information about the individual, or at least to recognize that the work refers to something real. I would argue that eighteenth century readers of The Iliad therefore had an incomplete understanding of the epic. Where works of fiction refer to real things, full understanding of the fiction is reflected at the level of referential content. But although this sort of collateral information may be necessary for full or better understanding of a work, we want to allow that readers can satisfy prescriptions to imagine without possessing it; otherwise we cannot account for the sense in which multiple individuals with different levels of background information can share a common core of basic understanding.

By contrast, there is no more complete understanding of The Metamorphosis than is reflected by gappy propositions (imagined in certain ways) or by network content. Of course, if Kafka had been fictionalizing a real individual rather than inventing a character, there would be a singular proposition representing full understanding of the story. But in fact, the truth of (1) does not depend on there being anyone who is the root of the network, of whom we are supposed to imagine that he turned into a beetle. That is why appeal to notion networks reflects how the fiction prescribes imagining; and that is why, if Nabokov’s interpretation is correct, (1) is true and (2) is false.

It is worth mentioning, in closing, that the account offered here has applications that go beyond its immediate target of utterances such as (1) and (2). One criticism of story operator analyses is that they cannot be extended to other sorts of statements about fiction, such as “Gregor Samsa is a fictional character,” “Gregor suffers more than Joseph K.,” “I think about Gregor,” “John pities Gregor,” and “Gregor does not exist,” none of which can be prefixed by ‘According to the story.’ The same limitation might appear to affect my account: if “Gregor Samsa is a fictional character” is true, it is not so because The Metamorphosis prescribes imagining that he is (in the story he is every bit as real as you or I). However, because all of the various statements about Gregor require participation in the Gregor-network, this element of the analysis will be relevant to understanding the sense in which they can be “about Gregor” and can thus be used to specify their truth conditions.35 For example, on a disavowal-through-pretense analysis of negative existential statements, “Gregor does not exist” is treated as claiming that reference attempts of the Gregor-kind fail (Walton 1990, Chap. 11; Kroon 2000); but proponents of this analysis rarely explain what constitutes a “reference attempt of the Gregor-kind.” Similarly, addressing the problem posed by intensional transitive constructions such as “I think about Gregor” and “John pities Gregor” requires an explanation of how our thought or pity can be directed at Gregor when there is no such individual. Although I do not have the space to defend this claim here, I believe that the account I have proposed can be utilized to provide the relevant explanations in these and other cases.36 As such it offers a promising strategy for approaching the full range of fictional discourse.


On some accounts such statements cannot be true or false, but only correct or incorrect; I address this below.


In fact Eagle’s account of correctness conditions appears similar to my proposal, though it is not developed in detail.


Though Braun (2005) argues that gappy propositions are just false.


Hanley (2004) argues that Lewis should use the method of intersection to deal with many of these cases and, where the method of intersection fails, we need not necessarily accept a contradiction just because it is in the text. Even if Hanley’s defense of Lewis is sufficient—and I am skeptical—I think Walton’s approach is to be preferred, both because it captures the central role of imagining in response to fictions and because it applies to a wider range of fictional discourse (as discussed below). Thanks to an anonymous referee for pressing me on this point.


This does not mean that we are obligated to imagine everything that is fictionally true. To say that it is fictionally true that Anna Karenina needs oxygen is just to say that if the question arises—which it usually doesn’t—and we have to decide between imagining that Anna needs oxygen and imagining that she does not, we should imagine the former (Walton 1990, p. 40).


Thanks to an anonymous referee for pointing this out.


This is merely an assumption. While Nabokov makes a good argument, there is no fact of the matter whether we are supposed to imagine that Gregor becomes a beetle or any other kind of vermin. This indeterminacy does not indicate anything mysterious about the nature of fictional characters, but is rather a result of the fact that there is a limit to the amount of information a work of fiction gives us in prescribing imaginings.


Not all languages translate the Greek and Latin names for this character differently, and this is one reason to think that the English names co-identify. But the justification of any claims about reference in works of fiction must appeal to the interpretation of those works. I would defend the identity of the character between Homer’s Iliad and Dante’s Inferno, for example, by arguing that one could not properly understand the latter without recognizing the character as Homer’s.


There are differences among these theories. One view is that the names of fictional characters function as variables in existentially quantified sentences prefixed by a fictionality operator; the name’s semantic contribution consists in all the predicates associated with it in the fiction (Kaplan 1973; Currie 1990). An alternative view is that a name appearing within the scope of a fictionality operator refers to its “customary sense,” which is given by the core of salient properties associated with the name (Lamarque and Olsen 1994). My objections are meant to apply to both approaches.


For more detailed criticisms of realism, see Everett (2005) and Friend (2007).


As Walton has pointed out to me, the solution I offer below can be construed as a way of individuating kinds of pretense within the constraints of his account.


See Braun (1993) for an argument in favor of the latter (which he there calls the ‘unfilled proposition’ view) over the former. For further development of the gappy proposition view, see also Adams et al. (1997), Braun (2005) and the articles cited in the next note.


Adams and Stecker (1994), Taylor (2000) and Adams and Dietrich (2004) defend versions of this view. See Reimer (2001), Everett (2003) and Green (2007) for criticisms.


Adams and Dietrich (2004) also invoke the “causal history of the name” (p. 128ff) to fill out the pragmatic portion of the gappy proposition view, but do not develop the idea in detail.


While it makes a difference that we represent contents with structured propositions (rather than possible worlds), nothing in my argument depends on how these propositions are construed. I take propositions to be nothing more than a way of representing truth conditions.


This claim purposely leaves open the possibility that in other fictions in which the name ‘London’ is used, our interpretation of the work would lead us to conclude that there is no reference to the real city.


As Anthony Everett has pointed out to me, there could be fictions that require us to imagine about real individuals under distinct modes of presentation, for instance a story in which “Chomsky the linguist” and “Chomsky the political activist” are two different people. I take it that such stories are rare, though they do deserve fuller discussion than I can offer here.


Ian Proops drew this to my attention.


The assumption that propositional attitudes can be understood as involving structured mental representations is not uncontroversial, but in my view it is the best explanation of many features of propositional attitudes; see, e.g., Fodor (1987, 1990). Further, research on narrative comprehension in cognitive psychology and linguistics supports the hypothesis that we create and update mental representations of characters while reading (see, e.g., the papers in Goldman et al. 1999). I say nothing here about the relations that constitute attitude-types, such as imagining, believing, desiring, etc., though I assume they are distinguished functionally.


I owe this terminology to Ian Proops (in conversation).


This terminology is introduced in Crimmins and Perry (1989).


The term dossier originates in Grice (1969).


We can, of course, have notions that are not labeled with names, including notions of fictional characters. In these cases we usually appeal to a description as a way of accessing information stored in memory.


This is not to say that the psychological mode of presentation is identical to the (linguistic) character (see Recanati 1993, p. 69ff). The distinction is not important for present purposes, however, so long as there is a systematic correspondence.


I take the term ‘root’ from Perry’s unpublished “Saying Nothing?” (1997).


If it is possible to refer to abstract objects that lack causal powers, we should doubt that this relation can be reduced to purely causal connections; but the basic idea of notion networks does not assume any such reduction.


A theory need not assume referentialism about names to invoke something like notion networks (see, e.g., Sainsbury 2005).


It is possible that there be more than one notion network rooted in a real individual. I address this issue below. For now I assume that Pierre, Frank, etc., participate in only one network.


More needs to be said about how to distinguish different “branches” of notion networks (see Everett 2000). One other advantage of appealing to segments of notion networks is that they help explain cases in which our intuitions about identification diverge. For instance, we typically identify Superman and Clark Kent but there are (Frege-style) contexts in which we might not. We typically identify Odysseus/Ulysses as the same character in Homer, Virgil, and Dante; but sometimes it is useful to speak of distinct characters. The contrasts can be explained by appealing to the whole network or to different branches. (Thanks to an anonymous referee for drawing my attention to this point.).


To say that the convention was instituted in the story is a simplification, since Kafka would have come up with the name and the character beforehand, for example in working the story out in rough drafts. I ignore this complication here.


Perry offers a brief account of fictional discourse (Perry 2001, pp. 170–172), but it differs from mine in several respects.


Thanks to Jason Stanley.


I owe this suggestion and its elaboration to Anthony Everett.


John Perry suggested this line of argument.


For example, one could argue that “Gregor Samsa is a fictional character” is true so long as (very roughly) (i) the Gregor-network originates in a fiction, and (ii) that fiction prescribes imagining that the network has a root.


Thanks to an anonymous referee for emphasizing these further applications.



Thanks to Kendall Walton, lan Proops, Peter Ludlow, Jason Stanley, Jessica Wilson, Gabriel Segal and Anthony Everett for very helpful comments on previous drafts. Thanks especially to John Perry, whose assistance from the beginning of this project has been invaluable. Thanks also to audiences at the Institut Jean Nicod (March 2005), the University of Michigan Aesthetics Discussion Group (November 2002) and the NEH Institute in Art, Mind, and Cognitive Science at College Park, Maryland (July 2002) where I presented early versions of this paper.

Copyright information

© Springer Science+Business Media B.V. 2009