This is a post by Manuel García-Carpintero (University of Barcelona).
My recently published book Truth and Reference in the Making of Fiction (CUP 2025) puts together work on philosophical issues about fiction that I have been doing over the past decade and a half. The core topic of the book is ontological: what is the nature of fiction, in contrast with non-fiction? Say, what is, if any, the metaphysical difference between Peter Weir’s 1982 fictional film The Year of Living Dangerously, and Joshua Oppenheimer’s 2012 documentary The Act of Killing, both intuitively about the same Indonesian 1965-6 events?
I approach the topic as the specific form of metaphysical inquiry nowadays called metasemantics – an investigation into the metaphysical underpinnings of meaning. I adopted Andrea Bonomi’s classification of the three most relevant types of discourse whose metasemantics we must study to build and evaluate theories of fictionality. Consider the sentence ‘Holmes is (now) famous’. We should distinguish textual utterances thereof, those that might be found in a fictional story by Conan Doyle, purportedly reproducing the fictional stories that, in the stories, Watson circulates about the exploits of his friend; paratextual utterances, used by readers to characterize the content of a fiction such as Conan Doyle’s Holmes stories; and metatextual utterances used to state facts in the actual world about the stories, beyond their plot or narrative. Textual utterances intuitively are not evaluable as true or false, while paratextual and metatextual utterances are. The former are made true or false by what the fictional stories recount, but not the latter, which intuitively are made true or false by what transpires in the actual world. This is why the truth-value of paratextual but not metatextual utterances is preserved when they are paraphrased by means of an operator, “according to/in the Doyle stories Holmes is famous”; this might be true even if the metatextual “Holmes is famous” is false, and also the other way around.
An interpretive decision that structures the book is a taxonomy of theories of fictionality. I distinguish, first, the mere pretense theory influentially articulated and defended by Searle (“The Logical Status of Fictional Discourse”, 1975) and Margaret Macdonald in a sharper 1955 paper. It was taken for granted by most Analytic philosophers from Frege until the 1970s, including Austin, Kripke, and Lewis. Second, a skeptical view that denies any robust fiction-nonfiction distinction, which Nelson Goodman advanced and which has been compellingly articulated in recent years by Stacie Friend. Finally, the dedicated representation theory classically articulated by Kendall Walton (and Moore and Wolterstorff, among others), a version of which the book endorses. These are all theories of textual utterances; it is intuitively reasonable that any proper account of paratextual and metatextual utterances should flow from such an account.
Many philosophers misclassify in one category all theories of fictionality that mention pretense; on this criterion, both Walton and Searle advance theories in the same class. But this is the wrong classification, putting together views that, when it comes to the theoretical resources they provide to address the main philosophical puzzles that fiction poses, should be in contending taxa, like the two just mentioned. Of course, every theorist should acknowledge that many fictions are generated by pretense. Drama is a clear-cut case; Olivier pretends to be Hamlet in the 1948 film adaptation he directed, generating in this way the fictional content of the play; and, as many philosophers point out, we can think of Conan Doyle as “playing” Watson in writing the stories. What is distinctive of the mere pretense view upheld by philosophers from Frege to Kripke, Searle, and Lewis is that they claim that pretense is all that there is to T(extual)-uses. There are the standard speech acts that Watson appears to be doing in fictionally producing the utterances constituting the stories – mostly, assertions reporting Holmes’ exploits. The fiction is just the pretense (by Conan Doyle, as suggested) that all such acts are really taking place. In contrast, dedicated representation theories of T-uses like Walton’s take those utterances to be of a distinctive representational kind of their own, usually referred to nowadays as fiction-making. As Walton points out, the crucial difference between the two views doesn’t lie in the invocation of pretense, which both may share in the cases we are considering. The crucial difference is that the mere pretense view takes speech acts like assertion to be explanatorily prior, and fiction to be understood as the pretense of performing them. On dedicated representation theories like Walton’s, T-uses are not in any way derivative from assertions, questions, or commands; for all the account proposes, acts of fiction-making might be explanatorily prior to the others, or (more plausibly) they might be explanatorily on a par with them.
Mere pretense theorists like Macdonald and Searle in fact present dedicated representation accounts as the relevant alternative to the views they put forward. The characterization that Macdonald offers (suggested by Moore, she says) is an inchoate Waltonian account; as she puts it, while core assertions are designed to share knowledge or belief, fictions are devised to share imaginings. Macdonald and Searle offer related arguments against such views; here too, Macdonald’s presentation is much more compelling than Searle’s. Their central objection is that being able to understand standard speech acts like assertions, requests, and acts of reference is sufficient to understand fiction. In the book, I argue that this is not the case: audiences need to know that T-uses are put forward with a force or point of their own, so that, for instance, a manifest contradiction or impossibility might not be in any way problematic for the job that T-uses are meant to perform.
In turn, dedicated representation theorists object to mere pretense accounts by presenting counterexamples. Against sufficiency, they mention cases in which we pretend to do ordinary speech acts without any fiction (or T-use) being generated. We may, for instance, rehearse a speech or simply clear up our voice. Against necessity, cases of fictions that do not involve any pretense of making assertions or any other speech act are mentioned; for instance, modernist fictions whose authors are at pains to make clear to their audiences that they are just producing fiction. In the book, I argue that such counterexamples are a manifestation of the central objection to the mere pretense view, which is that it lacks the theoretical resources needed to offer minimally prima facie compelling answers to the philosophical concerns that fictions generate, especially in comparison with dedicated representation views. Most of the book is devoted to defending, on abductive grounds, the version of the view that I uphold. I’ll outline three illustrative examples, two of them adduced by Searle in support of his view.
The first concerns reference in fiction. Several theorists in the analytic tradition, including Kripke, van Inwagen, and Thomasson, have defended a view that Roman Ingarden argued for in the phenomenological tradition. This is the view that fictional names like ‘Holmes’ are not empty; according to these theorists, when we say that Holmes doesn’t exist what we mean is only (roughly) that he is not among the living. ‘Holmes’ does refer, to an abstract entity, a “part” of the works in which reference to “him” occurs. Dedicated representation theorists can develop this view in precise ways. On their view, T-uses are standard literal uses of language with a specific point (like suppositions or guesses), and hence they have a semantics. Such semantics affords candidate referents for ‘Holmes’, for instance, what semanticists call a “discourse referent”. Now, Searle articulates this “realist” view of fictional characters and claims that his pretense account properly grounds it. He suggests that the explanation is that, in pretending to refer with ‘Holmes’, Doyle creates an object that is a target for reference. Without further elaboration, this obviously lacks any explanatory power; typically, pretending to do A doesn’t generate a real instance of A.
The second philosophical issue that Searle claims his theory illuminates is the intuitively manifest fact that fictions assume assertoric commitments. Searle says that the mere pretense theorist affords the explanation that not all assertoric utterances in a fiction are uttered in pretense; some of them are truly asserted, without pretense. He offers as an example the famous first sentence of Tolstoy’s Anna Karenina about happy and unhappy families. But this example makes it easy to overlook the problem with this account, which is that assertorically committed utterances in fictions are intertwined with fictional ones and contribute to the fictional plot. For instance, Doyle presupposes the geography of London at the time of his writing, so that we are expected to assume that Waterloo is farther away from where Holmes and Watson live than Paddington; but the place in Baker Street where they live and from where they walk is fictional, as of course they also are. The book discusses other philosophical issues for which dedicated representation theorists have the resources to provide compelling explanations, which mere pretense theorists lack. As a final example that the book discusses, I’ll mention the issue of the interpretation of fictions, i.e., the determination of which P-utterances are true, in particular interesting ones constituting controversial, disputed interpretations.
The most popular alternative to dedicated representation theories nowadays is not the mere pretense theory, which most philosophers have rejected after Walton for considerations like the ones just rehearsed, but the very well-developed and interesting version of Goodmanian skepticism articulated by Stacie Friend in several articles and now in a 2026 OUP book, Matters of Fact and Fiction. She relies on the point that fictions typically include contents presented with assertive force. Greg Currie deals with this with the analogy of impure water: like it, fiction works are mixtures of fictive and non-fictive utterances. Friend points out in response that theories of fictionality are primarily accounts of the taxonomic fiction/non-fiction distinction, the one picked out by the concept that librarians use to classify them; and this applies primarily to whole works, because such categorizations determine the values that guide our appraisals – say, truth for non-fiction, perhaps affording enjoyable imaginings to the expected audience for fiction. She observes that, at the other end of the spectrum, non-fiction works typically also include little fictions: thought experiments in philosophy, counterfactual alternatives in history, ideal models in empirical science. She then argues that containing made-up content that invites imagining is just a standard feature of fiction, while including true content that invites belief is not. However, these traits are, according to her, not essential; fictions may even be fully constituted by contents that invite belief (think of “autofiction”), while non-fiction may mostly invite imaginings without belief (a well-researched counterfactual history).
What, then, are the categorical features distinguishing the two kinds? Friend appeals to relational features, in particular authors’ intentions validated by the relevant authorities, like publishers and critics. I argue in response that this is an unjustifiably skeptical account of the fiction-nonfiction divide. A main consideration is that it doesn’t capture the correct critical point that Friend makes against Currie that good theories should explain the appreciative point guiding classification of works into such categories. Why would we care about what authors and publishers classify as non-fiction, if they may classify as such works that are not meant to be true? Likewise with fiction, if the authorities may classify as such works that engage our imagination as much as a phone book. As an alternative, I rely on Williamson’s influential normative account of assertion as a good account of core non-fiction, and I develop a similarly normative account of fiction on which it is essential for them to invite imaginings, on the presumption that engaging in them will be worthwhile for us.
References:
García-Carpintero, Manuel (2025): Truth and Reference in the Making of Fiction, Cambridge, Cambridge University Press.
Manuel García-Carpintero got his PhD at the University of Barcelona in 1988 and has taught there since then; he is also an invited professor at the University of Lisbon. He works on the philosophy of language and mind, and related epistemological and metaphysical issues. A book putting together his work on assertion, ‘Tell Me What You Know’, will be published by Oxford UP the next year.
