Robert Levine (levine@ling.ohio-state.edu)
Tue, 28 Jan 97 17:10:58 EST
So apparently the elimination of traces in favour of a valence-reduction mechanism has brought us to the point where what would once upon a time have been a set of separate lexical entries now consists of a single ur-entry that can be fleshed out in accordance with a potentially infinite set of dijunctive constraints. But once you start going down that route, what *principled* reason do you have for ever distinguishing between two lexical entries, ever? For example, once upon a time we would have regarded _give_ as corresponding to two distinct addresses in the lexicon, distinguished by their COMPS specification. Now we can have instead a single lexeme _give_ which satisfies the disjunctive constraint [COMPS<NP, PP]>] \/ [COMPS<NP, NP]>] I know that this isn't what Ivan said in his last posting, but I don't see what there is in the logic of the proposal he was citing that allows you to treat multiple verbs projected from a single lexeme, distinguished both in their adjuncts-valence property and their semantics, as a single lexical entry while requiring that a pair of verbs, distinguished *only* in their complements-valence property, have to be distinct. And why stop there? Since there's nothing privileged about the PHON attribute, as vs. SYNSEM|CAT|HEAD|VAL|COMPS, why could we not say that _go_ and _went_, or indeed any two suppletive forms, were a single lexical entry projected under the appropriate conditions from one lexeme underspecified for PHON, subject to disjunctive constraints which, inter alia, specified the phonetic realization of this projected word either as /gow/ or /wEnt/? What blocks identification of the active and passive forms of transitive verbs as the same lexical item? For that matter, what blocks the implosion of the lexicon into a single entry, subject to just the set of constraints that would realize versions of it as all of the various words in the language with their combinatorial possibilities, and only those words? One consequence of this dissolution of the lexicon is that the kind of evidence the AGKMS group has proposed for the complementhood of French negative adverbs is now absolutely irrelevant. Since a lexeme can be projected to a single word which satisfies a disjunction of constraints on how it can be realized in a feature structure, the argument that verbs which occur with _pas_ etc. have different subcategorization properties from those which don't now signifies exactly nothing about the distinctness of the lexical entries for the former and latter respectively. We have no a priori reason to view the difference in selectional properties between verbs cooccurring with negative advers vs. verbs cooccurring with nonnegative adverbs as anything more than the manifestation of a set of possible conditions on the realization of a *single* word in syntactic representations, where both negative and nonnegative adverbs are equally entitled to be part of the COMPS list. So the fact that propositional attitude verbs, which take an indicative complement clause when unnegated and either an indicative or subjunctive complement when negated does not, in fact argue for two verbal forms on this approach, contrary to Abeille and Godard's view in their paper `The syntax of French negative adverbs'; they are no longer allowed to reach that conclusion from this kind of evidence, because on the approach Ivan is advocating the evidence simply points to a disjunctive constraint that each of these verbs, taken to be a single word, can satisfy: COMPS[S[ind]] \/ COMPS[adv, S[ind]] \/ COMPS[adv[neg], S[subj]]. The dichotomy between the difference in selectional possibilities in the case of negative adverbs on the one hand and the lack of such difference in the case of nonnegative adverbs is essentially buried in this characterization of the constraints on verbs of propositional attitude; all that it means is that there is an additional constraint available which happens to explicitly mention the [neg] value on the adverb, whereas there is no such constraint that mentions a [nonneg] value on the adverb---a purely accidental fact with no systemic implications for anything, as far as I can tell. In fact, at this point it becomes very difficult to see what the empirical status of the notion `distinct lexical item' is. Bob
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:36:03 PST