I’ve been thinking about contrastivism lately, and just noticed an endnote (I think it’s number 34) in Jonathan Schaffer’s forthcoming piece in Phil Studies that I hadn’t noticed before. The topic of the note concerns enlarging our body of knowledge by deducing one claim from another. For contrastivists, the claim that Moore knows he has hands involves an implicit contrast, so that the proposition expressed is something perhaps to the effect that Moore knows he has hands rather than stumps.
Suppose Moore deduces from the claim that he has hands that he’s not a brain-in-a-vat. Schaffer says that’s OK, since his knowledge is now something like: knowing he is not a brain in a vat rather than having stumps in place of hands. He says that all it takes to know this is to be able to rule out the contrast claim: “Expand p [the principle that allows one to deduce not-BIV from having hands] does allow Moore to know that he is not a brain-in-a-vat rather than someone with stumps. But this is actually a very modest bit of knowledge, secured not by Moore’s possessing any sort of positive evidence for not being a brain-in-a-vat, but rather by Moore’s possessing conclusive negative evidence against having stumps (reminder: the stumps-possibility is restricted to worlds in which stumpedness would be apparent).”
Two points here. The first is the minimal standard given for knowing here: you know because you can rule out the contrast, even though you don’t have any evidence that you’re not a BIV. I don’t think Schaffer needed to say this, though. He could have said that Moore knows because he can rule out the contrast and deduction transmits evidence from the hands claim to the non-BIV claim.
The second is this. Shouldn’t contrastivism cite genuine contrasts in knowledge attributions? E.g., if “Joe knows that Texas is south of Canada,” involves some implicit contrast, I don’t see how a suitable theory would interpret it the missing parameter so as to yield “Joe knows that Texas is south of Canada rather than that North Dakota is north of Nebraska.” If I heard the latter claim, I’d interpret it as indicating that Joe knows the first claim and doesn’t know the second. When I hear a “rather than” knowledge attribution involving no logical contrast, it seems to me to have the form of a conjunction, not a three-place knowledge attribution of the sort contrastivists want us to focus on when we say things like, “Moore knows he has hands rather than stumps.”
You may be tempted to construe the case like this: Moore infers he’s not a BIV (understood to mean that all of his sensory modalities give him false beliefs) rather than having stumps (and being able to tell that this is so by looking). Schaffer notes parethentically in the passage quoted above that the stumps-possibility is restricted to worlds in which this is apparent, but that isn’t the same as building the restriction into the semantic content of the contrast. Instead, it only means that there are certain worlds epistemically relevant to whether you can rule out having stumps. Building it into the content gets the wrong results, since it makes the contrast harder to rule out (since it is now a conjunction).