I was reading Margolis and Lawrence’s introduction to their anthology Concepts a couple of weeks ago. I got the impression that (perhaps for limitations of space) they didn’t do justice to the view that concepts are best thought of as abstracta, and especially the argument from what they call “the idiosyncrasy of mental representations” to that view. I want to do justice to that argument here, and maybe also air out my thoughts about the metaphysics of concepts in the process.
Consider two speech situations, S1 and S2. The utterances and contexts in S1 and S2 are entirely identical, except for the identity of the speaker – A in S1 and B in S2. Let’s assume that their utterances contain no first person pronouns or other self-referential indexical elements, such as “Some dogs have no hair” or “Would you like more hot pepper?” We have an inconsistent tetrad of propositions:
(1) A’s utterance in S1 means the same as B’s utterance in S2.
(2) A’s mental representation of the meaning of at least one sub-sentential expression in her utterance (“dogs,” “hair,” “hot pepper”) differs greatly in type from B’s mental representation of the meaning of that same expression.
(3) For any two sentential utterance-tokens U1 and U2, ceteris paribus, if U1 and U2 contain one expression-type e, such that the token of e in U1 and the token of e in U2 do not express significantly type-identical concepts, then U1 does not mean the same as U2.
(4) Concepts are (types of) mental representations of the meanings of sub-sentential expressions.
Another slightly different version of the tetrad, corresponding to another slightly different version of the view that concepts are mental representations replaces (2) and (4) with, respectively:
(2`) A’s mental representation of what A believes to be the extension of at least one sub-sentential expression in her utterance (“dogs,” “hair,” “hot pepper”) differs greatly in type from B’s mental representation of what B believes to be the extension of that same expression.
(4`) Concepts of referential expressions are (types of) mental representations of the presumed extensions of referential expressions.
We could generate different “prime” versions of the basic tetrad for all sorts of different views of concepts and mental representation. (I’ll limit the discussion here to the first version, assuming that everything I have to say holds, mutatis mutandis, for the other versions.) Despite the vagueness of “significant difference,” I take it that all of these different versions are inconsistent; we have to deny at least one premise in order to avoid inconsistency. I also take it, however, that one ends up thinking about mental representations, (2) is supposed to be true by hypothesis. The question, then, is whether to drop (1), (3), or (4). Here is how I see the scene, now:
> I probably won’t give up (1). My intuitions – about utterance-meaning, anyway – are semantic-externalist enough at this point that (1) is, to me, pre-theoretically obvious. As far as theory goes, I think that linguistic conventions fix the semantic properties of utterances, and that whatever determines linguistic conventions, whatever those ultimately turn out to be, cannot tap so deeply into our minds as to draw distinctions in utterance meaning solely on the basis of differences like those between A’s mental representations and B’s mental representations. That said, I bet that all sorts of psycholinguists would prefer to drop (1).
> I would give up (3) if I thought that theories of concepts weren’t supposed to explain the semantic properties of utterances. Sometimes I think that is the case; sometimes I think that theories of concepts, as developed by psychologists, are just designed to explain typicality effects and related, more-or-less overt behavioral facts and dispositions. To the extent that the facts about typicality effects are not also facts about the meanings of words, this should lead me to drop (3). But I get the sense that I am really deeply confused about what theories of concepts are supposed to explain, and that my confusion obscures the role of concepts in semantic theory. That’s why I’m inclined not to give up (3).
> That leaves me with (4). I am motivated to reject (4) both because of the tetrad and because I don’t see a natural, sensible way to preserve locutions of the form “the concept of x” if we think of concepts as mental representations. I am motivated not to reject (4) by what I take to be the fact that most psychologists think of concepts as mental representations of one sort or another. On the plausible meta-philosophical view that what we philosophers are trying to do with our discussions of concepts and mental representations is to shed light on these terms as they are used by psychologists, it tells against an explication of “concept” if it does not also come with a truth- or plausibility-preserving interpretation of psychologists’ mental-representational talk of concepts. But what tells against an explication is not always what ultimately defeats it, and theories of concepts might be one of those areas in philosophy in which merely scrutinizing accepted scientific terminology ought to lead to broad change in the way scientists construct their theories. (How different would this be from the sort of verificationist thinking that motivated Einstein to develop the relativistic conception of simultaneity?) To what extent treating concepts as abstracta would lead to such a change is not clear to me, though, since there might be a simple, painless translation of most psychological, mental-representational talk about concepts into the “abstracta” idiom. (Perhaps the mental representations psychologists call “concepts” are best construed as mental representations of the abstracta that are called “concepts” in expressions of the form “the concept of x.”) Anyway, I lean towards giving up (4) because I like (1)-(3).