I'm in the DallasFortWorth airport, en route to a better place. But on the plane here, I reread Gil Harman's paper "Meaning and Semantics." Wish I had done this before my previous post. But in any case, it was time well spent. I hadn't looked at that paper in ages. Norbert mentioned it a couple weeks back. I vaguely remembered disagreeing with Harman, which should have made me suspicious, since I usually don't.
What I had forgotten is that when I first read "Meaning and Semantics," I was far friendlier to truth-theoretic semantics than I am now. (I was young, and back then, all the kids were annotating trees with Tarskian graffiti.) When I reread the paper, I realized that it must have made a deep impression on me. For a while now, I've been unconsciously channelling Harman with respect to being less than overwhelmed by (i) general objections to theories of meaning that connect grammatical structures with mental representations, and (ii) the case for thinking that a theory of meaning needs to take the form of a Tarski-style theory of truth. (Mea culpa, Gil.) I'm still less optimistic about the role of conceptual roles than Harman was. But the bulk of his paper now strikes me as blisteringly clear and fundamentally correct. In particular, I think Harman was spot on about the (unwise) speed with which people inferred from the inadequacy of a mere translation manual to the inadequacy of any kind of translational theory of meaning. And he saw, pretty much right away, the need to not conflate semantic theories (in the Tarski-Lewis sense) with empirically motivated theories of meaning, regardless of what one thinks about how meaning is related to truth.
If you're interested in these issues, I highly recommend (re)reading Harman's paper.