CNL Lunch

Elixabete Murguia
(w/ Amy Weinberg & Juan Uriagereka)

How is Verbal Ellipsis Processed?

Thursday March 20th, 12:30pm, 3416 Marie Mount Hall

In this work, I look at different elliptical constructions, namely Verb Phrase Ellipsis (VPE), Pseudogapping, and Gapping. These are exemplified in (1), (2), and (3) respectively. I address the questions of how those elliptical constructions are parsed, in other words, how the elided constituent is resolved; as well as why different locality restrictions apply (see (1b) and (2b) versus (3b)):
(1) a. John blamed Susan, and Bill did too.
b. John blamed Susan, and Peter thinks Bill did too.
(2) a. You don’t believe me, but you will the weatherman.
b. You don’t believe me, but I think you will the weatherman.
(3) a. Mary called Bill, and Beth John.
b. *Mary called Bill, and I think that Beth John.
The proposal I offer addresses the locality restrictions issue, and it is driven by efficiency and economy considerations. I assume a minimalist algorithm based on the operations of Merge, Move and Spell-out (as defined by Weinberg (2000)), and I account for the presence/absence of locality effects between antecedent-gap as a result of (i) auxiliary presence/absence (already discussed by Fodor (1985), Berwick and Weinberg (1985)), (ii) initial low attachment of coordinates, and (iii) multiple Spell-out operations which render syntactic material unavailable (Uriagereka (2000)). These last two offer a natural explanation for when left context (i.e. the antecedent) is available.
In the case of VPE and pseudogapping, since there is an auxiliary overtly realized, a verb phrase can be predicted without accessing the antecedent; consequently, the antecedent does not need to be local. In the case of gapping, however, there is no overt auxiliary or verb, and the antecedent must be accessed so as to interpret the gap. If the antecedent has already been spelled-out, it is no longer available, and the gap cannot be resolved—this is the case when the antecedent and the gap are not local (see (3b)).
To finish interpreting that VP which is predicted in the phrase-structure building component (both in VPE and pseudogapping), the antecedent is accessed, but no syntactic structure is copied from the antecedent into the gap. It is the antecedent lexical structure that is used to finish interpreting the elision site, along the lines of Lappin and McCord (1990). Thus, a two-fold process is proposed: (i) some syntactic structure is computed by using available local information, and (ii) the antecedent lexical structure is retrieved to finish interpreting the gap.