Hall, K. . & Yoshida, M.
Northwestern University
During the processing of ellipsis, the parser must recognize an ellipsis site, access the antecedent, and integrate the antecedent material into the current structure. The current study explores how antecedent size may affect processing. Recent studies on antecedent retrieval have investigated whether syntactic structural information of the antecedent is retrieved at the ellipsis site. Previous results suggest that length or complexity of an antecedent does not affect processing of ellipsis, lending support to retrieval theories that assume a cue-based pointer to content-addressable memory. We conducted an eye tracking while reading experiment to investigate length effects in ellipsis processing, and manipulated Ellipsis Type (sluicing vs. sprouting) and Antecedent Length (bi-clausal vs. mono-clausal) in a 2x2 design. Our results indicate that both Ellipsis Type and Antecedent Length affect the processing of the ellipsis site (interaction: t=2.43, p<.05): while sprouting conditions were marginally harder to process than sluicing constructions (main effect: t=1.8, p=.08), larger antecedents resulted in more processing difficulty than shorter antecedents (main effect: t=-2.07, p<.05). The finding that antecedent retrieval is sensitive to antecedent length directly contradicts predictions of unconstrained cue-based theories, and support retrieval models that are sensitive to structural cues.