Keith’s response to Ernie’s complaint that contextualism is not a theory of knowledge is to grant that it isn’t, but maintain that it is important to epistemology nonetheless. Keith says, “To the extent that contextualism/invariantism is an issue in the philosophy of language, it’s a piece of philosophy of language that is of profound importance epistemology. How we should proceed in studying knowledge will be greatly affected by how we come down on the contextualism/invariantism issue. For contextualism opens up possibilities for dealing with issues and puzzles in epistemology which, of course, must be rejected if invariantism is instead correct the correct position. And how could it be otherwise?”
Basic areas of philosophy are surely related in interesting ways to semantical projects, and it is important to see exactly how. For example, how are semantics and metaphysics related? I would have thought that the answer can’t be at the substantive level of telling us what things there are, or what the nature of things is; instead, the connection will be, I suspect, at the level of logical form.
So here’s my worry about the semantic approach to epistemological theorizing. I would expect some connection between semantics and epistemology at the level of logical form, but not in a way that tells us which particular epistemological theory is true. So, for example, the question of whether K is an operator or a predicate may tell us something about what kind of thing knowledge is. And the issue between contrastivism, understood as the view that knowledge is a three-place relation, and more standard theories, on which knowledge is a two-place relation, will be important to epistemological theorizing.
This is not meant to say that epistemological viewpoints don’t presuppose viewpoints in the philosophy of language (and vice-versa). Maybe it’s just my semantic skepticism (in Jason Stanley’s sense of that term) coming through, but I’m happier seeing theory construction with tests run on usage to see where the theory forces the pragmatic/semantic distinction to be drawn, rather than starting out thinking that we have this distinction under control and that we can isolate the semantics first and engage in theory construction in epistemology based on those results. The former approach has a long and distinguished position in epistemology already: consider what all of us have to say about perfectly ordinary and acceptable uses of ‘know’-sentences such as “I just knew I’d make an A on that test, but I failed again.”
I don’t think any of this cuts against contextualism, because that view needn’t be conceived of to require a commitment to the priority of the semantic project. In fact, it can be conceived in the opposite direction. Start with a relativistic theory of knowledge (or, if one wishes, call it epistemic pluralism to indicate the idea that there is at least one knowledge relation but not only one), a theory designed to find a middle ground between skepticism and Mooreanism. Then adopt a contextualist semantics to, as Williamson puts it, tame the relativism.
Maybe what’s bugging me here is the idea that order of inquiry privileges addressing the contextualism/invariantism issue over traditional theory construction in epistemology. I’m not sure Keith thinks this; in fact, I expect he will deny that this is his view. I expect we agree on something like the following: We are lost in an ocean of perplexing philosophical issues, and most all of them are connected in one way or another with all the others. So wherever you start, you’ll be presupposing things outside of the initial place of inquiry. In the process of trying to sort it all out, there is simply no required starting point, and also no point immune from being undermined outside the boundaries of one’s subdiscipline.
If we disagree, it’s probably about the degree to which one ought to be a semantic skeptic. Keith tends to appreciate ordinary language approaches to philosophy more than I, and uses arguments of this sort in defense of contextualism. Anyone who is as much a semantic skeptic as I am will find this approach unenlightening, unless there are clear tests to differentiate semantics and pragmatics. So maybe Jason should clear all this up for us, and then we can make some progress!