Relativism about Epistemic Modals: Some Experimental Data

An expert is talking on the TV news about a mafia boss who just disappeared. There is a lot of evidence suggesting that the mafia boss is dead, but the expert is not sure. He says, ‘The mafia boss might be dead.’ As it happens, though, the mafia boss is not dead at all. In fact, he is watching TV right then, listening to the expert speak. In a case like this, what should the mafia boss say about the expert’s claim? Should he say ‘What the expert said is true,’ or should he say ‘What the expert said is false’?

This is a paradigm example of the sort of case — a so-called ‘eavesdropper case’ — that has led to a revolution in recent theories about epistemic modals. It is widely thought that the intuitive response in this case would be to say that the mafia boss should say that what the expert said was false, but traditional theories of epistemic modals seems to predict that the mafia boss should say that what the expert said was true. This thought is one of the main sources of motivation behind the idea that we might need to turn to a radically different form of semantic theory involving what has come to be called ‘relativism.’

In this brief manuscript, we provide new evidence that calls into question the need for a theory of this type. Specifically, we ran five experiments exploring people’s intuitions about precisely the kinds of cases that were supposed to motivate the theory, but when people received these cases, they simply did not have the intuitions that had been adduced in support of relativist theories. For example, in the case described above, when participants were asked how they would react if the mafia boss said that what the expert said was false, the overall tendency was for people to strongly disagree.

We would love to hear any thoughts you all might have about these experiments or about the larger philosophical and semantic issues. (Feel free to write in with comments even if you haven’t read the actual paper.)

Joshua Knobe and Seth Yalcin


Relativism about Epistemic Modals: Some Experimental Data — 6 Comments

  1. I’ve only had time for the briskest of skims. But here is a quick thought.

    Since the participants do not have the distinction between metaphysical and epistemic modality down, it might be hypothesized that some of them take the ‘might’ to denote alethic, rather than epistemic, modality. Perhaps such an interpretation is even encouraged by including, as you did, an athletic term in the question – “What expert B said is false.” At any rate, the hypothesis would explain the disagreement with (Modal-False). So, maybe it’d be worth controlling for.

  2. The expert also might have said ‘the mafia boss might be alive’. The mafia boss watching TV would always say the expert had the right of it. The phrasing could reasonably be assumed to nuance “might be dead” as countering the certainty of the expert’s audience conclusion of _very likely dead_. But the expert can accomplish the same end by using “might be alive”. Life and death are on/off like a light switch so _might_ is the right verb in the case of uncertainty. However, I think most often the expert would express his caution by employing “might be dead” to correct premature popular conclusions that the mafia boss was still alive; so I think this thought experiment is contrived unless someone can show that “might be dead” and “might be alive” don’t describe the same uncertain reality, so that both phrases don’t mean the same thing. I don’t see the conundrum for the mafia boss if the phrase “might be alive” is used and I think it means the same as “might be dead”.

    There used to be a game show with three doors. The host would explain to the contestants that the grand prize might be behind Door #1 or it might be behind Door #2 or it might be behind Door #3. Because the game show host didn’t reveal that he actually already knew which door the grand prize was behind, doesn’t mean that he illegally used the word “might” when describing the situation to the contestants because he knew which door the grand prize was actually behind so that there wasn’t any “might” involved in describing the two empty (or trivial prize) doors.

  3. When asked the question, “what should the Mafia boss say”, implicit in the question is, “from whose perspective should we interpret the world”. Second, if we take “truth” to mean, “the objective representation of reality from the perspective of that reality’s inhabitants”, then the interpretation, “the boss might be dead” was truthful, based on the evidence the expert accumulated.
    If the mafia boss was also a philosophy student, and accepted my proposed definition of truth, then he should agree that the expert’s statement was true…acknowledging that he is answering the question by looking at reality from the perspective of the expert.

  4. How about doing cases where the subject is unambiguously in the knowing context. For instance, ask if what an expert who said a long time ago that cigarettes might not cause cancer spoke truly. Or do an experiment involving live participants where one is actually in the eavesdropper position.

  5. I had always taken it for granted that, for epistemic uses of ‘might’, the claims “It might be that p” and “For all I know, it might be that p” mean the same and are inter-substitutable. (Is that naive? – the latter seems a bit weaker as a pragmatic matter, in that it makes the speaker sound more uncertain, but that’s the only difference I can see.) If that is correct, then having the expert instead say “For all I know, the mafia boss might be dead” should merely clarify the case, but I can’t see anyone thinking the boss should hold the expert to have said something false then.

    (The gameshow example in (2) above leads me to think that sometimes it should be a ‘for all you know’ instead of ‘for all I know’, though only in a context where the speaker is taking on the listener’s perspective to explain how they ought to be reasoning.)

  6. Thanks for all these comments! Just a few super-quick comments in reply.


    The approach to the semantics of epistemic modals you outline here is a version of the classic contextualist theory. Defenders of relativist views then challenged this contextualist theory by saying that it fails to correctly capture people’s intuitions in ‘eavesdropper’ cases like the one about the mafia boss described above. Just as you suggest, the results of our experiments show that intuitions about these sorts of cases actually don’t provide any reason to reject the classic contextualist theory.


    I definitely see the worry about distinguishing epistemic modals from modals of other flavors, but we had been thinking that this worry wouldn’t arise for the specific sentence used in the study. For example, if participants had received the sentence ‘The mafia boss has to be in his office,’ they might have interpreted the sentence either as an epistemic modal or as a deontic modal. By contrast, we were thinking that there is really only one reading of ‘The mafia boss might be dead,’ and that is as an epistemic modal.

Leave a Reply

Your email address will not be published. Required fields are marked *