In their sustained “defence of insensitivity”:http://www.amazon.com/exec/obidos/ASIN/1405126752/caoineorg-20?creative=327641&camp=14573&link_code=as1, Cappelen and Lepore rely quite a bit on indirect speech reports. So, for instance, the fact that the following bit of discourse always seems natural
bq. A: S knows that p
B [later, in different context]: A said that S knows that p
is taken as evidence against contextualism. I think this is fairly strong evidence, but not everyone agrees. It might be argued that speech reports are messy things, and no one really understands them. Fair enough, perhaps. It might be worth noting though that the same kind of phenomenon occurs with questions. Consider the following scenario. A is walking aimlessly around Ithaca. Every five minutes, she asks “Is Tamar here?”. All through this time, Tamar is working in her office in Goldwin Smith Hall, so her location doesn’t change. B, who knows this, changes her answer to A’s question, depending on whether or not A is in Goldwin Smith Hall. When A asks the question downtown, B says “No”. When she asks it again in Goldwin Smith, B says “Yes”. When she asks it yet again while climbing down Cascadilla Gorge, B says “No”. ‘Here’ is a true contextually sensitive term.
Note that if A and B are separated, B will answer according to A’s location, not her own. So if A is downtown (and B knows this) and is talking to B by phone, and asks “Is Tamar here?”, if B is in Goldwin Smith she can answer, “No, Tamar is here.” For any context sensitive term such that different speakers are in different contexts, this kind of speech act, where we answer “No” and then follow up by uttering the sentence that looks like the indicative form of the question, is possible.
Here we have two tests for context sensitivity. First test, can we change answers while the underlying facts stay the same? Second test, can we consistently answer “No” and then repeat the question? It seems ‘knows’ fails both tests for context-sensitivity. Since neither case involves speech reports, this means the contextualist has to posit semantic blindness that extends even to fairly simple question and answer conversations. Let’s see a couple of cases illustrating this.
Jack has an odd compulsion. Every five minutes he asks “Does John know that his car has been stolen?” As it turns out John’s car has been stolen, and John believes this on reasonably good authority. (He has testimony from the manager of the garage from where it was stolen. The manager is a somewhat shady character, but he’s telling the truth this time.) If contextualism about ‘knows’ is true, we should be able to find an environment where the following conditions are met.
# When Jack first asks, the right answer is “Yes”
# None of the facts about John, his car, his evidence about his car, defeaters etc change over the course of five minutes
# When Jack asks again, the right answer is “No”
And as far as I can tell, no such change is possible. Let any sceptical scenarios you like pervade Jack’s mind. Let the importance of the proposition _John’s car was stolen_ rise as much as you like. Change whatever features of Jack’s context you like. As long as you leave my context (as the answerer) fixed, and John’s relation to his car fixed, I won’t change my answers. Contextualism incorrectly predicts otherwise.
The one thing that _might_ make me change my mind is if Jack changes the inflection of his question. I can imagine circumstances where I would answer “Yes” to “Does John know his car has been STOLEN?” and “No” to “Does John KNOW his car has been stolen?”, and that’s some (weak) evidence for thinking that stress changes the semantic value of ‘know’. But this is obviously of no help to the contextualist.
What about the second test? On the assumption that contexts can vary between speakers, there should be a context in which I can answer Jack’s question “Does John know that his car has been stolen?” with “No, John knows his car has been stolen.” Clearly this is impossible. Keith DeRose, addressing a somewhat different point, has a response to this. (See “these”:http://pantheon.yale.edu/%7Ekd47/SSS.pdf “two”:http://pantheon.yale.edu/%7Ekd47/Bamboozled.pdf papers.) When I talk with Jack, I enter his context vis a vis ‘knows’. I don’t enter his context vis a vis ‘here’, but ‘knows’ is somehow different. And that explains why we can’t answer with “No, he knows that his car has been stolen.”
We can change the example a little to push the problem around a bit. Imagine I’m not answering Jack’s question, but eavesdropping on his conversation with high tech listening devices that carry his words from a conversation in New Jersey all the way back to my secret lair in Ithaca. So I’m not in any sense in Jack’s context. I have the habit of answering questions Jack asks. (I don’t really have a high-tech lair, but if I did have I would have this habit. I already do this when TV commentators ask questions.) But there’s no imaginable context where Jack could ask “Does John know that his car has been stolen?” and I could answer “No, John knows that his car has been stolen.”
If I’ve understood the appendix to DeRose’s “Bamboozled”:http://pantheon.yale.edu/%7Ekd47/Bamboozled.pdf correctly, his explanation of this is that Jack’s context half-seeps into mine, so if I answered “No, John knows that his car has been stolen.” I would say something neither true nor false. To say something true, I’d have to say something true in both Jack’s context, and mine.
Intuitions get murkier here, but this seems wrong twice over to me. First, the speech I made (‘No, John knows that his car has been stolen.”) seems contradictory, not neither true nor false. Second, it seems that if I’m in a low stakes context, and I answer “Yes” I say something true, though on DeRose’s account I would say something neither true nor false if Jack happens to be in a high-stakes context.