Philosophy 305                                                                                                                                                                            WGL
Spring, 2001

                                                                KATZ' MANY (ALLEGED) EXPLANANDA

     I complained in class that Katz is drastically unclear as to what a semantic theory of his sort is supposed to explain.  And the differences matter.  I'll document that a bit here.

     But first, I return to Ted's question: What has Katz' paper to do with concepts, either in M&L's special sense or in a more general sense?
     Directly, nothing.  Katz sketches a linguistic theory; its subject-matter is English, a public natural language. So far as it has to do with human psychology, it is a theory of language understanding and language processing.  If we construe it as a cognitive theory, it will posit mental representations, all right, but those representations will be representations of phrase markers, syntactic categories, semantic properties such as entailment, and so forth--not representations of cows or cats or can-openers, not concepts of the kind that concern M&L.
     Rather, it seems M&L included it because they take it to have suggested the Classical Theory.  I don't recall Katz' talking of necessary and sufficient conditions, but we may imagine that his kind of dictionary entry for an English morpheme could be rewritten as a set of necessary and sufficient conditions.  Thus, (4.10) on p. 135 suggests:

 x is a chair <=> x is furniture & x is portable & x has legs & x has a back & ....
     Assuming he'd accept that, then he is saying of something that it's a set of necessary and sufficient conditions.  But what he's saying it of is a word meaning.  To make this into a theory of concepts in any psychological sense, at least a bit of work would have to be done.  (N.b., though, since word meaning is a perfectly good thing to mean by "concept" if one chooses to, Katz' theory is in that way a theory of concepts.)  Remember DORIT'S MAXIM:  Don't simply assume that a property of language or of a linguistic feature has a psychological correlate, or vice versa.
     The quickest way to make Katz' theory into a theory of concepts in a psychological sense is to posit that concepts in that sense just are mental entities corresponding to Katzian word meanings.  If we think of his dictionaries as being psychologically real (realized in our heads), then the word meanings are real too, but they can't very well themselves serve as concepts, because they're passively impacted within the speech center; the semantic processor uses the dictionary as a lookup table, and that's about it.  Rather, there would have to be doppelgängers of the word meanings, that played a more active and constitutive role in ordinary thought about ordinary things.  And what would be the psychological argument for the claim that concepts are doppelgängers of dictionary entries, even if we're convinced that Katz' dictionaries are real?

     But back to Katz' paper.
     As I said, there is nearly constant unclarity as to whether Katz is (a) talking about actual cognitive processing, (b) offering a loose rational reconstruction of what might be going on, or (c) not talking about psychology at all, but only offering an abstract quasi-mathematical theory of semantic properties and relations.  In his later book, Language and Other Abstract Objects, he vigorously repudiated any psychological interpretation of syntax or semantics, and embraced (c); but in the 1972 work we've read he hadn't made up his mind.
     P. 127:  He says we want an "explanation of the correlation of semantic and phonetic interpretations."  Psychological connection, or merely abstract formal correlation?
     P. 128:  He says we want the theory to spit out a recursive enumeration of the set of possible senses.  That sounds more abstract and mathematical than psychological, but it could be abstracted from a serious psychological idea.
     Further down p. 128:  He says the explanation "must assume that the speaker possesses" semantic rules "that enable him to obtain the meaning...."  So now we're explaining something about the speaker, a presumed cognitive achievement.
     But then on p. 129:  "The basic task of this theory...is to explain...each...semantic property and relation."  One can (so far as has been shown) do that without any adversion to psychology at all.  And that's what it looks like the apparatus toward the end of the paper is doing (pp. 142ff.).
     P. 131:  We're to "explain how a speaker is able to understand sentences."  Then in the following sentence, we "must reconstruct the semantic knowledge an ideal speaker-hearer has...."  Whoa, knowledge all of a sudden?  And how ideal, and in what direction of idealization, is ideal?
     P. 148:  He says the semantic theory will make predictions about speakers' judgments about semantic properties and relations.  No doubt that's true if by "judgment" he means (in Sellars' useful terms) the judgeds.  But if he means the speakers' judgings, uh-uh.  Judgings are physical actions performed in real time by real people; a purely semantic theory couldn't predict those, or even generalizations about them.

     Well, you get the idea.  Different explananda can call for very different explanatory apparatus and resources.  In particular, explaining either behavior or cognitive capacities is a much more concrete sort of task than just explaining semantic properties and relations of linguistic expressions.  Though a theory of the latter might play a key role in a psychological theory, it couldn't possibly do so alone.