Soames on A Priori Knowledge
I'm trying to catch up with Dave Chalmers's reading of Scott Soames's Reference and Description. I'm still at chapter 4, and my reaction to it is not quite the same as Dave's. (I began this entry as a comment over there, but it somehow grew way too long.)
Let's stipulate that "Lee" (rigidly) denotes the youngest spy (if there is one). Soames argues that if
1) "Lee is the youngest spy (if anyone is)" is contingent a priori,then
2) we can know a priori that Lee is the youngest spy (if anyone is),and so
3) we can know a priori of Lee that he is the youngest spy (if anyone is)
But (3) is false, so (1) must also be false.
As Dave notes, the move from (2) to (3) presupposes a non-trivial theory of knowledge attribution. If (2) expresses that we can be knowledge-related to some kind of singular proposition involving Lee, the transition to (3) seems acceptable. Likewise if (2) says that we can rule out worlds at which Lee is not the youngest spy (i.e. if the proposition we're supposed to know is the 2-intension of "Lee is the youngest spy").
Soames does not defend or even explain his theory of knowledge attribution in any detail. He does however argue against a simple alternative, which would not license the move from (2) to (3). On this alternative, "x knows that S" is true if x understands the sentence S and knows it to be true. (Soames points out that this leads to trouble in cases like Kripke's Pierre.)
I agree that the alternative is too simple. And Soames may be right that Kripke and, more clearly, Kaplan leaned towards a theory of knowledge attributions on which the move from (2) to (3) is valid. I think even two-dimensionalists could accept such a theory.
A more common, and better, view among two-dimensionalists is that there is no general assignment of propositions p to sentences S such that "x believes that S" is true iff x stands in some kind of belief-relation to p (say, iff p is a subset of x's belief worlds). This is certainly Lewis's view, and I think also Stalnaker's and Jackson's (not sure about Chalmers). Such a two-dimensionalist could say that there are readings of (2) and (3) on which (3) follows from (2), and there are other readings on which it does not. (Adding that (2) is meant "de dicto" encourages readings on which the inference fails. But I wouldn't say that there are just two readings, one de re and one de dicto.)
The move from (1) to (2) looks more problematic to me. Though this is in part a matter of terminology. One can probably define "a priori" in such a way that the transition is trivial. On such a definition, apriority inherits the complicated ambiguity, vagueness and indeterminacy of knowledge attributions, and I would agree that, so understood, it is far from clear whether "Lee is the youngest spy" is a priori.
(Another example: Suppose Fred knows that for no x,y,z and n>2 is x^n + y^n = z^n; but he doesn't know that this is Fermat's Theorem; he believes that Fermat's Theorem says that every number is the sum of two primes. Then in some contexts one can truly say that Fred doesn't know that Fermat's Theorem is true, and that he could not find it out by reasoning alone. If apriority is defined in terms of knowledge attributions, we could then conclude that "Fermat's Theorem is true" is not a priori in that context.)
I would prefer to characterize apriority in some way such that (1) is clearly true, but the transition to (2) becomes problematic. But I guess the other usage on which (1) implies (2) is just as common.
So I've learned that I should be careful when I say that sentences are a priori iff they have a universal A-intension.