A Common Bias of Uniformity

Paul Benacerraf in Mathematical Truth suggests that a semantic theory should be uniform for both mathematical discourse and everyday discourse. His idea, as I remember it, is that if we think of reference and satisfaction applying to talk of ordinary objects then we should think of them as applying to mathematical objects also. So that we should think of number terms as referential terms (at least if we think of ordinary terms as referential). I don't know much about the philosophy of math---take a look at Antimeta for some cutting edge stuff though--but I've often wondered how general such a claim should be--i.e. how much we should read ontology off surface grammar. Now I know, this isn't exactly the most original thought in the world--Quine, Carnap and all that. But my specific question is this: why should we expect a semantic theory to be uniform across different chunks of discourse. I mean I've always thought that there's one theory of truth for talk about chairs and tables and another one for talk about, say literature. For instance, when I talk about wine with some of my buddy's I tend to think that what determines the truth of what we say is very different--if equally objective--than what does when we talk about say where tables and chairs are (something we rarely do). Now I understand various ways one might get at a general worry about ontology and all that, but it doesn't seem like the idea that there has to be a uniform theory of meaning (or reference, or truth) is a very good one. In fact, I think every time one broaches a new subject the best thing to say about what the theory of truth is changes. Philosophical analysis is supposed to be the attempt at making the best sense of how we think about things--why the bias towards uniform accounts. And don't give me that science is uniform stuff. Science seems to me about as uniform as the contents of a typical issue of the Philosophical Review, think of the worlds apart between comp sci, quantum mechanics, evolutionary psychology et cetera. Well, I better let this go...