Encoding Actions and Verbs: Tracking the Time-Course of Relational Encoding During Message and Sentence Formulation

Two experiments tracked the encoding of relational information (actions at the level of the prelinguistic message and verbs at the level of the sentence) during formulation of transitive event descriptions (e.g., The tiger is scratching the photographer). At what point during message and sentence formulation do speakers encode actions and verbs? Participants described pictures of transitive events in response to neutral questions ("What is happening?"), agent questions ("What is [the agent] doing?"), and patient questions ("What is happening with [the patient]?"). The agent and patient questions were intended to change the message-level focus of speakers' responses and to induce priority encoding of the event action and the sentence verb. The questions had a nearly categorical effect on speakers' choice of sentence form in their responses (characters mentioned in the questions were produced in subject position. as expected) and a strong effect on the time-course of sentence formulation: Speakers rapidly directed their gaze to the part of the event needed to encode contextually new, task-relevant information-first the event action and the sentence verb, then the sentence object. The distribution of fixations during the "verb-encoding" window showed that speakers encode relational information by fixating both event characters. Comparing formulation of sentences describing events with action-informative agents and action-informative patients showed a small preference for fixating the more informative character both immediately after picture onset (message-level encoding) and during the verb-encoding window (sentence-level encoding). The results identify action-specific and verb-specific eye movement signatures in message and sentence formulation.