Ellipsis and Acceptable Ungrammaticality

Monday, 25 January 2010, Colloquium

Lynn Frazier, University of Massachusetts, Amherst.

Abstract

Elided phrases typically have a syntactically matching antecedent (morphological features aside). However, naturally attested examples occur where the antecedent and elided clause do not match syntactically, e.g., in voice as in This information needed to be released, but Gorbachov didn’t. (Hardt 1993, Kehler 2002) In earlier work (Arregui et al., Frazier 2008), my colleague Chuck Clifton and I have argued that the processor repairs mismatching antecedents at LF using the same repair mechanisms needed for garden-path sentences. According to this ‘Recycling hypothesis’ such mismatches are relatively acceptable when they correspond to common syntactic blends, and thus sound familiar to the listener.

In the talk, mismatch ellipsis will be argued to be most acceptable when attention is drawn away from syntactic form. Some clauses contain words such as the verb want in Ben wanted to go to Paris which support an implicature that the state of affairs described does not hold in the actual world. Such Non-Actuality Implicatures (NAIs) facilitate mismatch ellipsis because they implicitly focus on the modality or polarity of the state of affairs described rather than focusing on the description of that state of affairs. The contrast between the state of affairs described and the actual world serves as an implicit focus and serves to increase the salience of the clause that conveys the NAI, as well as constraining likely discourse continuations. The results of several experiments will be presented in support of this claim (Grant, Clifton and Frazier, in progress). They show that NAIs increase the acceptability of mismatch ellipsis more than the acceptability of other sentences such as sentences with matching ellipsis or no ellipsis. They also show that, in sentences with more than one potential antecedent, the probability of choosing an antecedent in a particular clause is increased if the clause contains an NAI trigger, as would be expected by the ‘implicit focus’ hypothesis laid out above. The fact that NAIs facilitate mismatch ellipsis most fits with the assumption that mismatch ellipsis is ungrammatical, though acceptable under favorable circumstances.

The Recycling hypothesis will be extended to cover examples of ellipsis with split-antecedents, as in Bruce wants to sail around the world and Wendy wants to climb Kilimanjaro, but they can’t [ ] because money is too tight (Webber 1978). Split-antecedent ellipsis will be argued to be ungrammatical, but acceptable under circumstances where the antecedent can be recycled, e.g., when the source clauses for the antecedent are highly parallel. They will be analyzed as the product of the same mechanisms that apply in the conjoined VP interpretation of examples with conjoined clauses: Ian travels in summer and he stays home in winter. George does too.

This approach to ellipsis allows a maximally simple grammar of ellipsis while also providing a detailed account of the range of acceptability found in mismatch ellipsis. Further the same kind of mechanisms that account for mismatch ellipsps are needed elsewhere. Time permitting, the results of a study on double quantification will be presented showing that listeners and readers compensate for likely speech errors, resulting in ungrammatical interpretations being assigned to perfectly grammatical sentences with computable ‘normal’ interpretations.