Individual differences in the relation between perception and production and the mechanisms of phonetic imitation


This study uses phonetic imitation to understand more about how individuals perceive and produce speech and to explore the link between the two. Previous research looking for a perceptionproduction link within individuals has been mixed. We used phonetic imitation and manipulated stimuli with the goal of more directly probing the link and to test (1) whether individual listeners’ perceptual cue weights are related to their patterns of phonetic imitation and (2) the underlying mechanisms of phonetic imitation. Twenty-three native speakers of English completed a twoalternative forced choice identification task followed by a baseline production and a forced imitation task. Perception stimuli were created from productions of head and had recorded by a native speaker of English. Seven steps varying in formant frequency (created with TANDEM-STRAIGHT) were crossed with 7 duration steps (PSOLA in Praat). Imitation stimuli were a subset of stimuli from the perception task plus extended and shortened vowel durations. We found that natural, shortened, and extended vowel durations were imitated well, indicating fine-grained sensitivity for imitation. Natural formant frequencies were imitated well but ambiguous formant frequencies were not, suggesting an effect of phonological categorization. Thus, preserving phonetic details in imitation may depend on the nature of target stimuli. The results from the relation between cue weights and degree of imitation suggest that individuals with greater ability to use formant frequency (higher weights) in perception showed more imitation of vowel duration. This may indicate that better phonetic perception leads to more fine-grained imitation in dimensions that are not constrained by phonological categorization. Our results suggest that phonetic imitation is mediated in part by a low-level cognitive process involving a link between perception and production, as evidenced by imitation of vowel duration. As indicated by imitation of formant frequencies, however, this study also suggests that imitation is mediated by a high-level linguistic component, i.e., phonological contrasts, which is a selective rather than an automatic process.

International Conference on Speech-language Pathology and Audiology
University of Montreal, Montreal, Canada