New empirical studies on epistemic contextualism

Epistemic contextualism is the view that the verb “know” is a context sensitive expression. As a first approximation, epistemic contextualism states that in order for us to truthfully say a person “knows” a proposition, that person must meet the standards set by our context and, critically, the standards change across contexts. The variation is thought to be theoretically important partly because it might indicate an ingredient of (the truth conditions of) “knowledge” statements beyond the traditional factors of belief, evidence, and truth.

Contextualists motivate their view based on a set of empirical claims about competent speakers’ linguistic behavior in certain situations. A famous way of illustrating the idea involves a pair of cases about a man who wants to deposit a check and is deciding whether to wait in a long line at the bank on a Friday afternoon, or come back on Saturday morning when the line would be short. But the question arises: is this bank actually open Saturday morning? The man visited this bank two Saturdays ago and it was open then, but banks do sometimes change their hours. In the “low stakes” version of the case, nothing serious hinges on whether he deposits the check before the weekend is over, and the man says, “I know that the bank is open tomorrow.” In the “high stakes” version of the case, something very serious hinges on whether he deposits the check before the weekend is over, and the man says, “I don’t know that the bank is open tomorrow.”

Contextualists claim that competent speakers will judge that the man truthfully says he “knows” in the low stakes version, and that the man truthfully says he “doesn’t know” in the high stakes version.

Do people behave as contextualists predict? Prior research on this empirical question has yielded mixed results. Taking into account methodological objections raised by contextualists,* I ran another series of studies to investigate the issue.** I found that, just as contextualists predicted, people judged that the man truthfully says he “knows” in the low stakes case, and that the man truthfully says he “doesn’t know” in the high-stakes case.


However, I also found two other things that make it very difficult to interpret this as evidence for contextualism.

On the one hand, there is the general shifting problem: switching from low to high stakes also significantly decreased belief attributions, the evaluation of the agent’s evidence, and people’s confidence that the bank was open tomorrow. Thus, the contextualist hypothesis is not needed to explain the observed pattern in knowledge judgments: it could easily be explained by changes in underlying judgments about belief, evidence, and truth. This confirms some suspicions voiced by some philosophers over the years (e.g. Kent Bach and Jennifer Nagel).

On the other hand, there is the deferral confound. Even if nothing about stakes or error possibilities is mentioned, people tend to agree that the agent speaks truthfully when he says “I know,” and they also tend to agree when he says “I don’t know.” In short, people tend to defer to another person’s self-regarding knowledge statement. (Some evidence for this might also be gleaned from earlier work by Wesley Buckwalter and Nat Hansen & Emmanuel Chemla.)*** Unfortunately, this low-level agreement bias has been obscured because contextualist test cases are multiply confounded.

To illustrate the deferral confound, consider a pair of cases I tested, which involve not a low/high manipulation but rather a yes/no manipulation:

(Yes/No) Keith and his wife Jane are driving home from work on Friday afternoon. They just received a check from a client, which Keith plans to deposit in their bank account. As they drive past the bank, they see that the lines inside are very long. Keith says, “I hate waiting in line. I’ll just come back tomorrow morning instead.” Jane asks, “Do you know that our bank is open tomorrow?” Keith answers, “It was two Saturdays ago that I went to our bank, and it was open. So, [yes, I do/no, I don’t] know that our bank is open tomorrow.”

People tended to agree with the agent in both cases. Thus manipulating stakes (low/high) is not needed to produce the basic pattern in knowledge judgments that contextualists have focused on. Moreover, a follow-up study revealed that when the deferral confound is removed, manipulating stakes does not produce the relevant pattern in knowledge judgments: instead, people tend to (meta-linguistically) attribute knowledge in both low and high stakes cases (i.e. they say that the man should say “I know” in order to speak truthfully).

Based on these findings — all of which were replicated across multiple cover stories — I conclude that the principal extant motivation for contextualism fails. Contextualists still owe us a distinguishing prediction of their view, something we would confidently expect only if contextualism were true, or which contextualism seems uniquely suited to explain. Absent that, contextualism is an idle hypothesis and we should not accept it.

Of course, it is consistent with these findings that such a prediction will be made and vindicated. That would be an interesting development! But, moving forward, hopefully the theoretical debate will not get too far out ahead of available evidence. The fact that experimental epistemology is now a firmly established and growing interdisciplinary field will likely help in this regard.

* DeRose, K. (2011). Contextualism, contrastivism, and X-Phi surveys. Philosophical Studies, 156(1), 81–110.

** Turri, J. (in press). Epistemic contextualism: an idle hypothesis. Australasian Journal of Philosophy.

*** For discussion, see Buckwalter, W. (in press). Epistemic contextualism and linguistic behavior. In Ichikawa, J. J. (Ed.), Handbook of Epistemic Contextualism. Routledge.


New empirical studies on epistemic contextualism — 5 Comments

  1. Contextualists have been on to the pressure to interpret speakers’ claims using context-sensitive terms as true. We think that’s to be expected where context-sensitive terms are being used, due to the operation of accommodation. I’ve called accommodation “an important and powerful aide” to my case for contextualism (p. 88 of the paper Turri marks *, but referring back to my earlier work, and also to Lewis’s “Scorekeeping,” which is where this use of “accommodation” comes from). I write:
    “I take accommodation to be a very important force in the interpretation of context-sensitive language. In fact, I think this force is often responsible for speakers being in a situation in which they’ll be asserting a truth whether they assert either a certain sentence featuring a context-sensitive term or that sentence’s negation! So, for instance, there may be situations in which you and I are both looking at Tom, and we both have a pretty accurate idea of just how tall Tom is, and there’s no particularly strong conversational forces militating either for standards that Tom meets for what counts as “tall” or for standards he doesn’t satisfy, and I will be speaking truthfully if I say to you “Tom is tall” or if I say “Tom is not tall,” because an appropriate rule of accommodation will have it that the standards for “tall” governing our conversation will adjust to make what I say to you count as true.” *, p. 88.
    But it’s just Lewis’s old observation that it isn’t as easy as one might think to say something false, where context-sensitivity is involved.

    • Hi Keith,

      Thanks for your reply and for including those quotes and references. (In the paper, I also cite Lewis.)

      You say that contextualists think that accommodation is “to be expected where context-sensitive terms are being used.” But agreement bias and default trust in testimony are robust tendencies that occur regardless of whether the statement features context-sensitive terms. So why do you take agreement as evidence of context-sensitivity?

  2. “Moreover, a follow-up study revealed that when the deferral confound is removed, manipulating stakes does not produce the relevant pattern in knowledge judgments”

    On that, the relevant part of * is pp. 89-91

  3. The rules for the proper use of the English verb ‘know’, as with practically all the lexical categories of English and every other human natural language, are context sensitive, so I’m not disputing the importance of the notion of context sensitivity. And this goes for the objectlanguage use of the verb ‘know’ by “a man”/Keith in the example given, as well as the “metalinguistic” use of it by the observer offering a critique of the sentence in question.

    But it strikes me that the use of the verb ‘know’ by the objectlanguage speaker in the above example is unnatural and artificial. First of all, in sentences of the type “I know (that) S”, it is only the embedded clause S that is meant to be judged true or false (as a description of the world). ‘know’ is a factive verb, which means that its proper use presupposes the truth of the embedded clause object. For the main clause first person subj. – verb expression “I know”, the relevant criterion is first of all sincerity, not truth: if the “man” is aware that he does not believe that the bank will be open, and if it in fact will not be, he could be accused by the critic of lying. If he sincerely believes that the bank will be open and it in fact will not be, then we would say he is mistaken (an epistemic status that deserves more attention).

    But the example given seems artificial, because the objectlanguage speaker probably would not use the verb ‘know’ to express their epistemic state in that situation. In the “low-stakes” version, the man might say something like, “I’m pretty sure this bank opens on Saturdays, because I was here two Saturdays ago, and it was open.” (I changed the time reference of the embedded clause from future to present (generic) to make the truth- judgment clearer.) And in the high- stakes version, something like, “I think the bank is open on Saturdays, but I’m not absolutely sure. I’d better ask.” ‘think’ is a non-factive verb whose use does not presuppose the truth of the embedded clause object. Speakers are able to indicate their degree of certainty about the epistemic status of their beliefs or what they think of as their knowledge of the world, and which expression of which status is considered relevant would vary with the context; this is the function of the grammatical category of modality, for example. There are other contexts where ‘know’ would be natural: (On Saturday) wife says to man, “Why didn’t you deposit the check? Don’t you know that the bank is open on Saturdays?” Man says, “Yes, I know the bank opens on Saturdays, but I didn’t notice the time/I forgot [the fact, or that this was Saturday, etc.].” Here the fact that the bank opens on Saturdays is not in doubt, and the wife attributes (apparently correctly) the relevant level of “knowing” to the husband.

    More generally, not all sentences are meant to be judged true or false; there are several other types of declarative sentences with different criteria for their critical evaluation. Anscombe said as much quite some time ago, but it seems that nobody wanted to listen to her. Taking “truth” as the criterion for the critical judgment of all sentences will not work out. Furthermore, considering “truth” as an inherent property of sentences is also a mistake. “Truth” judgments, I would say, belong to the pragmatic level of analysis, not the semantic. Action and critique of action; and that will work for formal languages as well as natural languages. The unsolved problem for the quest to understand the possibility of meaning in natural language is the general relation of reference. Anyway, that’s what I propose or suggest; it’s a possible view that seems to work out. (I’m not sure philosophers would accept this last paragraph, as these are some of the problems that I’ve noticed in the approach to the understanding of meaning taken in the philosophy of language.)

Leave a Reply

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