Dropping Bowling Balls on Tomatoes: Representations of Object State-Changes During Sentence Processing

Previous research showed that verifying a pictured object mentioned in a preceding sentence takes less time when the pictured object shape is compatible with the described object location or spatial position. In the current work we asked if nonvisual information is integrated into the mental model when the target object shape is implied by virtue of a description of a heavy versus light item being dropped on it. Furthermore, we asked if the canonical target object state continues to play an important role when the context requires the activation of a noncanonical representation. In seven experiments the data provide an affirmative response to both questions. Participants (N = 766) first read sentences that implied target object state-changes as a function of the impact caused by differently weighted items (e.g., "You drop a balloon/a bowling ball on a tomato") and then verified pictures of "squashable" target objects in either a canonical (e.g., intact tomato) or a noncanonical (e.g., squashed tomato) state. A reaction time (RT) advantage was consistently observed when a "noncanonical" target was preceded by a "heavy" (e.g., bowling ball) sentence than a "light" (e.g., balloon) sentence. However, no such advantage was observed when a "canonical" target was preceded by a light sentence than a heavy sentence. This pattern of results remained unchanged regardless of the items used and the verbal tense of the sentence. These data suggest that when changes of state are inferred (i.e., not driven by lexical semantics), both the initial and resultant states are equally accessible.