Skip to main content.
September 23rd, 2005

A Cheer for Contextualism

Ishani and I were talking about who various people here in Canberra are, and we noticed that the following conversation seems coherent, and in an environment where the folk are ignorant of Zaphod Beeblebrox, even plausible.

A: Do you know who that guy is?
B: Yeah, that’s Zaphod Beeblebrox.
A: Who’s Zaphod Beeblebrox?
B: I don’t know. I was just told that’s who that guy is.

The puzzle is that B seems to have contradicted herself. It seems that with her first word she said that she knows who that guy is. And with her second sentence, she said that she doesn’t know who Zaphod Beeblebrox is. But she knows that guy is Zaphod Beeblebrox, so presumably she knows who that guy is iff she knows who Zaphod Beeblebrox is. Contradiction. Yet her words sound consistent, even plausible. What has gone wrong?

Suggestion: The problem arises because we substitute into a non-extensional context. It is possible to know that guy is at the party without knowing Zaphod Beeblebrox is at the party, and hence possible to know who that guy is without knowing who Zaphod Beeblebrox is.
Reply: True, these substitutions are sometimes problematic. But this is a situation where we don’t just have co-referring expressions, but known to be co-referring expressions. And the knowledge of their co-reference is clearly salient to the speaker. So this probably isn’t what is going wrong.

Suggestion: The problem arises because ‘that guy’ is a quantificational rather than a referential expression.
Reply: That may well be true. If it is, drop all uses of ‘guy’ from the conversation so we have a simple demonstrative rather than a complex demonstrative. The conversation sounds a little more awkward then, but still coherent.

Suggestion: ‘Knows’ is context dependent, as Lewis, Cohen and DeRose have been arguing for donkey’s years.
Reply: Even if all the other objections to contextualism don’t work, it is hard to see how on any of their theories there was a change of context between A’s first utterance and B’s second utterance.

Suggestion: ‘Knows who’ is context dependent, even though ‘knows’ is not, as Boër and Lycan (among others) have suggested.

I think that’s basically right, and really I’m not sure how a non-contextualist can explain this data. The fact that B doesn’t even seem to have contradicted herself, or taken anything back, looks like a very strong argument for this kind of contextualism. (Compare the examples motivating ‘orthodox’ contextualism where there is a strong feeling that when a speaker first asserts, then rejects, that S knows that p after an alleged context shift. There there is a feeling the author is taking something back.) But in a recently posted paper David Braun has argued strongly against this kind of contextualism. The crucial argument, to my mind, concerns the following claim. David argues that if contextualism is true, claims like the following should be true, but they sound false.

‘Zaphod Beeblebrox is that guy’ is an answer to the question ‘Who is Zaphod Beeblebrox’, and B knows that Zaphod Beeblebrox is that guy, and so she knows an answer to the question of who Zaphod Beeblebrox is, but she does not know who Zaphod Beeblebrox is.

(This is a modified version of his (26) from page 23.) I’d like to have more to say here, but there’s not much more that I can say other that in this little story, this claim sounds true not false. If the claim was false (presumably because the last clause was false) then B’s last claim should sound like it is false. But it doesn’t, it sounds true. So I think the possibility of this conversation is a quite compelling argument for contextualism about ‘knows who’.

Posted by Brian Weatherson in Uncategorized

19 Comments »

This entry was posted on Friday, September 23rd, 2005 at 6:43 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.

19 Responses to “A Cheer for Contextualism”

  1. Vince says:

    Zaphod Beeblebrox is one of the main characters from The Hitchiker’s Guide To The Galaxy by Douglas Adams. No doubt you’ve heard of the film. I read half of the series. Funny stuff. You should read the books.

    Hope that cleared some stuff up for you.

    - Vincent Charrot.

  2. Vince says:

    Haha… I just read the article in full. Sorry, I didn’t really realize that you were just using the name as an example, I just skimmed the writing and saw ‘Who’s Zaphod Beeblebrox?’

    Sorry.

  3. David Braun says:

    Here’s the dialogue again.

    A: Do you know who that guy is?
    B: Yeah, that’s Zaphod Beeblebrox.
    A: Who’s Zaphod Beeblebrox?
    B: I don’t know. I was just told that’s who that guy is.

    The sentences B utters in the last entry do imply a contradiction, as Brian says (assuming that if B has been told who that guy is, then he knows who that guy is, which I think is plausible). I think this is what’s going on: B does know an answer to A’s second question, “Who’s Zaphod Beeblebrox?” (e.g., B knows that Zaphod is that guy over there) and so he knows who Zaphod Beeblebrox is. However, when A asks his second question, A already knows the answer that B knows, and B knows this. But A is clearly not satisfyed with this information. He clearly wants some other sort of information about Zaphod. B doesn’t know much else about Zaphod, so he knows he can’t satisfy A’s desire for further information. It would be uncooperative for B to utter the answer he does know, namely “That guy is Zaphod”, so he doesn’t want to utter that. B could at that point say to A “I don’t know an answer to your question that will satisfy your desire for further information.” But instead, he unreflectively says “I don’t know”. B is mistaken when he says this, but his utterance let’s A know that B does not know anything more about Zaphod (at least not anything that will satisfy A).

  4. P.D. Magnus says:

    I had a similar reaction to David Braun’s (above).

    However, we don’t have say that B is mistaken in the second utterance. It is plausible to understand “I don’t know” in this context as “I don’t know [anything further].” That is straightforwardly true and doesn’t require any contextualism.

  5. Carrie Jenkins says:

    “it is hard to see how on any of their theories there was a change of context between A’s first utterance and B’s second utterance …”

    So how does your proposed contextualist response work?

  6. Jason Stanley says:

    Dear Carrie,

    I’m not alone in thinking embedded questions are context-sensitive, and their context-sensitivity has nothing to do with the verb. For example, all question-embedding verbs, such as “know”, “tell”, “learn”, etc. have this context-sensitivity, suggesting that it is a feature of the embedded complement, rather than the verb chosen.

    Here is one way to deal with the context-sensitivity of questions (not the only way). Let the semantic value of a question be a function from contexts to propositions (this enables us to treat the distinction between intensional question-embedding verbs such as “ask” and “wonder”, on the one hand, and extensional question-embedding verbs such as “tell” and “know”, on the other; the former have, as their complements, the function from contexts to propositions). So a question is a function from a context to a complete answer. Let “who Bill is” in “John knows who Bill is” express a function from the context of use to a proposition; the sentence is true just in case John stands in the knowledge relation to that proposition (similarly for “John told Sally who Bill is”).

  7. Carrie Jenkins says:

    Hi Jason,

    Sure – that all makes sense (at least it’s surely an option in logical space). I was just thinking that to motivate any contextualist account of why B hasn’t contradicted herself, there should be some context shift between her two utterances, and I thought Brian seemed to be saying there wasn’t one. I thought that was his objection to ‘knows’ contextualism in the chunk I quoted. So I had in mind a dilemma: either there is a context shift or there isn’t. If there is, his objection to ‘knows’ contextualism doesn’t work. If there isn’t, his ‘knows-who’ contextualism doesn’t work.

  8. David Braun says:

    I would like to get straight on the view that Jason is expressing in his contribution. He said:

    “Here is one way to deal with the context-sensitivity of questions (not the only way). Let the semantic value of a question be a function from contexts to propositions (this enables us to treat the distinction between intensional question-embedding verbs such as “ask” and “wonder”, on the one hand, and extensional question-embedding verbs such as “tell” and “know”, on the other; the former have, as their complements, the function from contexts to propositions). So a question is a function from a context to a complete answer. Let “who Bill is” in “John knows who Bill is” express a function from the context of use to a proposition; the sentence is true just in case John stands in the knowledge relation to that proposition (similarly for “John told Sally who Bill is”). “

    I assume we need something similar to Kaplan’s character/content distinction for interrogative sentences. We want, in some sense, for an interrogative sentence that contains an indexical, such as “Who are you?”, to semantically express different semantic contents in different contexts, depending on who the agent of the context is addressing. Similarly, a sentence like “John asked who you are” should semantically express different propositions in different contexts, depending on who the agent of the context is addressing in that context.

    Jason says that the ‘semantic value’ of a question is a function from contexts to propositions. (I assume that by ‘question’ he means ‘interrogative sentence’.) Since the argument of this function is a context, this semantic value seems to resemble something more like a character for an interrogative sentence than a (semantic) content that the sentence would have with respect to a context. But doesn’t Jason want an analog of semantic content in his theory?

    It seems like the desired view might be something like this. We make a three-way distinction among character, semantic content with respect to a context, and extension with respect to a context. All of this is familiar for declaratives. We assume that the extension of an interrogative sentence, in a context, is a proposition that is (something like) a true complete answer to that interrogative sentence, in that context, in the world of that context. The semantic content of the interrogative, in a context, is a function (a kind of intension) from worlds to propositions—its value at any world is (something like) the true complete answer to the interrogative, in that context, with respect to that world. For example, the extension of the interrogative sentence ‘Who did you marry?’ in a context in the actual world in which the addressee is Bill Clinton would be the proposition that Bill Clinton married Hillary Clinton (and no one else). The semantic content of that interrogative sentence in that context(that is, the intension it semantically expresses) would be a function whose value at any world is (in effect) a true complete answer at that world: in some worlds it would be the proposition that Bill Clinton married Hillary Clinton (and no one else), in others the proposition that Bill Clinton married Monica Lewinsky (and no one else), and so on.

    Is this something like the view that Jason intends?

    Assume so. Then Jason holds that there is a sort of unobvious context-sensitivity for interrogative sentences (in both their direct and indirect forms). An interrogative sentence that contains no obvious indexicals, like “Who is Zaphod Beeblebrox?” may vary in both extension and semantic content from context to context within the actual world. Since it varies in extension from context to context in the actual world, this (intuitively) means that the true complete propositional answer to the interrogative sentence varies from one context to another (in the actual world): in one context, the true complete answer is proposition P1, in another context (in the actual world) it is proposition P2, etc.. For this to be the case, the semantic content of the interrogative sentence would also have to shift from context to context in the actual world (so it seems to me, at any rate).

    Carrie’s question to Brian did not really depend on any of these details of the semantic theory of interrogative sentences. In Brian’s reply to his third suggestion, Brian says “it is hard to see how on any of their [Cohen’s, DeRose’s, Lewis’s] theories there was a change of context between A’s first utterance and B’s second utterance.” Carrie reasonably inferred that Brian thought that nothing in the context that is relevant to the semantic content of any expression shifted through the conversation. But Brian didn’t quite say that. Since Brian was discussing theories which say that ‘know’ is context-sensitive, he might have meant that nothing in the context changes that is relevant to the semantic content of ‘knows’. Jason probably agrees with Brian on this, but he may think that something else shifts in the context which is relevant to the semantic content (and extension) of the phrase “who that guy is”.

    My apologies for being both slow to understand and pedantic.

  9. Jason Stanley says:

    Many semantic theories for embedded questions (and unembedded questions) take the question words to be quantificational expressions of some sort (as I recall, even Boer and Lycan treat ‘who’ as a quantifier over properties, though I don’t have their book here). So embedded questions do not contain obvious context-sensitive expressions, namely question words themselves; they’re quantifiers.

    There are a number of different ways for treating the semantics of embedded questions; different ways would involve distinct choices for treatment of the context-sensitivity of the embedded questions. I’m trying not to commit myself to any specific approach right now. But in response to David’s question: he’s of course right that I agree with Brian that nothing in the context is relevant to the semantic content of ‘knows’, but I am a contextualist about quantifiers, and I assume ‘who’ is some sort of quantifier, that has a restricted domain.

  10. Brian Weatherson says:

    I’ve been agreeing with Jason about everything to date, so I haven’t chimed in much. But I thought I’d make a couple of the agreements explicit.

    I agree it’s possible to argue that B’s utterances in the original dialogues are false but helpful, just as when we say “Everyone liked the play” we might convey the true information that everyone we were with liked the play, even if some other people disliked it. But I don’t see what would motivate such a move. If anything, it seems even less motivated in the case of the ‘knows who’ claim, because it isn’t too hard to expand the intuitive quantifier domain with focus (“Really? EVERYONE?”) in the play case, but I don’t see how focus could do a similar job in the knows who case. If I agreed with David’s intuitions about principles like the inset one in the original entry I would agree that’s a motivation, but that principle doesn’t seem true.

    Actually, if I was going to be an invariantist about ‘knows who’ claims, I’d be tempted to go to the opposite extreme to David, and say S knows who x is iff for every true proposition about x, S knows that proposition. I suspect that ‘universalist’ position has just as much force as David’s ‘existentialist’ version that – after all we can coherently say “I know a bit about his background, but I don’t really know who he is.” That is, we’d say that to know who x is, you have to know the complete answer to the question “Who is x?”.

    I was sloppy in talking about context in the post, but I think Jason cleared this up well. I meant that there is nothing that could change the contextual value of ‘knows’ between one usage and the next. But obviously some things in the context change – for one thing it becomes common knowledge that that guy is Zaphod Beeblebrox. It’s plausible I think to take these embedded questions to be quantifiers, with the quantifier domain to shift with changes in the context. As Jason has nicely demonstrated elsewhere, quantifier domains can vary at the drop of a hat, so it is plausible it changes between B’s two utterances.

    I don’t have a theory of what generates the quantifier domain, nor even whether it is a universal or an existential quantifier. I think it won’t be trivial to work such a theory out, but it does seem like a project worth doing.

  11. David Braun says:

    Brian,

    Thanks for reading my paper and taking the time to discuss it.

    Yes, in my paper, I adopt what you call the existentialist view: X knows who Y is iff X knows an answer to the question of who Y is.

    The universalist view (“X knows who Y is iff X knows every answer to the question of who Y is”) has the consequence that no one knows who Y is, especially if we take every true proposition about Y to be an answer to the question of who Y is, as Brian does. For example, since no one knows every true proposition about the person Bill Clinton married, it follows on the universalist view that no one knows who Bill Clinton married. A similar point goes for analogous universalist views about when, where, how, and why questions. E.g., on the universalist view of knowing-why, to know why John ate pizza for lunch one would (seemingly) have to know all of the propositions that give a reason for, and/or explain, his eating lunch. So no one knows why John ate pizza. This is why I favor the existential view. I don’t mention this in my paper.

    When I said that, on Jason’s view, interrogative sentences are unobvious context-sensitive expressions, I meant only to contrast them with such obviously context-sensitive expressions as ‘today’ and ‘here’. I assume that quantifiers, and wh-words and wh-phrases, are not obviously context-sensitive expressions in this sense.

    I do not discuss the details of the semantics of interrogatives in my paper. But I do not take interrogative words to be quantifiers. I also think the view that interrogative sentences are context-sensitive has various odd consequences.

  12. Andrew Jorgensen says:

    The dialogue and inconsistency look a little like Eubulides’ ‘Hooded Man’ paradox to me. (A hooded man walks in to the room. You don’t know who the hooded man is. In fact he’s your brother. So you don’t know who your brother is). Graham Priest has a piece on the paradox in the Journal of Philosophical Logic 31(5), 2002: 445-467.

  13. Sam C says:

    I’m really interested in this stuff; a couple of points.

    NB Jason that on your proposal ‘wonder’ is a monster.

    There might be an equivocation in Braun’s argument since there may be semantically distinct english copulas (the verb ‘be’). Higgins wrote his dissertation (1978) on this, and I believe it began with Chomsky and others noticing the following sort of minimal pair:

    (1) John’s favourite hobby is keeping him from sleeping.

    (2) John’s favourite hobby is keeping himself from sleeping.

    While the ‘is’ in (1) is ‘predicational’, the ‘is’ in (2) is ‘specificational’ and somehow allows a non-c-commanding antecedent to bind the anaphor ‘himself’ in the complement, something not normally allowed:

    (3) #John’s mother loves himself.

    There might similarly be different copulas occurring in the dialogue:

    A: Do you know who that guy is?
    B: Yeah, that’s Zaphod Beeblebrox.
    A: Who’s Zaphod Beeblebrox?
    B: I don’t know. I was just told that’s who that guy is.

    The copula in the first two sentences appears to belong to yet another category that Higgins calls ‘identificational’ (one of his tests for the identificationality is appearing with bare ‘that’ subject). Unfortunately (for one hoping to use Higgins’ taxonomy to resolve the issue) the next copular sentence (A’s second question) would also come under the heading of ‘identificational’.

    It seems there might be more fine-grained distinctions in the vicinity; perhaps we can find subtypes of Higgins’ identificational rubric. For instance, notice that A cannot ask ‘Who’s that?’ (with neutral intonation) instead of ‘Who’s Z.B.?’, suggesting that the latter ‘is’ might be somewhat different from the usual type of identificational copula.

    As a very sketchy first pass, there might be different copulas corresponding to the different sort of counterpart relations (or world-lines) proposed by Hintikka (passim.). So there would be a sort of ‘perceptual identificational’ copula used when picking out people at a party, and a ‘descriptive identificational’ copula used when narrowing down an individual by his/her properties. It would be interesting if the same distinction appeared in copular sentences as did with flavours of modality, although at this stage it would be tendentious to prefer a semantic over a pragmatic account.

  14. Brian Weatherson says:

    David,

    I agree that the universalist will say unintuitive things about particular cases. But as you note in the paper, the existentialist also has to explain away some deviant intuitions. My gut feeling is that the universalist has the easier task here. I can find it easier to believe that I don’t strictly and literally know who Bill Clinton is (though it might be helpful to falsely say that I do) than that I do strictly and literally know who that is (said while pointing at someone across the quad who I know is a human in Australia).

    Both of these positions strike me as fantastic, but there is a live question here. On the view that the quantifier domains are contextually restricted, is the relevant quantifier existential or universal (or something else, e.g. generic)? I don’t think that will be an easy case to answer.

    Sam,

    I think you’re right that I’ve been playing fast and loose with the copulas. I’ve got some thoughts about how this interacts with cases innvolving plural referring terms in the ‘knows who’ construction that hopefully I’ll get posted later today.

  15. Sam C says:

    Notice the following difference (from Higgins 1979/Percus 2002):

    A: That is Z.B.
    B: And who would you say Z.B. is, pray?
    *And who would you say is Z.B., pray?

    Adding ‘would you say’ disambiguates between two structures of ‘who is Z.B.?’ – one where ‘who’ starts out in object position (unstarred) and one where it starts out in subject position (starred). The latter question is only good if we haven’t perceptually singled out Z.B.

    This gives us a way of pinpointing the equivocation in the Braun argument, by rewriting it as follows:

    ‘That guy is Zaphod Beeblebrox’ is an answer to the question ‘Who is Zaphod Beeblebrox’, and B knows that that guy is Zaphod Beeblebrox, and so she knows an answer to the question of who is Zaphod Beeblebrox, but she does not know who Zaphod Beeblebrox is.

  16. Matt Weiner says:

    But I think, abusing stress a little, we can come up with a non-perceptual case that creates a similar ambiguity:

    “Every schoolchild knows who Gavrilo Princip was: Franz Ferdinand’s assassin. But not until this shocking new biography did we know who Gavrilo Princip was. Described as ‘A Chekhov character who knew how to shoot’ blah blah blah…” and yet, looking at a picture of Black Hand members, we don’t know who (which one) is Gavrilo Princip.

    So I’d say here we have a case in which in one sense every schoolchild knows who G.P. is and in another sense every schoolchild does not know who G.P. is. It might be thought that we’re speaking figuratively when we say we don’t know who G.P. was—or that universalism or existentialism is the way to go here. So non-contextualists might have some outs here. But the passage about G.P. doesn’t seem to me contradictory on its face, nor to contain the equivocation Sam C points out.

  17. David Braun says:

    In my paper, I give a sentence like Brian’s sentence (which follows) and claim that my sentence is at least odd or peculiar.

    ‘Zaphod Beeblebrox is that guy’ is an answer to the question ‘Who is Zaphod Beeblebrox’, and B knows that Zaphod Beeblebrox is that guy, and so she knows an answer to the question of who Zaphod Beeblebrox is, but she does not know who Zaphod Beeblebrox is.

    Sam claims (roughly) that ‘is’ is ambiguous. I’m not sure what to think about this. But if it is ambiguous, then we can ask ourselves ‘Is the displayed sentence false, contradictory, or at least semantically odd, when ‘is’ is read in the same way throughout?’. I think it is. If so, then I can use it to argue against contextualism about ‘knows-wh’.

    In any case, I need not rely on questions and answers that contain the copula in order to argue in roughly the same way against contextualism for ‘knows who’. Consider the following.

    ‘Zaphod Beeblebrox kissed that guy’ is an answer to the question ‘Who did Zaphod Beeblebrox kiss?’, and B knows that Zaphod Beeblebrox kissed that guy, and so she knows an answer to the question of who Zaphod Beeblebrox kissed, but she does not know who Zaphod Beeblebrox kissed.

    There is no copula in the question and answer here. I claim that the sentence as a whole is necessarily false. If I’m correct, then I could use this sentence (rather than Brian’s sentence or the sentence I actually use in my paper) as a premise in my argument against contextualism about ‘knows who’ (and ‘knows wh’ more generally), and thus avoid issues about ‘is’ altogether.

  18. Sam C says:

    To recap: Brian reprised the ‘crucial argument’ in David Braun’s paper and reproduced its central claim. To David’s ear, this claim sounded false (though a form of contextualism supposedly entailed its truth). To Brian’s ear, it sounded true. I chimed in, suggesting that there was an equivocation (though not necessarily a ‘fallacy of equivocation’) in the central claim of the argument, which might explain the differing judgments – basically, Brian was hearing a different copula to David in the last sentence.

    Of course, if semantic differences between various copular sentences can be used to account for all the readings of ‘Who is NP?’ questions, then we don’t require such sentences to be context-sensitive. Personally I’m skeptical that this can be done. However, I also feel that copular questions deserve seperate treatment, for their many different readings, and shouldn’t be lumped in along with questions like ‘Who came to the party’, either by locating their context-sensitivity in their status as questions, or in the quantificational nature of ‘who’.

  19. Erik says:

    And here I thought B was going to respond with “Vell, Zaphod’s jist zis guy, you know?”