The final nail

Going back to at least G.E. Moore, philosophers have steadily built a strong case that the social practice of assertion is constituted by a knowledge norm: you should assert Q only if you know Q is true. As best I can tell, this “knowledge account,” as it’s often called, is by now the best supported hypothesis in contemporary epistemology. It teaches us something deep and important about an absolutely central aspect of our lives as social beings. It’s a hard-won discovery that illustrates philosophy’s value, exemplifies genuine philosophical progress, and is something we as a discipline can be proud of.

Whenever I explain the knowledge account as part of a presentation or in conversation, the primary — indeed, pretty much the only — objection I hear is that it’s counterintuitive and, by way of illustration, purported counterexamples are adduced. The examples feature subjects who assert false but well justified beliefs. This is usually followed by the proposal that justified belief, not knowledge, is the norm of assertion. I’m also often told that this is the “ordinary” or “commonsense” view and that it’s a foible of epistemologists to place knowledge, or anything other factive state, into that role.

In a new paper to appear in Cognition, I report a series of experimental studies which strongly suggest that these objections and examples do not reflect the commonsense view. Instead, just the opposite seems to be true: the ordinary view is that we should not assert false claims, even ones extremely well supported by the evidence. By contrast, asserting well justified true claims is perfectly fine.

Interestingly, it turns out that resistance to a factive account of assertion’s norm is at least partly driven by a general psychological tendency that many people have to excuse blameless transgressions by (falsely) denying that any transgression has occurred.


The final nail — 29 Comments

  1. I.

    John, “should” is said in many ways. And *that* I think is what is really at issue here: What kind of “should” is evidenced here for the K-norm. There’s no question that Monday morning surveys will show that the quarter back “shouldn’t have thrown deep” because the receiver was out of bounds. But this “best outcome should” just isn’t relevant to people asking a different kind of question about what the QB should have done. I.e. is he a good QB. That question can only be answered by adopting the perspective of the QB. So if he couldn’t see that the receiver was out of bounds and he looked wide-open, he did the right thing, he did what he should have done (the thing with the greatest *expected* utility, not necessarily the best outcome).

    And the kind of What should I believe and What should I assert questions I am interested in at the intersection of linquistics and epistemology is like an expected utility investigation, not a best outcome kind of investigation.



    Another kind of question about the “should” here is whether it is a “constitutive” norm or not and whether we should care about this kind of norm. For example, suppose the K-norm of assertion is correct (this assumes we’ve settled the question of what it even means) and it is “constitutive” of the act of assertion. Well, fine, then I don’t care about assertion as such any more than I care about knowledge as such. WRT belief, I’ll just stay focused on the probability of propositions I’m interested in and I’ll “bassert” propositions when the probability is sufficiently high. I can live with whatever “criticism” I’m liable to in “violating” your “norm”.


    I hope you won’t see II.A as mere flippancy or ball-go-home-ery. Their is a certain conventionalism that runs from Carnap to Jeffrey which notes that this stuff is up to us. These norms, if such norms their be, are not like moral norms, they have no intrinsic per se force. Unlike moral norms I can say to “constitutive norms” (if such there be) “I just don’t care.”


    All that I care about in this domain can be accounted for by inductive logic and standard norms of morality and practical reasoning.

  2. III.

    Unless and until I am confident that the people in the surveys have thought about these issues a bit, their raw “intuitions” aren’t going to count much for me and my ilk. I think intuitions count as evidence in the context of a reflective equilibrium. How the folk talk, by itself, is of little interest to philosophers as such. Give me the chance to interview some of your subjects, then we’ll talk! 😉

  3. Hi Trent,

    I acknowledge (here and in the paper) that ‘should’ can be read in multiple ways. I took some steps to show that responses aren’t keyed primarily to the perceived costs of speaking falsely or to damaging certain sorts of relationships. I’ve tried not to claim false precision about what sort of standard is embodied in the norm (e.g. perfect vs imperfect duty, asserting permissibly vs asserting well). As of now, the evidence arguably doesn’t discriminate among some importantly different ways in which knowledge norms assertion, even if knowledge does norm assertion. (I discuss this in “Knowledge and Suberogatory Assertion.”)

    It sounds like we’re just interested in very different projects. I’m interested in the actual practice and its norms. You don’t care about the actual practice or my project (at least, if the actual practice turns out to be a certain way), and you’ve got something different in mind that you do care about. I’m glad you’re working on that, because I’m sure that we’ll be able to learn interesting things about it from your work.


  4. Hey, Trent. Sorry, I missed your second comment while I was composing my first. Reflective equilibrium is good stuff, I agree. A couple of the experiments in the paper actually followed up with people and asked them to identify what principles guided their evaluation of the assertion. Their answer to this question correlated very well with their first-order judgments.

    Btw, I’m a philosopher and, as a philosopher, I care about, among other things, how people talk and think about important categories. I don’t claim that philosophers as such need to care about such things, and I can see value in philosophical projects that don’t attend to ordinary thought and talk. But –– and I don’t take this to be in disagreement with anything you said –– it seems to me a great mistake to suppose that the habits and principles of ordinary thought don’t deeply and irreversibly shape how philosophers think, even in reflective equilibrium.

  5. To date, I don’t think I’ve seen any argument at all in print from the claim that knowledge is *constitutive* of assertion. My impression is a lot of people are still rather unsure what that even means. What’s the strong case for that supposed to be?

  6. John, I don’t care about unreflective ordinary language, and I don’t think it reflects anything very deep (other than how bad our educational system is), but I care about you and thus your project! 🙂

    I hope it was clear that I find your work fascinating on this, and I wish you were on facebook, because about a month or so ago I posted abut how great your work was.

    It’s just that, like Aidan, I’m still waiting for clarification, some 13 years later, as to what a “constitutive” norm really is. And so even though, yes, I think we’re principally interested in different projects (let a thousand flowers bloom), I do think there is *considerable* uncertainty about the nature of the norm and about it’s “status”.

    This is a sort of “value-driven” question, to use Jon’s term. Why should I care about these norms, whatever they are? Why should they matter to me? Especially since I don’t think knowledge is important. Why think the norms these people are allegedly judging by have a stronger status than regional rules of etiquette?

    There are some places I’ll get criticized for putting my elbows on the table and some places I’ll get criticized for making assertions of a well-justified false belief. I don’t want to hang out in either place!

  7. Hi Trent,

    The paper contains an actual argument for why the judgments of competent speakers matter on this question (my question, at least).

    It’s actually pretty clear what a constitutive rule is. It’s a rule that helps to individuate the practice (activity, etc.) — that is, helps make it the practice that it is. This category is at least as clear and familiar as the categories of moral norms or norms of practical reasoning (to take two examples you cited earlier). For my part, I suffer no uncertainty about this.

    Rather than saying why you specifically should care, I’ll say this: the average person should care about the rule because they should care about correctly engaging in the practice of assertion, which benefits them and the people they care about.

    Whether someone should be criticized for breaking the rule is a separate question from what the rule is and whether it has been broken, and my results show that these questions are readily and reliably recognized as distinct. Basically everyone agrees that a justified false assertion doesn’t merit criticism or blame (so there’s no need for you to feel uncomfortable hanging around these parts). Despite that, most people still judge that such assertions should not be made.


  8. No problem, Aidan.

    I don’t think I’ve put this into print yet, but here’s an argument I’ve advanced many times in conversation.

    It’s clear that knowledge norms assertion. What sort of normativity is this? It’s neither moral, nor epistemic, nor prudential, nor legal, nor aesthetic, nor a matter of etiquette. What’s left? Presumptive answer: constitutive.

    The weakest link here is the claim that it isn’t epistemic. I don’t think it’s too hard to see why it isn’t, but at least we’re down to at most two choices. (Note: don’t let the fact that the norm features an epistemic status fool you into thinking that it’s epistemic rather than constitutive.)

    There’s at least one other argument, but I think I’ll save it for the book.

  9. In defense of the knowledge norm of assertion, many philosophers appeal to the distinction between ‘wrong’ and ‘blameworthy’. This distinction is most familiar in the ethics literature, where it is typically the distinction between ‘morally wrong’ and ‘blameworthy’. There, philosophers will often say that what someone did was morally wrong, but not blameworthy because they had such and such justified belief(s). But why do (the relevant) justified beliefs free one from blame? Is this a brute fact about blameworthiness? I doubt it. The most plausible account would seem to be that the relevant justified beliefs free one from blame because they make it rational to do what is in fact morally wrong. If so, then blameless wrongdoing is rational wrongdoing. This notion of ‘rational wrongdoing’ is coherent because the wrongness in question is moral wrongness. But the knowledge norm of assertion is standardly offered as a norm of rationality, not a moral norm. So a blameless violation of the knowledge norm of assertion would be a case of rational irrationality. Something has gone wrong.

  10. Dustin,

    That’s interesting about the ethics literature. But it’s just not true that “the knowledge norm of assertion is standardly offered as a norm of rationality.” It’s put forward as a constitutive rule of assertion. Nothing goes wrong here.

    • It’s being a constitutive norm does not preclude it from being a norm of rationality. As you put it in the paper, the claim that knowledge is the constitutive norm of assertion is the claim that ‘assertion is the unique speech act that has a certain standard of correctness [namely, knowledge]’. This leaves open the question of what sort of correctness is at issue. As far as I can tell, the sort of correctness that most philosophers have in mind is rational correctness. On this reading, the claim that knowledge is the constitutive norm of assertion is the claim that assertion is the unique speech act that has knowledge as its standard of rational correctness. Do you have something else in mind by ‘correctness’?

      • Yes, I do: correct according to the rules of the practice.

        I guess a constitutive rule could also be a norm of rationality, though in this case I don’t think it is. It’s an optional, additional thesis to claim that “assert only if you know” is also a norm of rationality. And if the argument you gave up-thread is on the right track, then there’s compelling reason to not opt for this optional thesis and, other things being equal, to not interpret others as opting for it either.

        • I see. Can I then ask you why you asked the questions you did in your surveys?

          If I want to know participants’ views about the constitutive rules of chess, I ask them things like ‘In chess, can you do such and such?’ I don’t ask them things like ‘In chess, should you do such and such?’ Why then did you go with the latter rather than the former formulations when testing participant views about the constitutive rules of assertion?

          • In other words, it seemed to me more likely that your questions were testing participant views on the rational rules of assertion, rather than the constitutive rules.

          • My reason: the “can” question is just too easily read as a question about ability rather than evaluative status. Of course you can say false things, and of course you can say what you know to be false. This greatly complicates the interpretation of affirmative answers.

            Of course, opting for “should” raises concerns about interpreting negative answers, so this choice is not costless. But it seemed obvious to me how to design follow-up studies that addressed the concerns about interpreting negative answers to the “should” formulation (Exps. 2–4), whereas there was no easy way –– that I could think of, at least –– to address the concerns about interpreting affirmative answers to a “can” formulation.

  11. I thought one of the fascinating things about John’s paper was the discovery of the role of what he calls “Excuse Validation”. For those who haven’t read the entire paper, excuse validation is this seemingly general cognitive bias where the desire to say that someone who engages in blameless and excusable yet objective rule­breaking “doesn’t do anything wrong” can lead to the denial that a rule was actually broken at all.

    There were some worries above about interpreting ‘unreflective’ ordinary language evidence as supporting the factive account. But it looks like if John is right then maybe at least part of what the evidence is suggesting is the existence of this psychological bias. Interestingly, the further claim to be made is that perhaps one major philosophical argument that many have considered and found quite intuitive against the factive account could actually be fuelled by it.

  12. If I may chime in, John, regarding your very interesting paper, I have two questions about Experiment 1A. Did you try asking Question 3, the “should” question, without asking (or without *first* asking) the first two questions. It seems to me that eliciting responses to those questions induces a certain bias favorable to the factive intuition.

    Also, did you try asking the “should” question in the neutral condition (as opposed to the True and the False conditions), that is, without specifying whether the inventory is right or wrong about the Rolex Submariner? I’d be *very* interested to find out how people would respond.

    By the way, although I do say (in the paper you cite) that belief, not knowledge, is the norm on assertion, I immediately add that if (as is plausible) knowledge is the norm on belief, in which case we can derive the knowledge norm on assertion by combining the knowledge norm on belief with the belief norm on assertion.

  13. Hi Kent,

    Good questions!

    Experiment 1B tested the same False story without asking a comprehension question. Once again, 80% of participants said that the well justified but false assertion should not be made. Many of the subsequent studies omit comprehension questions too and I kept getting similar results.

    I didn’t do a neutral condition in this paper. But Wesley Buckwalter and I have done further work where we don’t stipulate whether the proposition (= Q) is true. Instead, we asked people to judge whether Q is true, based on the evidence provided. In this particular study, we didn’t asked specifically whether the protagonist should “say” that Q is true; instead we asked whether the protagonist should “write” in her report that Q is true — which, I would think, is in the ballpark of counting as an assertion. Anyway, it turns out that in this neutral context participants’ truth-judgments were predictive of their answer to the “should” question.

    Interestingly, in that same study I was just mentioning, the strongest predictor of the “should” judgment was, *by far*, their judgment about whether the protagonist *knows Q*. By contrast, participants’ judgments about whether the protagonist believes Q were not predictive of their “should” judgment.

    I always thought it was ingenious to propose that belief was the norm of assertion, and knowledge in turn the norm of belief. I discuss your view whenever I teach a module on norms of assertion.

  14. Thanks for the quick response, John. Just to elaborate on one thing, the reason I thought it would be interesting to test the neutral condition is this. Suppose it turns outs that some (many? most?) people think it would be OK for Maria (in that example) to assert in the neutral condition that she has a 1990 Rolex Submariner. These subjects could then be told that after consulting her inventory Maria she finds that she doesn’t have such a Rolex, and you could ask them if she still should have asserted (i.e. *before* checking) that does have one. I’d be curious about the result.

    I don’t know how ingenious that little idea was. From my perspective on speech acts in general, the belief condition on asserting is not relevantly different from the regret condition on apologizing or the desire condition on requesting. In fact, I’ve never understood all the fuss about the norm of assertion in particular, considering that there are all sorts of different speech acts, which have central things in common along with their individual differences. Yes, I do recall Williamson maintaining that there was something distinctive about assertion that made it special somehow. But from the aforementioned perspective of mine, I just don’t see it.

  15. Thanks for your further thoughts, Kent. I have not run such a study, but I do wonder what the result would be. What do you think the result would tell us?

    I think assertion has come in for special attention partly because it’s so central to communication and partly just because the researchers who have worked on it are so influential more generally. In any case, I think the norms of other speech acts are interesting too.

  16. My guess is that, given the neutral condition (i.e. where it is not specified whether the information from the inventory list is correct or not), many if not most subjects would say that Maria should assert (rather than not assert) that she has a 1990 Rolex Submariner. But I really have no idea what the distribution among them would be, once they are informed that the list was in error, regarding what she should have said. I suspect, though, that something like excuse validation would come into play for some of them.

    I’m afraid I’ve never gotten a handle on what constitutive norms of assertion or other speech acts are supposed to be, at least if they are needed for a theory of speech acts or are to be otherwise of theoretical interest. But I say that as a beleaguered old speech act theorist, whose theory has no place for constitutive norms (or, for that matter, constitutive rules, in Searle’s sense). And, as a former tournament chess player, I’ve never been keen on the game analogy (or much of anything philosophers have said when they mention chess).

    Finally, you may have noticed that in sec. 6 the paper you cite, “Applying Pragmatics to Epistemology,” I make similar disparaging remarks about the neglect of speech act theory and pragmatics in then recent debates about testimony. But that’s another story!

  17. Kent, illuminating, as always.

    John and Wes,

    “general cognitive bias where the desire to say that someone who engages in blameless and excusable yet objective rule­breaking “doesn’t do anything wrong” can lead to the denial that a rule was actually broken at all.”

    It kind of sounds like you are treating this as a modus tollens on the initial bifurcation. But it seems just as easy to take it as a modus tollens on saying that there *was* no rule broken. That is my pre-pluralist position and my even present inclination. I have strong ought-implies-can inclinations and others which incline me to think the discovery is that there are no such rules, rules that don’t start out as “You should try to …”

    This is kind of how Chisholm’s thoughts evolved on epistemic normativity.

  18. Hi Kent,

    On the yes/no “should” question, my expectation is similar to yours. But I would further predict that participants’ truth-judgments would mediate their answer to the “should” question.

    I’m not sure what would happen, though, upon stipulating further information and asking them whether they want to change their minds.

  19. Hi Trent,

    I think it’s important to be clear on just what exactly people were split on.

    Here is the case:

    “Doreen just had her car serviced and is driving home from the mechanic’s shop. She wants to get home without unnecessary delay, but she does not want to break the speed limit. The speed limit is 55 miles per hour, so she looks down to see how fast she is going. The speedometer says that she is going 55 miles per hour. But the speedometer is wrong, since the mechanic disrupted a setting on the speedometer without anyone noticing. As a result, Doreen is actually driving 60 miles per hour.”

    It is an objective fact of the case that Doreen is going 60 mph in a 55 mph zone. In fact, these were comprehension questions. Then people were asked whether Doreen should be criticized for driving that fast, and whether “there a sense in which it is incorrect for Doreen to drive that speed.”

    Basically everyone answered that she should not be criticized, and roughly half answered that her speed is incorrect in *some* sense.

    But it’s utterly obvious that Doreen’s speed is incorrect in the following sense: she’s breaking the speed limit. She’s not blameworthy for doing so, but she still is breaking the speed limit.

    To top it all off –– and this is not actually reported in this paper on assertion we’re discussing, but it’s in work under review –– if we run the exact same case and ask people if Doreen is “breaking the traffic rules,” we again see the roughly 50/50 split. However, if we ask people if Doreen is “unintentionally breaking the rules,” then virtually everyone answers “yes”!

    So everyone agrees that she unintentionally breaks the rules, but only half of people agree that she breaks the rules. Clearly, half the people who answer the “did she break the rules?” question are doing something other than competently & literally answering the question. For it’s impossible that she unintentionally breaks the rules without breaking the rules.

    Participant response to analogous cases involving blameless false assertions turn out very similarly.

Leave a Reply

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