The Weekly Newsletter of MIT Linguistics

LingLunch 2/25 - Paul Crowley

Speaker: Paul Andrew Crowley (MIT)
Title:Imparallel VP ellipsis
Date/Time:Thursday February 25/12:30pm-1:50pm
Location: 32-D461

This talk will be concerned with a class of VP ellipsis expressions illustrated by the sentence in (1) where VP ellipsis is licensed despite an imparallelism between the antecedent VP and the interpretation the ellipsis site. CAPS indicates obligatory contrastive accenting.

(1) JOHN is expecting NOT to pass the exam but MARY IS .

The imparallelism is attributed to the presence of the negation within the matrix VP in the first conjunct, which is acting as the antecedent to an elided VP that does not contain that negation. Under no formulations of the identity conditions on ellipsis—whether syntactic or semantic—is the ellipsis in (1) expected.

The proposed analysis of the effect, which will be referred to as Imparallel VP Ellipsis (IVPE), will not treat it as a case of non-identity tolerance but rather as an illusion of non-identity created by an LF opacity effect. It will be proposed that the problematic negation in the first conjunct is situated outside of the antecedent VP at LF, where the identity conditions on ellipsis are taken to apply. It will be shown that this assumption is necessary in order to resolve an additional problem of imparallelism that arises where IVPE appears in ACD environments in which VP ellipsis and an NPI are both licensed despite having prima facie conflicting scope requirements.

Assuming that the negation is really situated high at LF, the task of deriving IVPE expressions will then be split into two pieces. The first is to explain why the negation is pronounced low when it is high at LF. The second is to explain why is the negation interpreted with narrow scope when it is high at LF. An account of the first question will be shown to come from the assumption that there is a syntactic Neg-raising operation active in a single-output grammar that creates Neg copy chains, which are interpreted separately by each interface component. An answer to the second question will come from a generalization that will be observed for the IVPE phenomenon where the effect is only felicitous if the verb heading the antecedent and elided VPs is a Neg Raising verb. By treating the Neg Raising phenomenon as the result of a pragmatic strengthening effect which gives the negation narrow scope in the truth conditions post-derivationally, we can account for the disconnect between the LF and output truth conditions in IVPE expressions. Independent evidence will be provided for this approach to Neg Raising, which involves cases of VPE similar to (1) but lacking the problematic imparallelism.

Finally, two points of overgeneration will be shown to come from the assumptions used here in light of the Neg Raising generalization on IVPE and a means of overcoming them will be proposed by way of pragmatic principles.