Ivan A. Sag (sag@csli.stanford.edu)
Fri, 24 Jan 97 17:24:36 PST
In the recent discussion about adverb extraction, Bob Levine has pointed out a number of consequences of any analysis of adverb extraction that places adverbs on the verb's ARG-ST list. Paramount among these is the worry that: >if you treat adjunct *clauses* as complements then you both lose the >possibility of an account of the antireconstruction cases and, less >exotically, you lose a good story about the contrast between (1)a and >b: > >(1)a. *You can't tell them{i} that the twins{i} are being offensive. > b. You can't say anything to them{i} without the twins{i} getting offended. > >and similarly for (2): > >(2)a. *I told them{i} about the twins'(i} BIRTHDAY. > b. I only get them{i} presents on the twins'{i} BIRTHDAY. > The kind of account Bob has in mind uses relative obliqueness relations defined on ARG-ST. True complementation should induce o-command relations, hence blocking (as Principle C violations) any example like (1)a or (2)a, where a pronoun o-commands a coreferential NP (cf. the definition of Principle C in Pollard and Sag 94 - chap. 6). On this view, the grammaticality of (1)b and (2)b is to be explained by treating these in terms of head-adjunct structures that do not induce o-command relations. Hence Bob raises the perfectly reasonable objection that all `adverb-as-argument' theories (like the ones Gosse and others (myself included) have been investigating) will incorrectly let adverbs enter into obliqueness relations, just as complements do, hence blurring the distinction that is crucial for an account of the binding-theoretic contrasts in (1) and (2). I just wanted to point out that the assumption Bob is making (and that Pollard and Sag 94 also made), namely that pronoun-antecedent relations are mediated by some kind of grammatical structure (whether it's phrase structure or argument structure is immaterial to the point I want to make here), has been quite controversial in the literature. In particular, Bolinger (79), McCray (80), and Bresnan (ms) [see also work by Reinhart] have argued that an orthogonal functional dimension should explain data of the sort Bob brings up. Bolinger (1979:306) [see also the discussion in Bresnan ms.] states the following principle (`Reidentified' means `renamed by a full NP instead of a pronoun'.): Noncorefernce Principle: The topic may be reidentified easily in the theme, but in the rheme only if the theme lacks a normally topical form (subject noun or subject pronoun). Now this may all seem a little hard to pin down, but so is the relevant data. Compare the following examples to those cited by Bob (I draw freely from the sources cited above): (3) He-i did what John-i always does... (4) She-i was told that if she wanted to get anywhere in this dog-eat-dog world, Mary-i was going to have to start stepping on some people. (5) The teacher warned him-i that in order to succeed Walter-i was going to have to work an awful lot harder from now on. (6) It was rather indelicately pointed out to him-i that Walter-i would never become a successful accountant. (7) If you try to tell him-i that the reason why John's-i dog was taken away from him was rabies, he'll get very upset. I believe all of these examples are perfectly grammatical on the indicated readings. And though I'm not sure exactly how to apply Bolinger's account to all these cases, I believe it's true that these stand as counterexamples to all extant binding theories, including that of Pollard and Sag (1994: ch. 6) and the modified version of it that (I believe) Bob has in mind. So, though I realize this is far from conclusive, I just thought it should be brought into the discussion that perhaps `Principle C effects' have, as many have suggested, a more functional explanation. If this is right, then binding theory would provide no argument whatsoever against an account of adverb extraction that places adverbs on the ARG-ST of verbs. Best, Ivan
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:36:03 PST