Aesthetics for Birds

Aesthetics and Philosophy of Art for Everyone

Are There No Fictional Truths?



What follows is a guest post by Roy T. Cook. Roy is an extremely nerdy associate professor of philosophy at the University of Minnesota – Twin Cities, a resident fellow of the Minnesota Center for Philosophy of Science, and an associate fellow of the Northern Institute of Philosophy – Aberdeen, Scotland. He has published over fifty articles and book chapters on logic, the philosophy of mathematics, the philosophy of art (especially popular art). He co-edited The Art of Comics: A Philosophical Approach (Wiley-Blackwell 2012) with Aaron Meskin, and his monograph on the Yablo Paradox is forthcoming from Oxford University Press. He is also a co-founder of the interdisciplinary comics studies blog PencilPanelPage, which recently took up residence at the Hooded Utilitarian, and hopes to someday write a book about the Sensational She-Hulk. He lives in Minneapolis, Minnesota with his wife, two cats (Freckles and Mr. Prickley), and approximately 2.5 million LEGO bricks.

Standard approaches to the interpretation and evaluation of a work of fiction have it that some claims are true-in-the-fiction, such as “Sherlock Holmes lives on Baker Street”, and other claims, such as “Sherlock Holmes is a Martian”, are false-in-the-fiction. Put simply, they adopt an alethic approach to fiction. In this post I want to challenge that assumption, and propose an alternate, probabilistic account.

Consider “The Final Problem”, the 1893 Arthur Conan Doyle story where Holmes is killed while battling Moriarty at Reichenbach Falls. This story was followed by the Great Hiatus, a ten-year period during which Conan Doyle published no new Holmes stories and the great detective remained dead. In 1903, however, due to pressure from the public, his publishers, and his creditors, Doyle returned to write “The Adventure of the Empty House”, which details the shocking survival and return of the world’s most famous detective.

Now, the critical question here is the semantic status of the sentence “Holmes is dead”, uttered by a reader – let us call her Betty – between 1893 and 1903 who has read all of the relevant stories up to the Great Hiatus.  Clearly, Betty is correct in some sense when she utters “Holmes is dead” during this period. Not only does the fiction clearly indicate that Holmes died, but Doyle himself wrote this story with the intention of definitively killing off Holmes so that he could concentrate on his ‘serious’ writing – the historical novels that he thought would be his true literary legacy (insert ironic comment here!). Further, on most accounts of fictional truth, the claim “Holmes is dead” is true-in-the-fiction during the Great Hiatus (for example, on a Waltonian account, the Holmes story as it existed at that point clearly prescribes that we make-believe that Holmes died). But, as we learn in 1903, Holmes isn’t dead! So Betty’s utterance of “Holmes is dead” during the Great Hiatus is (apparently) somehow faulty, since the same utterance is false-in-the-fiction in 1903, despite the fact that her interpretative and evaluative actions were beyond reproach during the Great Hiatus.

So what is the status of “Holmes is dead” in 1893, and in 1903? If we wish to retain an alethic semantics for fiction that adheres to the true/false dichotomy (which is in-and-of-itself compatible with the plausible thought that some claims are indeterminate-in-the-fiction), then there seem to be a limited number of options:

"1. 'Holmes is dead' is a false-in-the-fiction in 1893 (and in 1903), because of the story published in 1903.
2. 'Holmes is dead' is true-in-the-fiction in 1893, but false-in-the-fiction in 1903. 
3. 'Holmes is dead' is true (in both 1893 and 1903) relative to the corpus of stories published as of 1893 (i.e. true -in-the-1893-fiction), but is false (in 1903 and beyond) relative to the corpus of stories published as of 1903 (i.e. false-in-the-1903-fiction)."

Option (1) is unattractive for obvious reasons: it requires that facts about what is true-in-the-fiction in 1893 depend on contingent events that occurred in 1903 – in short, we would be faced with something very akin to backwards causation.

Options (2) makes better sense of Betty’s puzzle, since she would be correct in asserting “Holmes is dead” in 1893, but it is unattractive for other reasons: it entails that truth- and falsity-in-the-fiction behave very differently than truth and falsity simpliciter. On this sort of account, Holmes is the sort of beast that can be dead, but later come back to life. One of the reasons we value fiction is that it provides a means to reflect on the nature of our own world and our own lives. If the logic of standard fictions were this different from the logic of the actual world, then it would be mysterious how and why reading such stories could have any lessons to teach us.

Option (3) is a sophisticated sort of relativism, and this solution (or something much like it) is defended by Andrew McGonigal in “Truth, Relativism, and Serial Fiction” (British Journal of Aesthetics 2014). The problem with this view, however, is that it seems somewhat metaphysically excessive: If Holmes is dead in the up-to-1893 fiction, but alive in the up-to-1903 fiction, then the Holmes that appears in the first fiction is a distinct character from the Holmes that appears in the 1903 fiction (since they have incompatible properties). But surely if Betty says “Holmes is dead” in 1893 and “Holmes is alive” in 1903, she is speaking of the same character (regardless of whether her utterances about that character are correct).

(1), (2), and (3) above are probably not the only options, and variations on these themes of various sorts have been explored by Ross Cameron and Ben Caplan as well. But all of these approaches proceed by retaining a broadly alethic approach to fiction, and then attempting to solve puzzles raised by serial fiction (and other, related puzzles) by adopting some specific kind of account of truth (e.g. relativist, contextualist, etc.)

Here I want to suggest a much more radical approach: that we reject the idea that claims are true-in-fiction, or false-in-fiction, altogether. Instead, claims are more-or-less-likely-in-fiction. To motivate this move, consider a variation on the example given above: it is now 1903, but Betty has only read the Holmes stories up to “The Final Solution”. She then utters “Holmes is alive” (assume this is meant to be a sincere report regarding the content of the Holmes fiction, and not a wishful thinking, “I believe in Sherlock” sort of claim). It would be perverse to respond to Betty by saying “You are correct.” Clearly, even though stories detailing Holmes survival have been published by this point, Betty is clearly making some sort of interpretative mistake if she makes this utterance before reading the post-Great Hiatus stories. A better response might be “You have guessed wisely.” But the best response, it seems, would be something along the lines of “You haven’t got evidence for that (yet).”

Compare the following situation: You and I both roll a single die, each of which is hidden under a cup. The higher roll wins. At 2PM the referee, who can peek under the cup, informs us that we rolled different numbers. I can now justifiably utter “The probability that I will win is ½.” At 3PM the referee informs us that I rolled an even number. I now (again, correctly) utter “The probability that I will win is 3/5.” Note that the truth-value of the claim “I will win” did not change between 2PM and 3PM (that claim is already determined, and the referee knows its truth-value) – what changed is the evidence that I have with regard to this claim. If, however, I utter “The probability that I will win is 3/5” before the referee’s second pronouncement, it is unlikely that the referee would say (or merely say) “You are correct”. Instead, he might point out that I have no evidence for that claim (keeping in the mind that the referee, who has seen the dice, might already know that I have lost).

My suggestion is that, in one respect, fiction works similarly to this: Betty has very good evidence for the claim “Holmes is dead” in 1893. As a result, the probability that Holmes is (fictionally) dead is (relative to the evidence Betty has to hand in 1893) high. But reading the 1903 story provides her with conflicting evidence, which significantly lowers the (fictional) probability of the claim “Holmes is dead” (but does not make it definitely false, since Doyle could have written a story in 1913 in which the 1903 story was merely a dream!) Thus, Betty’s utterance of “Holmes is dead” is correct in 1893 (since exceedingly likely given the evidence she has), correct in 1903 if she has not read the 1903 story (for the same reason), and incorrect in 1903 if she has read the 1903 story (since the new evidence cancels out the old).

Notice, however, that unlike the dice example, there is no independently existing reality (corresponding to the dice) against which these probabilities are to be compared. All we have is the evidence, in the form of claims asserted (or depicted, in comics, films, etc.) within the fiction. In short, with fiction all we have to go on, and the only grounds available to support the objective correctness of assertions like “Holmes is dead”, is the (defeasible) evidence provided by the text (film, etc.) itself.

If this is right, then fictional ‘objectivity’ is more like (subjective) probability than it is like (bivalent) truth – new evidence, in the form of new installments to the story, can increase or decrease our commitment to particular statements in the fiction (and should do so, in rule-governed and predictable ways). Further, no statement is completely-true-in-the-fiction, or completely-false-in-the-fiction (i.e. no statement ever has probability 0 or probability 1), since any statement whatsoever (including mathematical and logical statements – consider Graham Priest’s Sylvan’s Box) is probabilistically hostage to the possibility of countervailing evidence in future installments, sequels, or related works (a good example: consider the modal and psychological claims one is tempted to make regarding Frank L. Baum’s Dorothy before and after reading Gregory McGuire’s Wicked).

Of course, for those of us (like myself) who ascribe to a broadly Waltonian account of fiction, where fictions are games whose rules prescribe what we are to make-believe regarding the characters, objects, locations, and events described in the fiction, all of this shouldn’t be all that surprising. It is widely (although not universally) accepted within epistemological circles that belief is a matter of degree. Thus, it seems natural to think that make-belief is also analog rather than digital, and that when experiencing and evaluating fictions, we do not make-believe that certain claims are definitively and determinately true, or definitively and determinately false, but rather (implicitly) assign higher or lower probabilities to claims as new evidence (in the form of new installments, or chapters, or paragraphs, or even individual sentences) come to light – that is, we assign degrees-of-make-belief to various claims in or relevant to the fiction.

Of course, the above isn’t a fully worked out account of a degrees-of-make-belief based probabilistic semantics for fiction. But the basic framework promises to have important applications to a number of interesting issues in the philosophy of fiction, including: unreliable narrators (indications of unreliable narrators are prescriptions to assign uniformly lower degrees-of-make-belief to the relevant assertions), contradictions in fiction (there is nothing incoherent about conflicting pieces of probabilistic evidence), the phenomenology of appreciation (reading a novel involves an ongoing process of revising assigned degrees-of-make-belief in light of new evidence), and pluralism regarding interpretation (there need be no single correct way to coherently ‘balance’ degrees-of-make-belief given a particular body of evidence). Working these ideas out will have to wait for another day, however.


  1. Roy (if I may),

    Thanks for the post. I found it very interesting, especially since I have a soft spot for Margolis's “robust relativism” about interpretation. I'm not quite convinced, though. Here are a few potential problems that come to mind:

    1.) I'm not sure to what extent anyone does or should buy into bivalence for stories in the first place. This is because everyone tends to agree that stories are necessarily incomplete: some of their features are, of necessity, left indeterminate. How many moles does Holmes have? It's some number between 0 and infinity, but we have no way of knowing which (even a probabilistic account won't get us very far with respect to these kinds of facts!).

    But it seems to me that any theory of truth in fiction that accepts that some fictional “facts” are indeterminate has a ready explanation for Betty's problem (and can accomodate your probabilistic leanings too!): when she utters “Holmes is dead” in 1893, she is either (A) expressing an *interpretation* of the story (in which case robust relativism and probabilistic accounts seem entirely appropriate: hers is a perfectly plausible/likely statement), or (B) she is attempting to assign a truth-value to a “fact” that is actually left indeterminate in the story (even if the balance of evidence lies on one side—just like the balance of evidence indicates that Superman has more than a handful of hairs on his head, but it's still fictionally indeterminate [especially because he's an alien!]).

    2.) I'm a little leery of your analysis in paragraph 11 (echoed earlier in paragraph 9). In that paragraph, you make it sound as though the truth (or plausibility, or whatever) of Betty's remark depends *on whether she's read the later stories*: “Thus, Betty’s utterance of “Holmes is dead” is correct in 1893 (since exceedingly likely given the evidence she has), correct in 1903 if she has not read the 1903 story (for the same reason), and incorrect in 1903 if she has read the 1903 story (since the new evidence cancels out the old).” So it's *reading* the 1903 story that grants the conflicting evidence, not *the existence of the 1903 story itself*.

    I'll readily confess that I've got something of an intuitive block here, since I'm more inclined to think that the story's existence matters more to a narrative arc's content than my having read it (unless we're restricting her/our quantification to just the evidence she has/we have—but then, all she's/we're doing is guessing, nothing more). I've something of an externalist bent, I guess.

    But even leaving that aside, I think that the emphasis on what Betty's read generates a more serious consequence: the story Betty reads, and the story I read, are different stories. Our Holmses are different Holmses. And that seems like a hefty burden for a theory of fiction to bear.

    3.) Just a quick set of questions, since I've already abused your time with some hefty paragraphs: how would your probabilistic model handle implicit/background truths (e.g. “the laws of gravity apply”) and carryover content? And, on a related note, what happens to deductive closure and logical consequence?

  2. FROM JOE LEVY (latest victim of my comment-napping blog):

    There is a synergy between probability and fiction, as probabilities are properties of minds and the information available to them (demonstrated by the dice example), rather than properties of the events they regard.

  3. With regard to (1): If the 1893 story leaves the fact of Holmes' death indeterminate, then it seems like every story will leave everything indeterminate (and merely likely) – after all, the story really does indicate with absolutely no ambiguity that Holmes was dead. So construing the objectivity of fiction in terms of likelihood rather than truth seems like the only option if we want any sort of objectivity at all (of course, your first option – relativism – might still be on the table, but I am less attracted to that route for reasons already given – as well as, I must admit – a serious allergy to the “r”-word more generally).

    With regard to (2): I think you have my view exactly right (perhaps I wasn't clear, however): The story provides evidence if it exists, but statements by a particular agent are correct or incorrect relative to the evidence they have. So probabilities (or degrees of make-belief) are relative to what stories we have read (but truths aren't, since there aren't any truths). So it is a relativism of sorts, I guess.

    Of course, responsible agents also in general have a responsibility to obtain better evidence if such is available. Thus we can say that the Betty who hasn't read the 1903 story is correct in asserting Holmes is dead since the evidence she has weighs heavily in that direction, but we can also say that, insofar as the correctness of such claims is important to her, she also has a responsibility to read the 1903 story in order to obtain the additional available evidence with regard to her claim.

    I am not sure why we need to have read about different Holmes, based on the fact that we have read different stories, and thus assign different probabilities to different claims. We would work in different science labs, and obtain different evidence, with regard to the atomic weight of some element, but we could still be talking about the same element.

    With regard to (3): I have no idea how implicit or background truths, etc. work. Is it true that the standard laws of gravity apply in the Holmes stories? Or merely very likely? This strikes me as one of the most important questions about fiction generally (especially since I tend to think about hard cases like science fiction and superhero comics).

    One thing that is clear to me is that the logic of fictions can be different from the actual correct logic (or logics, for us pluralists!) But I don't have anything insightful to say beyond that. I am particularly worried about how different such logics can be (I think that excluded middle is up for grabs, for example, but probably not and-introduction. But I need to think about this more).

    Apologies for the late replies!

  4. Roy,

    Thanks for your response to my wall of text (and apologies for my own delayed response). If you'll permit a (smaller) wall:

    On (1): I just re-read The Final Problem, and I have to say that I don't think Holmes's death is as determinate a fact as you seem to think. The story unambiguously leads us to that conclusion through Watson's eyes and inferences, so it's true that Watson believes Holmes is dead. But the only real evidence are a “suicide” note and a set of footprints that don't double back. It's pretty clearly indicated that Holmes and Moriarty go over the fall, but that's where the evidence ends (there are no corpses, etc.). So it still seems as though all the usual theories of fiction have a readymade explanation for Betty's utterance, which is just that we can't say whether Holmes is alive or dead, but he seems dead (and Watson and the other characters believe he's dead).

    The question of how much ends up actually being determinate is a different one, and I think a very interesting one. Now, your approach has a neat explanation for indeterminacy built into it, since the result seems to be that nothing in a fiction is 100% determinate. But if the point goes to you, it looks to me like it goes to you over *that* fact, not over the explanation of Betty's utterance.

    On (2), thanks for the clarification. It seems to me like we've read about different Holmeses because their properties, adjusted for individual credence, end up being different. For individual properties, the differences are pretty minimal (e.g. 60% vs. 50% credence for right-handedness). But the accumulation seems like it'll total out to something more significant, like the difference between Harry Potter and Percy Jackson. I guess vagueness looms. It just seems like this generates some serious problems for reference unless we're each referring to a separate entity whose properties track our beliefs, or unless we think Holmes is a natural kind of some sort–which would seem to stretch natural kinds, and which would leave us on the wrong side of Putnam and twin-earth. Maybe we can go for a functional description of Holmes-characters instead, but then I have a harder time seeing how to preserve the initial probabilism/relativism. From reading the same story, we'd have read about characters as different from one another as Doyle's Holmes is from Benedict Cumberbatch's. In any case, it seems like this approach shares the same kinds of problems with reference that possibilism has, but I'm not sure that possibilism's solutions/workarounds are particularly plausible in this instance.

  5. Hi Roy,

    Given the way that you set things up, a relativist view of the kind I described will actually embrace your (1), not (3) I think. You've made the corrective 1903 episode salient in our context, and so when we assess “Holmes is dead in the 1893 story”, we should say that that is false, since it is false-relative-to-our-current-context-of-assessment, that selects the longer narrative which has the two temporally separated episodes (1893 and 1903) as non-conflicting parts. I don't see that that commits us to any backwards causation though, any more than thinking that a proposition can be true-relative-to-world-A and false-relative-to-world-B commits you to 'transworld causation' driving the transworld pattern of change. We can also truly say that English utterances of “Holmes is dead”, as assessed from an 1893-perspective, are true, that “The same character appears in succeeding episodes without being at once dead and alive” is true simpliciter when assessed from our context, etc. There does seem something weird about the view – something to do with the way that it puts pressure on the idea of a claim being made *about* the fiction – but I'm just not sure how much weight to give pre-theoretic notions of aboutness when we're trying to best systematize our messy storytelling practices. I do have sympathy with the worry though.

Leave a Reply

Required fields are marked *.