The effects of prosody on on-line parsing were examined in ten experiments. Eight experiments used coordination items in which a relative clause (RC) was temporarily ambiguously attached to coordinated Noun Phrases (NPs):
The lawyer greeted the powerful barrister and the old judge who was/were walking to the courtroom.
A production study showed that subjects produced prosodic cues distinguishing the two structures before their eventual disambiguation. Two off-line written completion studies established that subjects preferred attachment of the RC to both conjuncts, regardless of order of NPs, and that commas were partially able to reverse this preference. An auditory off-line forced-choice disambiguation study showed that prosodic cues were able to a larger extent than commas to reverse attachment preferences.
Four on-line cross-modal naming studies, paired with subsidiary appropriateness rating tasks, showed that subjects were able to integrate the prosodic cues in the syntactic representation of spoken sentences. Moreover, it appeared that this integration occurred early in processing, as prosody's effect was reduced over time. However, prosody still affected appropriateness ratings. It was also shown that the effect was not confounded with temporal pattern of the prosody.
The final experiments employed shorter ambiguous sentence fragments (He recognises...), excised from temporarily ambiguous sentences in which verbs permitted direct object or sentence complement continuations:
He recognises them and he greets them./ He recognises they try to greet him.
A cross-modal naming study found no evidence that prosody cued syntactic structure, but did show that short verb durations led to the activation of both subcategorisation frames, whereas longer verb durations allowed time for direct object preferences to develop. An auditory off-line forced-choice disambiguation experiment confirmed the inability of prosody to cue syntax in these items.
KEYWORDS: INTONATION, PUNCTUATION, PITCH, FUNDAMENTAL FREQUENCY, CONSTRAINT
SATISFACTION, GARDEN PATH, MINIMAL ATTACHMENT, LATE CLOSURE.
Last updated 30 March 1998