Implicit Arguments in Sentence Processing

Abstract

Some arguments of verbs appear to be represented semantically even when they are not expressed in a sentence. We investigated a particular example of this phenomenon, short verbal passives (e.g., The game show′s wheel was spun), where intuition suggests that our understanding includes an implicit agent (e.g., the agent of the spinning event). In order to test the hypothesis that an implicit agent was present in short passives, short passive, full passive, active declarative and intransitive clauses were followed by rationale clauses (e.g., ... to win a prize and lots of cash), which are only felicitous when the clause that they modify provides an agent. In a word-by-word, stop-making-sense task, rationale clauses were difficult to process after intransitives only when the subject noun phrase was not a good agent for the event in the rationale clause (Experiment 1). Rationale clauses following short passives with the same explicit propositional content (i.e., the same subject and main verb) as intransitives with poor agents were considerably easier (Experiment 2). Moreover, rationale clauses were no more difficult to process after short passives than after full passive or active clauses that provided a semantically appropriate agent for the rationale clause (Experiment 3). Experiment 4 demonstrated that the stop-making-sense task was sensitive to small differences among simple bridging inferences. The results are interpreted as evidence that implicit agents are routinely encoded.