Overt prosody and plausibility as cues to relative-clause attachment in English spoken sentences

We investigated the interplay between overt prosodic cues and semantic cues on the structural interpretation of spoken sentences that permit either high- or low-attachment of a final relative clause. Prosodic cues were manipulated via the presence or absence of a strong prosodic boundary before the relative clause, and semantic cues were induced via plausibility restrictions (e.g., the servant of the actress who was { serving tea / very famous }). In the first two experiments, each type of cue was studied in isolation while keeping influences of the relevant other cue constant. Experiment 1 employed a standard off-line comprehension task and suggested that prosodic cues were not as effective as semantic cues in biasing participants’ attachment preferences. However, using a more implicit (and less biased) structural priming task, Experiment 2 showed that our overt prosody manipulation was actually no less effective than plausibility in biasing relative-clause attachments. Experiment 3 was, again, based on structural priming; here, the two factors were fully crossed to investigate the interaction between overt prosody and plausibility. This experiment showed that the two types of cues interact in a complex way, suggesting that (a) the amount of surprisal associated with cueing a generally dispreferred structure and (b) the type of revision necessary to resolve the ambiguity both play a major role in determining relative clause attachments.