Constant - I'm not sure I comprehend your distinction (could be lack of caffeine) but thanks for the recommendation.
poke - my friend likes to explain this to his undergrads by asking them how they would verify that a thermometer is accurate (check it against another thermometer, but how do you know that one is accurate . . . etc.) until they figure out that thermometers are only "accurate" according to custom or consensus. Then he asks them how they know their eyes work. And their memories.
Some of them cry.
Okay maybe not really. Anyway, aside from torturing undergrads, I agree that we could just as easily start to do science with "physical objects themselves as described in our best theories" - I just mean that in order to get those theories, we have to believe in our (or someone else's) perceptions, memory, ability to do logic, etc. Ultimately I don't think it's a problem, at least one serious enough to destroy science.
Also there's this cute little move from Warren Quinn where he asserts that no philosophical theory could ever convince him that his chair wasn't there, no matter how sound and careful the argument - because his sureness of his perceptions would outweigh any sureness he could possibly feel in any theory.
Any lit recommendations from poke?
poke, interesting that you're talking about the middle ground between skepticism and the infallibility of sense data, which is the sort of "defeasible warrant" idea - we DO take our sense data as a starting point (or, as you note, we take a theory as our starting point - though where did the theory come from, ultimately, but somebody's taking his sense data seriously and then interpreting them?) - but if we find a conflict, either among our sense data or between ours and somebody else's, we reduce our trust in our sense data (we don't just toss them out). People are starting to apply this line of thinking not just to straight epistemology but to ethics (and even aesthetics!).