Skip to main content.
November 26th, 2005

Schaffer on Contextualism and Knows-wh

In the latest Philosophical Studies, Jonathan Schaffer launches a series of objections to Interest Relative Invariantism. I suspect most of these will end up being clashes of intuitions, though maybe I’ll write something more about them later. What I want to focus on here are Jonathan’s arguments concerning knowledge claims involving embedded questions. Jonathan claims these support contextualism, but I don’t really understand how that could be true. What he does seem to raise is a problem about how to understand embedded questions.

Here is a small variant of a case that Jonathan uses.

CIA Headquarters have been burgled. Frank is the thief, though the CIA don’t know that. Indeed, Frank isn’t even on their list of suspects. After an investigation to find out what was stolen, the CIA have discovered that the only thing stolen is the director’s coffee cup.

A and B are having a conversation about the break-in. It is common ground among them that Frank was the thief, and the only thing he stole was the coffee cup. Further, it is common ground that the CIA were doing an investigation into what was stolen. But only A knows that the investigation has completed with the CIA discovering that only the coffee cup was stolen. First consider A making any one of the following claims.

(1) The CIA know that Frank stole the coffee cup.
(2) The CIA know what Frank stole.
(3) The CIA found out what Frank stole.
(4) The CIA discovered what Frank stole.
(5) The CIA were investigating what Frank stole.
(6) The CIA know who stole the coffee cup.

My instinct is that (1) is clearly false, as is (6) but it’s plausible that A could truly utter any of (2) through (5). I don’t think these four are clearly true, but I think they have a true reading. In a context where Frank’s identity as the thief is given, and its importance to the CIA not being considered (because we are talking for now about the investigation into what was stolen) we can bracket the fact that the CIA don’t know Frank is the thief. In other contexts I think it would be natural to say (2) is false.

Similarly, consider B asking one of the following two questions.

(7) Does the CIA know that Frank stole the coffee cup?
(8) Does the CIA know what Frank stole?
(9) Does the CIA know who stole the coffee cup?

The answer to (7) and (9) is clearly no, but the answer to (8) may well be yes. (At least in circumstances where A could truly say (2).) On the other hand, if A was asked (8) by Frank’s lawyer, or his brother, who is worried about Frank’s legal jeopordy, the proper answer may well be no. As we suggested above, if A was talking to these people, it is odd, and arguably false, for her to utter any of (2) through (5).

Jonathan suggests all of this supports some kind of contextualism, and it may well, but it isn’t contextualism about ‘knows’. I think it’s true that (2) to (5) are context-sensitive, but I don’t think that’s because of the verb ‘knows’. If it were, we could generate some context-sensitivity in (1), and we can’t. The real context-sensitivity is in the embedded question. Of course, all context-sensitivity in the sentence shows is that some term in the sentence is context-sensitive. And as I think Jonathan himself shows, the semantic value of the embedded question is context-sensitive.

Having said that, it’s tempting to make the following argument.

  1. If what Frank stole is the coffee cup, then the CIA knows what Frank stole iff the CIA knows that Frank stole the coffee cup.
  2. Whether ‘The CIA knows what Frank stole’ is true is context-sensitive.
  3. The coffee cup is what Frank stole.
  4. ‘The CIA knows what Frank stole’ is true iff the CIA knows what Frank stole.
  5. ‘The CIA knows that Frank stole the coffee up’ is true iff the CIA knows that Frank stole the coffee cup.
  1. So, whether ‘The CIA knows that Frank stole the coffee cup’ is context-sensitive.

That’s a bit rough, but hopefully you get the idea. It’s a very tempting argument, because every premise has at least a good ring of plausibility. But the conclusion is false. And the above example shows why; the very first premise fails in our example. I suspect in any argument for contextualism about knowledge that claims from contextualism about knowledge wh- claims, the equivalent premise linking the two types of knowledge claims will be false. Certainly Jonathan needs a premise like 1 for his argument to work, but as far as I can see he doesn’t defend one, and his very arguments provide reason to think that the premise would be false.

But this all raises a further question. If the CIA don’t need to know that Frank stole the coffee cup in order to know what Frank stole, what do they need to know? My best guess is that the following is true.

‘S knows what N F-ed’, where F is a transitive predicate, iff for some x, x is what N F-ed, and there is some description ‘the D’ denoting N such that S knows that the D F-ed x.

That’s a mouthful, but what it comes to in our case is that the CIA must know that Frank, under some description or other, stole the coffee cup, though they need not know that the description denotes Frank. And they do know that – they know that the thief stole the coffee cup, and ‘the thief’ denotes Frank. Obviously this would need a bit of work in order to get it to a general theory of embedded questions, but I think this is on the right track.

If it is on the right track, it explains where the contextualism comes from. The quantifier over descriptions in the paraphrase above is, like all quantifiers, potentially restricted by contextually supplied parameters. Which kinds of descriptions of N are allowed will be determined by context. When talking to someone who cares particularly about what the CIA knows about Frank, perhaps the only description that matters is ‘is identical with Frank’. In other contexts, descriptions like ‘the thief’ matter as well.

So the big epistemological point is that these cases provide no support for contextualism about knowledge that claims. But the semantic question, about just what the semantic value of embedded questions is seems hard and interesting. I certainly don’t know that my solution here works, but I hope it is on the right track.

Posted by Brian Weatherson in Uncategorized

14 Comments »

This entry was posted on Saturday, November 26th, 2005 at 3:35 pm and is filed under Uncategorized. You can follow any responses to this entry through the comments RSS 2.0 feed. Both comments and pings are currently closed.

14 Responses to “Schaffer on Contextualism and Knows-wh”

  1. Kenny Easwaran says:

    Just for the record on clashes of intuitions, I can hear (1) as being true.

    Consider a context where A and B know that Frank broke in and stole the coffee cup. They also know that he stole the coffee cup because the two of them wanted a sample of the director’s coffee grounds, so they could use this information in a future plot. They both also know that the CIA knows that they are the only two who are interested in the director’s coffee grounds. A and B know that the CIA have know idea who broke in, and don’t really care, but are interested in what was stolen for clues to future plots that might be afoot. B asks A, “Do they know what Frank stole?” In this context, A utters (1) to B, with “Frank” completely unstressed.

    Actually, in order to find (1) true, I only needed the word “Frank” to be completely unstressed and there to be no special reason why the CIA would care about the perpetrator more than the crime. But I think this context might make (1) sound better to others. Of course, (2) through (5) sound progressively better, but even (1) sounds ok to me.

    I don’t know what this says for either of your theories.

  2. Jonathan Schaffer says:

    Hi Brian,
    Thanks for the insightful discussion. Let me try to be clearer about what the argument for contextualism (/contrastivism) was supposed to be!
    WARNING: long post ahead.

    1. Background on the semantics of questions:
    Virtually everyone (you and I included) agrees that wh-clauses are semantically context-variable. Let me try to be a bit more precise about this.

    We should distinguish wh-clauses (in the language) from questions (the abstract entities they denote), just as we distinguish that-clauses from propositions. Questions are partitions on logical space. Each cell is a set of worlds (a proposition), so questions are sets of mutually exclusive propositions. These propositions are the possible answers to the question. So a question is essentially a multiple-choice slate.

    The contextual variability is in the mapping from wh-clauses to partitions. To take a standard example, if a non-academic asks you “Who came to your talk?” then the multiple choice slate might include “only philosophers,” while if I were to ask you the same question you might need to name names.

    The real issue is whether this contextual variability SUFFICES to explain the intuitive contextual variability of knowledge-wh ascriptions, or whether more is needed.

    2. Your argument:
    You do a heroic job of trying to construct an argument for me! But that was nothing like the argument I had in mind.

    In fact, if you think about it, you’ll see that I am committed to DENYING premises 1, 4, and 5 in your argument. I’m committed to denying 1 because of the contextual variability of wh-clauses. And I’m committed to denying 4 and 5 (as is any contextualist) because semantic descent is invalid where there is contextual sensitivity. (For the contextualist, 4 and 5 are like disquoting on indexical utterances containing “I”, while for the contrastivist, 4 and 5 are like disquoting on reduced utterances such as “Jonathan prefers chocolate” which leave the contrast implicit.)

    3. The intended argument:
    The argument that I had in mind is that there can be contextual variability in knowledge-wh ascriptions that seems solely driven by the NON-ACTUAL cells of the partition (the false answers to the question, the contrasts).

    To illustrate, suppose that there is a goldfinch in the garden, and consider the following questions:
    Q1: {there is a goldfinch in the garden, there is a blue jay in the garden}
    Q2: {there is a goldfinch in the garden, there is a canary in the garden}
    Q3: {there is a goldfinch in the garden, there is a goldfinch at the neighbor’s}

    Then the claim is that the following can differ in truth-value (with the interpretation of the wh-clauses fixed by the questions above):
    K1: I know whether there is a goldfinch in the garden, or a blue jay.
    K2: I know whether there is a goldfinch in the garden, or a canary.
    K3: I know whether there is a goldfinch in the garden, or at the neighbor’s.

    That is, just look at K1. The question of whether it is a goldfinch or a blue jay (Q1) is a pretty easy question. You might well think that I know the answer to that one. But now clear your mind and look at K2. The question of whether it is a goldfinch or a canary (Q2) is tricky. You might well think that I don’t know the answer to that one. Likewise, clear your mind and look at K3. The question of whether it is in the garden or at the neighbor’s (Q3) seems incommensurate with Q1 and Q2. The kind of person who would know the answer to Q3 would be the homeowner, while the kind of person who would know the answer to Q2 would be the ornithologist.

    To present the argument more formally:
    1. K1 and K2 can differ in truth-value.
    2. K1 and K2 involve the same subject (me).
    3. K1 and K2 involve the same true proposition (that there is a goldfinch in the garden).
    4. K1 and K2 involve the same time.
    2-4 show that there is no slot in the standard invariantist Ksp (at t) relation that can explain 1. The argument continues:
    5. The best explanation for 1 is that the knowledge relation has an additional slot for the contrast (that there is a blue jay in the garden, vs. that there is a canary in the garden).
    5 motivates the introduction of a CONTRAST slot: s knows that p rather than q, where differences in the q slot drive the intuitive differences in 1. So it is not enough just to recognize that wh-clauses are contextually variable. We also need a contrast slot in the knowledge relation, to pick up on how the wh-clause is contextually variable.

    By the way, this isn’t just an issue with wh-constructions. It seems a GENERAL feature of the language, arising wherever we have mechanisms for encoding contrasts, that the truth-value of knowledge ascriptions is sensitive to these contrasts. That was the idea behind the general pattern of intuitive differences I was reporting with the who/what, whether, rather, focus, cleft, and presupposition pairs.

    Of course I would be very interested to hear what you think of this argument! I hope this won’t just be a clash of intuitions.

    4. Your positive account
    You suggested (as a “best guess”):
    “S knows what N F-ed” where F is a transitive predicate, is true iff for some x, x is what N F-ed, and there is some description ‘the D’ denoting N such that S knows that the D F-ed x.

    I think there are at least three problems with that proposal. The first problem is that we should want a general account of knowledge-wh, not a specific account for “knowledge what N F-ed” for transitive F.

    The second problem is that it is doubtful whether your account will be compositional. I’m assuming that the wh-clause is a constituent of the VP, and that the meaning of the VP is basically composed by computing the meaning of the wh-clause (a partition), and combining it with KNOWS. So I’m assuming that all we can do compositionally is to combine KNOWS with some partition. You seem to want to give KNOWS access to a lot of non-local information that should be buried deep inside the wh-clause.

    The third problem is that the truth conditions are going to be nearly trivial. If Frank indeed stole the coffee cup, then everyone will count as ‘knowing’ what Frank stole, in virtue of knowing the trivial claim that the person who stole the coffee cup stole the coffee cup. So quantifier domain restriction will wind up doing virtually all the work here. But it seems hard to see how it could. If I ask “Who is the person who stole the coffee cup?” then the description “the person who stole the coffee cup” ought to be relevant if any description is. But that won’t make my question trivial (“Ah! It was the person who stole the coffee cup who stole the coffee cup!”) In general, the worry is that if quantifier domain restriction is the only thing preventing triviality, then it ought to be quite easy to generate triviality, when it isn’t.

    In any case, your account is just not going to touch the cases under discussion here. So, still supposing that Frank stole the coffee cup, consider the following pairs:
    Q3: {Frank stole the coffee cup, Frank stole the secret plans}
    Q4: {Frank stole the coffee cup, Frank stole the beer stein}
    Assume that it is very easy to tell the coffee cup from the secret plans, but that the coffee cup and beer stein are virtually indistinguishable. Then people are going to have the same intuition of inequivalence between “I know what Frank stole” as said in a context in which “what Frank stole” denotes Q3 (that will seem true – the CIA might be able to answer Q3), and “I knows what Frank stole” as said in a context in which “what Frank stole” denotes Q4 (that will seem false – the CIA might well bungle a question so subtle as Q4). The need for contrasts remains.

  3. Jonathan Schaffer says:

    Hi Kenny,

    I think we have the same intuitions about 1, namely that it matters where we put the focus.

    Thus compare:
    1a: The CIA knows that FRANK stole the coffee cup.
    With:
    1b: The CIA knows that Frank stole THE COFFEE CUP.
    In the context in which the CIA merely presupposed (with little evidence) that Frank did the deed, and focused their investigation into what he stole, I think we both accept 1b as true but reject 1a as false.

    Dretske makes a similar point about the role of focus in knowledge attributions:
    Someone claiming to know that Clyde sold his typewriter to Alex is not (necessarily) claiming the same thing as one who claims to know that Clyde sold his typewriter to Alex… A person who knows that Clyde sold his typewriter to Alex must be able to rule out the possibility that he gave it to him, or that he loaned it to him… But he needs only a nominal justification, if he needs any justification at all, for thinking it was Alex to whom he sold it. (1981: 373)

    This is evidence for contrastivity. Differences in the location of focus are differences in the contrasts. Thus 1a says something like: “The CIA know that Frank [rather than Margaret] stole the coffee cup,” while 1b says something like “The CIA know that Frank stole the cofee cup [rather than the secret plans].” To the extent that our intuitions allow 1a and 1b to diverge, they exhibit contrast-sensitivity.

  4. Lance says:

    Disclaimer: I’m more linguist than philosopher, and therefore I’m coming into the debate (a) late and (b) with a very different perspective.

    As a quick note: I agree with Kenny and Jonathan about (1), namely that lack of stress on “Frank” (and perhaps added stress to “coffee cup”) makes the sentence seem fairly true to me.

    But this means that there’s something fairly strange going on here, then. We’re used to this sort of referential paradox in belief-contexts with definite descriptions—so in addition to the morning/evening star, and Orkut the spy, and the author of Waverly, we have:

    (10) The CIA believes that the thief stole a coffee cup.
    (11) The thief is my brother Frank.
    (12) The CIA believes that my brother stole a coffee cup.

    If the CIA doesn’t know (11), we have no trouble judging (10) true and (12) false. I feel like the same thing is going on in (1), except that, unlike (10)-(12), we have (a) “know” instead of “believe” and (b) a name instead of a definite description.

    Even a question or a proposition-with-focus won’t get rid of that name—both “what did Frank steal?” and “Frank stole the COFFEE CUP” are going to produce, in some manner and for some usage, the set of propositions {Frank stole the coffee cup, Frank stole the secret plans, Frank stole a laptop…}. So you’re stuck with that name in there. I must admit that I’m puzzled by my own intuition that the CIA can know that Frank stole the coffee mug, even if they don’t know that it was Frank who stole the coffee mug.

    ———

    I also have qualms about Jonathan’s examples. “I know whether there is a goldfinch or a blue jay is in the garden” is all well and good—but what if there’s a canary in the garden?

    Suppose I hear a bird, and being utterly ignorant of ornithology (to the point that I can’t tell a blue jay from a canary!), I ask you, “Look out the window, would you? I heard something in the garden—a goldfinch or a blue jay, probably.” You look out, see a flash of yellow feathers, and tell me, “Ah. I know whether there is a goldfinch or a blue jay in the garden. Indeed, I know that there is a goldfinch.” But what there is in the garden is a canary; so is either of your statements true? Obviously not, because “know” is factive. But I can’t escape the feeling that Jonathan’s scenarios don’t take that into account: being able to tell a goldfinch from a blue jay isn’t going to be enough for you to know whether there’s a goldfinch or a blue jay in the garden. You have to be able to narrow it down to that point, which requires knowing that it’s not a canary.

    I’m not sure if I’m expressing this properly, because the factivity seems like such a trivial fact. And yet I can’t get past that fact: if you tell me that Alice know whether a goldfinch or a blue jay is in the garden, then, I think, Alice knows that it’s a goldfinch or a blue jay that’s in the garden—which entails that she knows that it’s not a canary. Which means that even what Jonathan considers non-ornithological knowledge (K1) does require ornithological knowledge, doesn’t it?

    Must give this more thought.

  5. Heath White says:

    My two cents on Brian’s positive proposal: it seems clear to me that there is a de re/de dicto switcheroo going on to explain how

    (1) The CIA knows that Frank stole the coffee cup
    is false but
    (2) The CIA knows what Frank stole
    can be heard as true.

    I would say that the CIA knows, of Frank, that he stole the coffee cup. This does not entail (1), but it can entail (2) on one reading of (2).

    This way of putting it means that there need be no definite description D for Frank, under which the CIA knows that D stole the coffee cup; the CIA might have identified the thief with a aliased proper name, not a definite description. And it also means that we don’t need to confine ourselves to the subject position of the sentence; the thing about which the CIA has knowledge de re (but not de dicto) might be in the object position of the sentence.

  6. Jonathan Weinberg says:

    Fwiw, Lance, I’ve thought for some time now that that is exactly one point where Jonathan’s account is vulnerable: to know whether it’s p or it’s q, you need also to know that it’s at least (p or q). Or, at least, in a great many ordinary cases one needs to do so (if one always needed to do so, then maybe there’s a risk of scepticism creeping in here). Several of his cases become problematic if we try to take this into account — if you can’t tell whether it’s a canary or it’s a goldfinch, then you can’t know whether it’s a goldfinch or it’s a jay, because you just know whether (goldfinch or jay) exhausts the relevant space of options. And if you did know that, then of course you also know that it’s not a canary, and so you do know whether it’s a canary or it’s a goldfinch. This makes it less obvious than Jonathan needs it to be, that his K1 and K2 can differ in truth value — thus making his premise 1. in his reconstruction above less attractive.

    I will be pushing something like this line of objection as a part of my commentary on Jonathan’s paper at the Eastern APA.

    Also, Jonathan, I think there’s something off here (which I haven’t quite figured out how best to say) when you write, “By the way, this isn’t just an issue with wh-constructions. It seems a GENERAL feature of the language…” If contrastivism is a general feature of language, then why claim that knowledge-wh is not reducible to knowledge-that? After all, the knowledge-that sentences will already have (if you’re right) the same amount of contrastivish yumminess that the knowledge-wh ones do. Indeed, I thought you were antecedently committed to its being the case that I can’t know that it’s a goldfinch in the yard simpliciter, but only know that it’s a goldfinch in the yard (as opposed to a jay; or as opposed to in your yard; etc.) So… what gives?

  7. Kenny Easwaran says:

    I would have thought that the proper way to deal with the issue raised by Lance is to say (something along the lines of) in order to know the answer to a question, it’s sufficient to have the knowledge conditional on the presupposition in the question. Lance’s example is then an example of presupposition failure, rather than a failure in the account of knowledge itself. If we falsely believe the presupposition, then we will be tempted to falsely attribute knowledge to Alice. We would just be wrong in that case, but if the presupposition had held, then Alice’s abilities are sufficient to tell the difference between a bluejay and a goldfinch, so we would have been right.

    However, in the question period after his talk at Berkeley, Jonathan seemed to not want to make this move. I seem to recall that he went instead for something like saying that there are often three options: “Bush is on TV”, “Jackson is on TV”, “none of the above”. But this seems to make the knowledge hard again, as Jonathan Weinberg points out.

  8. Matt Weiner says:

    I agree with Jonathan, Lance, and Kenny, and I think disagree with Heath, that I can hear (1) as true or at least assertable if “THE COFFEE CUP” is stressed.

    I think that, as Lance says, it’s an Ortcutt case. Consider this: If the CIA saw a dark figure (who happens to be Frank) skulking down the hallway, and then came in and discovered the coffee cup gone, we’d probably accept

    (1a) The CIA believes that Frank stole the coffee cup

    in at least certain contexts—even though the CIA might vehemently demur if asked “Did Frank steal the coffee cup?”

    Now, let’s suppose that there was no eyewitness, but that we are talking in a context in which it is well known that Frank was the thief (and that the CIA doesn’t know who the thief was). I think that (1b) would be assertable:

    (1b) The CIA believes that Frank stole THE COFFEE CUP

    as a way of figuring out what the CIA believes was stolen. On some theories (1a) is true on a de re reading, and (1b) is false if helpful, because the CIA does not believe a singular proposition but only something with a definite description like “The thief stole the coffee cup.” I don’t think this difference is particularly significant; but in any case, in order to make this difference out we need to do a lot of theorizing. We can’t just read off the example that (1a) is different from (1b).

    Now if we convert (1a) and (1b) from belief-ascriptions to knowledge-ascriptions I think we get the same results. That is, it can be OK to assert them because the CIA know that whoever-it-was stole the coffee cup, even if they don’t know whoever-it-was. So we can use (1) to convey that the CIA know that it was the coffee cup that was stolen, if the focus is on “the coffee cup” rather than on “Frank.”

    This suggests that there’s context-sensitivity in the assertibility of (1), if not its truth conditions, but that the context-sensitivity may come from “that Frank stole the coffee cup” rather than “knows.” That is, the context may affect whether it’s acceptable to use that phrase to describe the CIA’s belief/knowledge; and that may depend on whether or not it’s important to establish whether the CIA has the name right.

  9. Jonathan Schaffer says:

    Warning: another long post ahead!

    1. Contrastivity and opacity [for Lance, Heath, and Matt]:
    I think these are independent phenomena. Both exist, and neither can be reduced to the other. Of course these phenomena can interact, when the focused element and its contrasts differ in purely intensional ways (as in Matt’s nice example). But first, there can be contrast-sensitivity without any opacity issues, if we just stipulate that the subject is aware of all the relevant identities. So look back again at:
    K1: I know whether there is a golfinch in the garden or a blue jay
    and:
    K3: I know whether there is a golfinch in the garden or at the neighbors.
    This looks like a contrastivity effect without any opacity issue.
    In the other direction, we can get substitution failure without contrast shift, if we just fix the contrast explicitly. So consider:
    K4: I know that Frank stole the coffee cup rather than the secret plans.
    K5: I know that the guy in dark glasses stole the coffee cup rather than the secret plans.
    This possible difference in truth-value looks like an opacity effect without any contrastivity issue.

    In any case, there seems to be a certain pattern of contrast-sensitivity present with our knowledge intuitions. Thus consider the following pairs (this is from my Irrelevance of the Subject paper):

    Who/what:
    (a) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know who stole the bicycle?
    (b) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know what Mary stole?

    Whether:
    (a) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know whether Mary or Peter stole the bicycle?
    (b) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know whether Mary stole the bicycle or the wagon?

    Rather:
    (a) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that Mary rather than Peter stole the bicycle?
    (b) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that Mary stole the bicycle rather than the wagon?

    Cleft:
    (a) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that it was Mary who stole the bicycle?
    (b) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that it was a bicycle that Mary stole?

    Focus:
    (a) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that MARY stole the bicycle?
    (b) Mary has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that Mary stole THE BICYCLE?

    Presupposition:
    (a) SOMEONE has stolen the bicycle from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that Mary stole the bicycle?
    (b) Mary has stolen SOMETHING from the toy store. The detective finds Mary’s fingerprints at the scene. Does the detective know that Mary stole the bicycle?

    If your intuitions are anything like mine, you will say that the detective knows in all the (a) cases, but does not know in any of the (b) cases. This looks like general contrast sensitivity. I find it hard to imagine explaining away each of these pairs (where each looks like a minimal pair differing only over contrast) as all somehow turning on opacity effects.

    Suggested moral: there are opacity effects, and there are contrast effects. Both exist. They should not be conflated.

    2. Factivity [for Lance]:
    Knowledge is factive. If s knows that p rather than q, then p is true. To answer Lance’s question, if there is (only) a canary in the garden, then you can’t know whether there is a goldfinch or a blue jay in the garden. (I had stipulated, when presenting the cases, that there was in fact a goldfinch in the garden.) Being able to discriminate p from q does not suffice for knowledge. I only meant it as a good intuition pump, at least for satisfying the evidential component of knowledge. The idea was that you can see how K1-K3 can differ, by seeing the different sorts of evidence each requires.

    3. Presuppositions and evidence [for Lance, Jonathan W, and Kenny]:
    Can I know that p rather than q, even if, for all I know, neither p nor q is the case? I think that the answer is yes, though I have a pragmatic story to tell about why it is often infelicitous to say as much.

    Here are three reasons why I think one can know that p rather than q, without knowing that either p or q is the case.

    First, it fits how we evaluate C students (and others with intermediate epistemic powers). Imagine a student who has learned enough to tell whether a given passage (from Hume) was from Hume rather than Plato, but not enough to tell Hume from Kant. In other words, if you ask this student ‘Hume or Plato?’ the student will get it right (and not by guessing), while if you ask the student ‘Hume or Kant?’ the student can only guess. What does this student know about who wrote the passage? Not nothing—this is the C student, not the F student. Not everything—this is the C student, not the A student. I think that (i) this student knows that Hume rather than Plato wrote the passage, though (ii) this student doesn’t know that either Hume or Plato wrote the passage (as opposed to it being Kant).

    Second, it dodges skepticism.

    Third, it fits a general and elegant distinction (due to Dretske) about the different roles played by presupposition and evidence in epistemic evaluation. The idea is that, if I know that there is a goldfinch rather than a blue jay in the garden, then the question posed is:
    Q: whether there is a goldfinch in the garden or a blue jay in the garden.
    And among the presuppositions of Q is:
    P: There is either a goldfinch or a blue jay in the garden.
    Call Q what is posed and P what is presupposed. Then Dretske notes that we need little to no evidence for what is presupposed, but we need conclusive evidence to close the inquiry into what is posed. So I might know the answer to Q, while merely presupposing P (and thus having little to no evidence, certainly not enough for knowledge, for P).

    Of course, it is odd to say, in one gulp, “I know that p rather than q, though I don’t know that (p or q) [rather than r]”. But the presupposition account explains the oddity pragmatically, for to claim not to know (p or q) [rather than r] puts strong pressure on us not to presuppose (p or q), since what was presupposed has now been questioned.

    (Kenny: I’m not sure why you thought I had said otherwise at Berkeley. But in any case, the bit about three options [e.g. whether ‘none of the above’ is on the multiple choice slate] concerns the contextually variable mapping from a whether-clause to a question. The clause ‘whether A or B’ could denote the options (a) A v B, (b) ~(A v B); or it could denote© A, (d) B; or it could denote© A, (d) B; or (e) none of the above; or it could denote© A, (d) B; or (f) both of the above, etc., etc. The bit about presupposition occurs once we fix the multiple choice slate.)

    4. The general view [for Jonathan W]:
    The general view is that ‘knows’ invariantly denotes a 3-place relation: Kspq. This is true for both knowledge-wh and knowledge-that ascriptions. So all knowledge ascriptions do have the same level of (to use Jonathan W’s memorable phrase) ‘contrastivish yumminess’. If you look at the final section of my paper, you’ll see that I explicitly argue for this.

    (I think you might have an older copy of my paper (one that still uses the misleading rhetoric of ‘reducing knowledge-wh to knowledge-that’), and an abridged one (which omits the final section on knowledge-that, to meet the APA space requirements). If you get a chance, check out the version on my website. Hopefully that will clear things up. Anyway, I’m looking forward to our session at the Eastern!)

  10. Jonathan Weinberg says:

    “Then Dretske notes that we need little to no evidence for what is presupposed, but we need conclusive evidence to close the inquiry into what is posed.” I think that we may have an explanatory cart-horse problem here. I think the following seem right:

    I. At any given time, there are many propositions that we can epistemically rely on & deploy without having anything like conclusive evidence for them, even though this is not true of most propositions. (Presumably ‘hinge’ type propositions are paradigm members of this category.)

    II. When a successful claim involves a presupposition, although the presupposition might itself be well-evidenced, it does not need to be such, and very well might just be among the propositions mentioned in I.

    But it doesn’t follow from those observations that merely relegating some proposition p into a presupposition thereby absolves us from needing evidence for p. It’s not the presuppositionness of p that explains our lack of need for evidence for p — it’s that p is a proposition for which we need little evidence that makes it such a good candidate for being presuppositionized (because you wouldn’t be able to get away with asserting it flat-out).

    So, in particular, if you’re in a situation in which you can tell A from B, but you can’t tell A from C, when is it alright to just ask whether you know whether it’s A or it’s B, without paying any mind to the possibility of C? Well, you need it to be already & independently the case that ~C is a good candidate for presuppositionizing.

    So this is why it’s hard to get the kind of case that the contrastivist really needs. If we’re in a situation where it’s true that I don’t know whether it’s A or it’s C, then we’re not going to be in a context where it’s true that I do know whether it’s A or it’s B (since it might be C). If C is somehow ruled out, then both propositions are true; if C is not somehow ruled out, then both propositions will be false.

    Maybe one way of seeing this, in terms of the machinery of presuppositions, is that there’s no context in which you can utter (without emphasizing the “know”) “I don’t know whether it’s A or it’s C”, and still have ~C lurking around as an operative presupposition for other utterances.

  11. Sam C says:

    Hi. Sorry if anything I’m about to say has been addressed, but I didn’t have time to read all posts.

    1. Something is fishy about J.S.‘s semantics for his embedded questions. Take ‘whether there is a finch in the garden’. On the Groenendijk & Stokhof semantics he endorses, this question would denote the propositional concept which partitions logical space into the propositions:

    ^There is a finch in the garden
    ^There is not a finch in the garden

    This is not the partition J.S. envisions, however. Which is more like:

    ^There is a finch in the garden
    ^There is a canary in the garden
    ^There is a bluejay in the garden

    The G&S semantics recursively generates this meaning, but for the question ‘what (bird) is there in the garden’.

    Nor is this the denotation of the question ‘whether there is a finch in the garden or a canary in the garden’, which would be:

    ^There is a finch in the garden or a canary in the garden
    ^It is not the case that there is a finch in the garden or a canary in the garden

    I don’t disagree that there is a question in the vicinity that is what J.S. wants: ‘Is it a finch or a canary that is in the garden?’, but it is not clear that this question can be agreeably embedded (as the awkwardness of J.S.‘s K-sentences attests), or how its denotation gets generated via a G&S semantics.

    2. The example: ‘The CIA knows what Frank stole’ is an interesting one. One possible avenue of explanation is that rather than being an embedded question construction, it is in fact a ‘concealed question’ (CQ) construction, a little bit like:

    (1) The CIA knows the object that Frank stole.

    It could be a CQ because ‘what Frank stole’ might occur as a free relative (something like a definite description) rather than a question-denoting interrogative phrase. There are different ways to test this (looking at languages which distinguish free-relatives from questions, for instance); an easy way would be to ask about the sentence:

    (2) The CIA knows who Frank was in league with.

    Can this be used to describe a situation in which the CIA knows the identites of the other conspirators, without knowing that Frank was involved? I’m not sure it can, but if it can, that rules out the CQ approach, since English doesn’t have ‘who’ free relatives. If the sentence cannot be used in this way, and the rest of my explanation is sound, then we may have found evidence for free-relative CQs in English.

    How does the CQ analysis help explain the example? A CQ like (1) means the same as the embedded question construction:

    (1’) The CIA knows what the object that Frank stole was.

    So rather than needing to substitute a definite description for ‘Frank’, all we need is for the description ‘the object Frank stole’ to take wide scope (actually, to my mind the right solution is to give the relative clause ‘that Frank stole’ wide scope, but I won’t go into detail about what that means).

    All this applies mutatis mutandis to the free-relative CQ, which is very similar semantically to (1).

    Sorry to have added so much to an already pendulous thread!

  12. Matt Weiner says:

    I don’t know anything about CQs and free relatives, but it seems to me that something like (2) is acceptable. Suppose that the CIA knows that there is a mole in their ranks, but they don’t know that Frank is the mole. Frank has just told the KGB that Valerie is a CIA agent. Frank’s KGB controllers, who all know that Frank is the mole and that the CIA doesn’t know this, are discussing if the CIA knows about Valerie’s cover being blown. One says:

    (2’) The CIA knows whose cover Frank blew.

    That seems to me as OK as the original (2).

    (And, as discussed above, I think “The CIA knows that Frank blew VALERIE’S cover” can also be OK.)

  13. Lance says:

    OK, so let’s see.

    As to J.S.‘s 11/29 post—I didn’t mean to suggest that contrastivity and opacity should be conflated. I think that, in fact, I merely confused myself over what can be opaque where.

    But for point 3, I’m not sure I agree with everything there, and I hope it’s not just intuition clash. I just can’t quite believe that it’s true to say “The C student knows that Hume and not Plato wrote this passage”, insofar as that sentence entails “The C student knows that Hume wrote this passage” and “The C student knows that Plato didn’t write this passage”, and while the latter is true, the former is false (because the student can’t rule out Kant, short of your asking him directly, “Hume or Plato? It’s one of the two”).

    It’s the same reason that I can’t quite accept Dretske’s distinction (bearing in mind that I haven’t read Dretske): I don’t see how your evidence for the question posed, “whether the bird is a goldfinch or a bluejay”, can be conclusive without conclusive evidence about what is presupposed. Suppose I hold up a small blue cubic object and ask you, “Quickly now, is this an apple or a wug?”, Would you be willing to say, “I know it’s a wug” (or even “I know it’s a wug and not an apple”), simply because the presupposition of the question I had asked—“this is an apple or this is a wug”—requires little evidence for you to accept it as true? I feel that I’d hesitate; I’d say, “I suppose it’s a wug”, or “I believe it’s a wug”, but never having heard of wugs, and never having seen blue cubic objects worthy of naming, I need more evidence that it’s a wug before I can feel that I know it’s a wug—i.e., evidence that the presupposition (“it’s either apple or wug”) is true.

    The point I skipped over (“it dodges skepticism”), I skipped because I’m utterly not qualified to comment on skepticism, and I mention it now because this last question—can you know that this blue cube is a wug?—may be dragging me back into that unfamiliar territory.

    ———

    With Sam C’s 11/30 post, I’m back into very familiar territory, having done no small amount of work on CQs. I think I can say with some confidence that a wh-question isn’t a CQ free relative, though I’ll admit that I’d want to sit down at a time that isn’t 6:15 am to work through some diagnostics before stating so with 100% certainty. (A quick diagnostic: “What did Frank steal?” “I don’t know. But the CIA knows.” — that has to be an actual question, and the same arguments should hold. Also: “*The CIA knows whatever Frank stole” is no good, although putting that into a clause, i.e. “The CIA knows what whatever Frank stole is”, is at the very least better, if not fine.)

    Of course, it’s my research in CQs that brings me to be reading this discussion, and even attempting to contribute to it, in the first place.

  14. Sam C says:

    Ah! I had a suspicion that Lance might be Lance Nathan, but thought it was too much of a long shot. I’m glad to finally cross swords. Hey I have a bunch of stuff on CQs (though my research has taken a slightly different turn) that I’d like to pester you about sometime, when this semester cools down.