Speaker: Jennifer Cole (University of Illinois, Urbana-Champaign)
When: Friday, February 22nd, 3:30 pm
Where: Education room 433
Title: Memory for Prosody
Abstract: Lacking the automatic playback of audio recording devices, humans recall a previously heard utterance on the basis of a cognitive representation of the linguistic object encoded in memory. When repeating or imitating a heard utterance, this representation serves to guide the phonology and phonetics of the spoken output. A number of recent studies show that the memory encoding of perceived speech includes (sub‐phonemic) phonetic detail that reflects the individual speaker’s voice and variation in the phonetic implementation of phones. The study presented in this talk asks about the memory encoding of prosody: Does the cognitive representation of the prosodic form of a heard utterance specify the abstract phonological features that encode pitch‐accents and prosodic phrase boundaries? Does it include speaker‐and utterance‐dependent phonetic detail?
Two different production tasks are used to explore the prosodic aspect of these representations: an imitation experiment in which speakers heard and then imitated spontaneous utterances from a Maptask corpus, and a read enactment task in which speakers read the same sentences aloud from text presentation. For each task, the resulting utterances were compared for similarity to the original Maptask utterance in their phonological prosodic features (prominences and boundaries) and in acoustic measures of the phonetic cues to those features. The main empirical findings from these studies are (i) imitators reproduce the phonological prosodic features more reliably than their specific phonetic cues, and (ii) the prosodic form of read‐enacted utterances is phonologically and phonetically more variable across speakers than with imitated utterances. These findings, considered in the context of our ongoing work on prosody perception, support a complex model of prosody encoding: Listeners encode prosody in terms of both phonetic detail and abstract prosodic features, but abstract features play a privileged role in later tasks involving recall for reproduction.