Testimonial Knowledge and Linguistic Competence

I am always looking for ways to connect epistemology to other subfields of philosophy (esp. within the “M” side of M&E). Here’s one I’ve been thinking of for some time.

If testimonial belief is to be knowledge, it must satisfy the anti-luck condition on knowledge. There are, it seems, (at least) three sources for luck to enter into the picture in a typical testimonial exchange: the testimony itself, though true, might be only luckily so (as when based on a guess); the recipient might be lucky in having accepted what in fact is solid testimony (as when one just happens upon the one truth-teller in a room full of expert liars); or the recipient might be lucky in having recovered the attested proposition (as when one didn’t really catch what was said, but guessed correctly). Of these the first two have been explored at length in the literature on the epistemology of testimony; the third not so.

(more: click on p 2 below)


Testimonial Knowledge and Linguistic Competence — 8 Comments

  1. Hi Sandy,
    I agree with you about the general point that the criteria of (at least) epistemic success ought to be more relaxed than they are when people insist on the very same proposition being transmitted. (I don’t know about all other kinds of success, though this seems to be a crucial one anyway.) Let us take the following scenario, in which a speaker S asserts ‘p or q’ (thereby testifying to it being the case that ‘p or q’) to a hearer H in response to H’s question. Let us say the conversation goes like this:
    H: Where is the necklace?
    S: Either it is in the safe or it is on the dresser.
    S always answers questions like this in the following way. When (and only when) he knows the answer, he produces a true assertion that is not maximally informative, since the first disjunct is true and the second is false (the necklace is in fact in the safe, and S knows it). When S does not know the answer, S says that he does not know. H does not know about this communicative peculiarity of S. Let us consider H’s epistemic standing in the following five cases.
    Case A: H understands the proposition being asserted by S perfectly well. Presumably, H comes to know that either the necklace is in the safe or on the dresser. (I wonder if anyone doubts that this could be a case of knowledge by testimony, assuming all the other criteria are met” … )
    Case B: H accidentally mishears S as saying ‘Either it is in the safe or it is on the safe’ (‘p or p’). H assumes that S either misspoke or hesitated about what he was going to say in the middle of the sentence, or something like that. In any case, H acquires the belief that the necklace is in the safe. Since H is normally a reliable hearer, had H not misheard S, he almost certainly would have acquired the belief that either the necklace is in the safe or it is on the dresser, which would have been an instance of knowledge (if H knows this proposition in Case A).
    Case C: H systematically fails to pay attention to the content of the second disjunct of S answers, though not because he believes that it ought to be ignored or has been conditioned to ignore it. This failure is a product of a weird cognitive defect on the part of H, which manifests itself only in his interaction with S. This failure does not affect H’s understanding of the content of the utterances of others. In this case, H only processes the first disjunct of the utterance, and so acquires the belief that the necklace is in the safe, where it is indeed.
    Case D: H accidentally mishears S as saying ‘Either it is in the safe or it is on the desktop’ (‘p or r’). H thereby acquires a true belief, which is, moreover, safely true (assuming that a disjunctive belief is at least as safe as the safest of its disjunts). Assuming that whatever epistemic failure is responsible for H’s mishearing only affects the reception of second part of the utterance, we may be inclined to conclude that S has acquired a bit of knowledge that is testimonial (or, at the very least, quasi-testimonial or testimonial enough, or something like that).
    Case E: H hears S’s whole utterance and understands its first part perfectly well. However, H does not know what a dresser is, and therefore is incapable of grasping the proposition asserted by S. Instead, he acquires the belief that the necklace could well be in the safe (or some belief like that). Note that this belief is true and possesses many other structural and functional epistemic desiderata.
    I take these cases to suggest that the requirement that the exact proposition be transmitted is probably too strict. It seems clear that there is some sense in which ‘p or q’ is “not far off” from ‘p’, though it would be interesting to spell it out, especially given that the case can be generalized to include an indefinitely large number of disjuncts.
    We may also imagine cases of the following variety. S frequently gives intentionally false answers to certain questions, but in a non-arbitrary manner. Whenever possible, S systematically replaces phonemes or letters in his answers to make them different from what the true answer would have. Thus, instead of saying ‘Jim bought a cow’, he might say ‘Jim bought a sow’ (or something like that). H, similarly to what happens in Cases B and C above, either accidentally or systematically (but abnormally) undoes the damage done by S to what would have been the proper sentence for expressing the proposition that is the appropriate answer to the question at hand (replacing letter S with C, etc.). It seems implausible to maintain that ‘Jim bought a cow’ is propositionally “not far off” from ‘Jim bought a sow’. It might turn out, therefore, that we need to specify not only the propositional sense of “not far off”, but the morphological one as well.

  2. Pavel,

    These are nice cases to think about.

    Regarding Case A, we are in agreement.

    Cases B, C, D, and E are harder. Part of the problem is unclarity in the core notion I am using: testimonial knowledge. Elsewhere I have tried to characterize this notion as I understand it. I think that there are conditions on both the recipient (hearer) and source (speaker). I try to spell these out in terms of two notions: the notion of epistemic reliance (on the side of the hearer), and the notion of authorization to believe (on the side of the speaker). Epistemic reliance: if my knowledge that p is to be testimonial, its status as knowledge must depend on (my reliance on the existence of) epistemically good-making properties of the testimony itself. (Roughly: I rely on you to have reliably gotten things right in your attestation. This needs qualification, but I won’t bother here.) Authorization: a speaker who testifies that p presents herself as in an epistemically happy position regarding the truth of the attested proposition, and in so doing does something akin to “authorizing” a hearer to accept it on the strength of her epistemic authority. (Elsewhere I try to give an account of this in terms of the norm of assertion.)

    With this in hand we can turn to the other cases.

    Regarding Case B, my concern is that the hearer’s belief is “risky” in a way that the testimony was not. Admittedly, you set up the case so that this sort of “riskiness” cannot be spelled out in terms of the belief’s safety. Still, it strikes me as too risky to count as testimonial knowledge, esp. since the element of risk (coming to believe something equivalent to the detachment of one of the disjuncts) is not sanctioned by the testimony itself. In a nutshell: what the believer believes she was not authorized to believe by the testimony. So I am inclined to say that even if it is knowledge (and I’m not sure it is), it is not TESTIMONIAL knowledge.

    I think Cases C and D might be handled similarly.

    Case E is interesting. My initial reaction is this: either it should be assimilated into the category of accepting testimony in another language one understands (in which case it can count as testimonial knowledge), else it should not be counted as testimonial knowledge at all (even if it is knowledge).

    In any case I agree that these sorts of cases need to be thought about carefully — certainly more carefully than I have done here.

  3. Just a few quick remarks…

    Epistemic reliance: And what about the epistemically bad-making properties of the testimony? It seems that whether my cases count as cases of testimonial knowledge depends in part on how the latter figure in the story…

    Authorization: Whether or not this is satisfied in my cases depends on whether you extend norms of assertion to cover more ground that people often take them to do. In particular, you must include conversational norms that go beyond what Williamson, Lackey, etc., talk about.

    Regarding your analysis of the cases themselves, I just want to suggest that (much like I indicated in my original comment) we might want to distinguish quasi-testimonial and/or partially testimonial knowledge, which could still be such that the correct account of them constrains accounts of linguistic competence.

  4. Is there a difference in scope between testimony in law and testimony in epistemology? That is, does ‘Authorization’ provide a theoretical backing to the practice of challenging testimony delivered by a witness in a court from entering the court’s record? Or, would the category of testimonial knowledge be ill-suited for this purpose, either by being too broad for law or too narrow?

  5. Pavel,

    The epistemic reliance claim — that if my knowledge that p is to be testimonial, its status as knowledge must depend on (my reliance on the existence of) epistemically good-making properties of the testimony itself — is an instance of a more general thesis of epistemic dependence, where the epistemic status of my belief depends (at least in part) on the epistemic properties of your testimony. (If my belief is to be knowledge, then your testimony must have some epistemically good-making features.) So I take your point here.

    Regarding your point about extending the norms of assertion, I think the point is complicated. It may be, as you appear to be suggesting, that a speaker authorizes a hearer to believe propositions that are merely implicated by what she said. This may be. My sense, though, is that if this is so, the content(s) authorized in this way can be derived by appeal to (i) a claim representing the speaker as having authorized belief in the attested content, together with (ii) claims regarding other (common knowledge) features of the context. So my hope would be able to account for these sorts of case without having to read anything more into the norm of assertion beyond what would be held by anyone who thinks there is such a norm (and who thinks that it is epistemic). Admittedly, though, this is just a promissory note at this time.

    Finally, your point about quasi-testimonial knowledge may be correct. It’s just that I want to see what we can get from the “pure” (or “purer”) cases first. But I am not opposed to extending the inquiry.

    Gregory: the question is an interesting one, and I confess that I am no expert regarding testimony and the law. I do know that in his seminal book Tony Coady does talk a bit about the connections (although if memory serves he is talking about British Common Law rather than US law). My initial thought is that the law places some substantial constraints on admissible testimony — constraints that go beyond anything that is imposed on the notion of testimonial knowledge in epistemology. But if details are wanted, it would perhaps be best to consult a legal source on testimony in the law.

  6. Gregory–I imagine I’m close to optimally situated to talk about the interaction of the epistemology of testimony and the law, since I’m a law professor whose dissertation was on the epistemology of testimony, but I’m not sure what you mean by “the practice of challenging testimony delivered by a witness in a court from entering the court’s record.” Could you elaborate a bit?

  7. Chris—There seem to be rules for admitting testimony from witnesses in court that are more stringent than those that feature in epistemology. For instance, I can imagine that an epistemology of testimony would allow for transitivity (i.e., Jones safely passes testimony to Smith, who in turn may safely pass it to Brown) but this admissibility condition would seem to run afoul of rules striking second-hand testimony from entering the court record.

    One could view this type of difference as one where both the law and epistemology agree on what testimonial knowledge is but differ on admissibility conditions. Legal testimony appears in more contentious circumstances than ordinary testimony, after all, which may account for the more stringent rules for admissibility. Or at least on some issues: a law court may admit expert testimony on a scientific matter that would not pass under that science’s system of peer review, for instance.

    My question was whether it made sense to think of the differences between epistemic testimony and legal testimony to be solely to do with admissibility conditions.

    If it is, then I would think that the issues of safety and admissibility that are raised above would need to be heavily tempered by empirical considerations. (My back-up set of examples would be systems of remote sensors, which is the AI analogue to models of witness testimony.)

    If it is not a sensible view, that is, if the difference between legal testimony and testimonial knowledge is a difference in kind, then it would be nice to know what that difference is, whether there are other types of testimony, and how they are thought to relate to each other.

  8. The hearsay rules generally exclude from evidence “a statement, other than one made by the declarant while testifying at the trial or hearing, offered in evidence to prove the truth of the matter asserted.” Federal Rule of Evidence 801(C) [that should be just a C, not a copyright sign] gives that definition, while Rule 802 states the general rule that it is inadmissible. Rule 803 then gives 23 exceptions for which the declarant’s availability is immaterial, and Rule 804 gives 5 more exceptions that apply if the declarant is immaterial. Rule 807 gives a residual exception for evidence “having equivalent circumstantial guarantees of trustworthiness.”

    I’m not sure that these rules are trying to give any sort of definition of testimonial knowledge that might compete with epistemologists’ definitions; the rules are trying to ensure that litigants have a chance to test the reliability of all the evidence that the factfinder considers.

Leave a Reply

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