Prosody Influences Real-Time Processing When Grammar Allows but Plays Less of a Role in Final Interpretation: An Eye-Tracking Study on VP Ellipsis and Stripping.

Abstract

This study examines whether prosody affects how people process and interpret two types of sentences that both have missing material but differ in ambiguity. Stripping sentences (e.g., Grandma hugged the kid at home and Grandpa too) are ambiguous because the argument after and can be read as either the subject of the missing verb or the second object of the audible/visible verb. In contrast, verb phrase (VP) ellipsis sentences (e.g., Grandma hugged the kid at home and Grandpa did too) are unambiguous, only allowing a subject reading. In a visual-world eye-tracking experiment, 43 English speakers heard VP ellipsis and Stripping sentences, with either the subject or object in the first conjunct made more prominent through prosody. The analysis of their mouse-click responses shows that prosody did not change their final interpretations: They always preferred the subject reading for VP ellipsis sentences and exhibited no preference for Stripping sentences, no matter which argument was prosodically prominent. The eye-movement data for VP ellipsis sentences also display a consistent pattern of early looks to the subject-reading image. However, the eye-movement data for Stripping sentences reveal a prosody effect: After sentence offset, listeners briefly showed divergent looks toward either the subject- or object-reading image, depending on whether the subject or object carried prosodic prominence. These findings suggest that prosody can influence temporary interpretations during real-time processing but has little impact on final interpretations.