Abstract
We present two experiments involving true scenes (motion pictures of events), manipulating verb class, sentence semantic context, and scene motion context. Experiment 1 constituted a replication of [reference omitted, 2019]. Participants (N=32) were presented with sentences containing either a causative or a perception/psychological (experiencer) verb (e.g., Before making the desert, the cook will crack/examine the eggs that are in the bowl). Scene context varied according to the action performed by the agent (cook), moving towards the target object (eggs), away from it, or remaining neutral. Results were similar to those obtained by our previous study: a main effect of motion and no main effect of verb type. We obtained faster saccades to the target object in the causative sentences than in the experiencer sentences, but only in the towards motion condition. As in our previous study, we did not find anticipatory effects to target objects. In Experiment 2 (N=46), in addition to verb type (causative vs. experiencer) and agent motion (towards vs. neutral) we introduced a sentence context manipulation, with the first clause denoting either a semantically restrictive activity (e.g., In order to make the omelet...) or a non-restrictive one (e.g., After pouring the flour into the bowl,...). We predicted that the stronger context would enhance attention to properties of verbs making the potential referents of their objects more salient, thus driving anticipatory effects found in other studies. We found an effect of motion and semantic context, with restrictive sentences driving faster saccades to target objects. We also found a verb effect with causatives yielding faster saccades than experiencer verbs only in the towards condition, but no anticipatory effects. These results suggest that early linguistic and visual processes are largely independent, interacting at a later, conceptual stage. We propose that the two systems interact using a common propositional code.