This site will look much better in a browser that supports web standards, but it is accessible to any browser or Internet device.

Graduate Program in Linguistics at the City University of New York

Abstract for Andrea Martin's talk

Comprehension of verb phrase ellipsis: Speed-accuracy tradeoff evidence for a content addressable pointer
Andrea Martin (New York University)
September 12, 2006 (Tuesday)
6:30 PM - ; Room 7102, The CUNY Graduate Center

Four speed-accuracy trade-off (SAT) experiments investigated the retrieval and interpretation processes involved in the on-line comprehension of verb-phrase ellipsis (VP ellipsis). To determine whether a content-addressable retrieval mechanism is used to find antecedents in memory, we manipulated the amount of interpolated material (distance) between antecedent and ellipsis site. To determine whether interpretation of ellipsis involves a copy or a pointer mechanism, we manipulated the size of the antecedent. Experiment 1 showed a significant difference in the likelihood of successful retrieval between near and distant antecedents. Distant antecedents were less likely to be retrieved than near antecedents, possibly due to retroactive interference. However, there were no differences in speed of retrieval between near and distant antecedents, as would be predicted by a serial search process. This finding therefore suggests that the mechanism that subserves the retrieval of the antecedent is content-addressable. Experiments 2 and 3 manipulated antecedent size to test the pointer hypothesis. Experiments 2 and 3 found that complex antecedents were no more likely to be retrieved than simple antecedents, and more importantly, that there was no difference in retrieval speed. The lack of SAT dynamics differences suggests interpretation of VP ellipsis involves a pointer to extant structures in memory rather than a copy or structure building mechanism at the ellipsis site. Experiment 4 investigated whether this content-addressable pointer mechanism mediates ellipsis when variables within the antecedent must be bound to new complements at the ellipsis site. We again found no differences in retrieval speed, suggesting that a pointer mechanism mediates interpretation of VP ellipsis even under increased binding constraints.