Tibor Kiss (tibor@heidelbg.ibm.com)
Tue, 21 Jan 1997 15:24:21 +0100
-- =========================================================================== PD Dr. Tibor Kiss IBM Germany Institute for Logic and Linguistics Vangerowstr. 18 69115 Heidelberg +49-6221-594483 (phone) +49-6221-593200 (fax) tibor@heidelbg.ibm.com ============================================================================
Dear (lexical) Rulers, I have two comments on the discussion about adjunct extraction. Perhaps the ackem (AGKMS) community can say more about French, but I still try. 1. The Unicity Property/Iterability of Adjuncts Bob has raised the issue of unicity. I haven't read the AGKMS papers, but I am surprised about the idea of taking the non-iterability of an adjunct as an argument in favour of putting the adjunct on a SUBCAT list. Particularly, I am wondering how the AALR accounts for the non-iterability of an adjunct. It seems to me that the AALR would need something similar to an "occurs check" added to it as a constraint. Something like X can only be added to the COMPS list if X is different from all Y already appearing on the SUBCAT list. The question then is: "Different with respect to what?" Assuming that the AALR only adds synsem-objects to the SUBCAT list, only material within SYNSEM is available for comparison. Assuming further that the synsem of "nullement" and "pas" is pretty similar, how would such an approach account for the contrast in (1)? (1) a. *Il ne pas pas venu. b. ?Il ne nullement pas venu. My french informant told me that she never would use (1b) but still that it is much better that (1a). Perhaps then, the occurs check must make reference to an articulated theory of lexical semantics, finding a hidden difference between nullement and pas ... As long as such an articulated theory is not present, the occurs check cannot be formulated. Moreover, the very necessity of such an occurs check turns the unicity argument against itself: Even if certain adjuncts behave like complements on the surface, the theory accounts for them in quite different ways: While "additional" complements are excluded by a violation of the subcategorization principle, the non-iterability of the adjuncts has to be accounted for by stipulating an additional occurs check. Ignoring the contrast between (1a) and (1b), the best way out would be to assume that "pas" is an optional argument of finite verbs in French. But then, this element does not count as an adjunct any longer and hence the whole phenomenon cannot be taken as a paradigm argument for the assumption that adjuncts are added to SUBCAT lists. Arguments in favour of this position can easily be given: Firstly, negation particles in other languages (such as German or English) can be iterated over and over. Given an approriate contexts like the claim "Er ist nicht gekommen." (He did not come.), the denial would be: "Er ist nicht nicht gekommen." (engl.) "He did not not come."). Secondly, other adverbs in French behave like adverbs should, e.g. frequency adverbs. They can be iterated in finite contexts (please note that it makes perfect sense to iterate frequency adverbs if the events quantified over are of different domains, similar to the iteration of local adverbs in examples like: "Fred handed the toy to the baby in a restaurant last Sunday in Back Bay."): (2) Il est souvent/rarement venu souvent/rarement. He has often/rarely come often/rarely. To sum up, as long as the non-iterability of French "pas" is a cornerstone of the adjunct-on-SUBCAT-theory, I don't find it convincing. 2. Adverbs and Modal Verbs Gosse has remarked that one of the arguments in favour for adjunct->SUBCAT is the scope of adjuncts in contexts of argument inheritance in Dutch. This argument carries over to German as well. Since the adverbs are added lexically, their scope is also defined lexically and hence by definition is always narrower than the scope of nominal complements of the verb. At least for German (and I assume for Dutch, too), the prediction is not borne out by the facts, witness (3). (3) Er hatte oft einen Jungen verletzt. He has often a boy hurt (3) can only have the interpretation often(a boy(he hurt him)). The LR-account would predict the reading a boy(often(he hurt him)) which is not possible. A way out would be to generalize Pollard's and Yoo's approach to quantifier scope to adverbs, but this would amount to Gosse's suggestion to use a non-configurational analysis of these facts (which would then lose any force as an argument in favour of adjunct->SUBCAT). As it stands, however, the approach makes the wrong predictions w.r.t. scope relations between arguments and "adjuncts". Again, I don't find the argument pretty convincing. A small remark: The binding differences between arguments and adjuncts noted by Bob are quite vivid in German, too. Greetings from the most active room on this floor. Tibor
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:36:03 PST