Linguistic representations of visual events.

摘要

This thesis explores the nature of linguistic representations that correspond to verbal descriptions of events. In two experiments, participants watched captioned videos and decided whether the captions accurately described the videos. In the videos, two geometric shapes moved around the screen. [In half of the trials, the geometric shapes had "eyes."] The verbs used to describe the shapes' actions were either source-to-goal verbs (chase, follow, trail) or goal-to-source verbs (flee, lead, guide ). Sometimes the captions were active sentences (e.g., The circle is chasing the square) and sometimes passive sentences ( The square is chased by the circle). Analyses of participants' reaction times indicate that the level of linguistic and visual detail encoded reflected the complexity of the task participants had to perform. These results are consistent with "good enough" models of language processing (e.g., Ferreira and Henderson (2007)) in which people process sentences heuristically or syntactically depending on the nature of the task they must perform.