The Weekly Newsletter of MIT Linguistics

LF Reading Group 5/6 - Mitya Privoznov (MIT)

Speaker: Mitya Privoznov (MIT)
Title: Structural islands and discourse anaphora
Time: Wednesday, May 6th, 1pm – 2pm

Abstract: Let us define discourse anaphora as a referential dependency between an indefinite noun phrase and a pronoun like in (1) which could be established across a sentence boundary. Descriptively speaking, the indefinite introduces a discourse referent that the pronoun picks up.

(1) a. A person who came in with a woman1 offered her1 drinks.

b. *A person who came in with her1 offered a woman1 drinks.

Looking at the contrast in (1) one might think that for this relation to hold the indefinite must linearly precede the pronoun. However, it is an established fact in the quite extensive literature on anaphora that discourse cataphora is also in principle possible, like in (2a), but not in all syntactic configurations, as (2a) stands in a contrast to (2b).

(2) a. The teacher said that she called his1 parents, after she caught a student1 smoking.

b. *His1 parents said that they went to the teacher, after they caught a student1 smoking.

The question of interest to me in this talk is when discourse anaphora is in principle possible and when it is not? That is, what explains the contrasts like (1-2)? There are theories of binding (especially within the dynamic framework) which can explain examples like (1-2). However, to my knowledge they provide different explanations for each contrast. This view seems to me to be missing or rather not taking into account one potentially important syntactic generalization. Namely, that discourse anaphora always obeys one structural condition. For a pronoun to be discourse anaphoric to an indefinite the constituent that contains the indefinite and c-commands a pronoun must be a maximal projection (aka structural island under a strict view of Condition on Extraction Domains: specifier, adjunct or conjunct). In my talk I will try to formulate, defend and explain this condition.

It is possible, of course, that this generalization is accidental and that the core explanation is semantic and different for each case. But I will try to see the data like (1-2) from a syntactic perspective, which seems to me to be an experiment worth undertaking. The data will come from Russian and English (elicited with small samples of speakers).