Lisa Davidson (NYU) will give the linguistics department colloquium on Friday, October 20, at 3:30 in the seminar hub. A title and abstract of the talk follows.
Stabilizing the production of non-native sequences with acoustic variability
In the processing of non-native consonant clusters, speakers’ systematic errors have often been attributed to the influence of universal factors or native-language phonotactics (e.g., Dupoux et al. 1999; Moreton 2002; Berent et al. 2007). There has been less focus, however, on whether speakers are also sensitive to fine acoustic details in non-native sequences. In this talk, I focus on two questions: (1) Do speakers use fine phonetic detail to determine what phonotactic structures are present in a non-native speech signal, and (2) Is sensitivity to non-contrastive phonetic detail attenuated when sufficient acoustic variability is contained in the input?
These questions were examined by presenting English speakers with ill-formed clusters (e.g. /bdafa/, /zgade/) containing systematically manipulated sub-phonemic acoustic properties, including duration and amplitude of stop bursts, and the onset and amplitude of voicing before voiced obstruents. These stimuli were presented to participants in two conditions. In the low-variability condition, participants heard the stimuli with the acoustic modifications produced by only one talker. In the high-variability condition, the stimuli were produced by three different talkers within one trial. Results show that in the low variability condition, the acoustic manipulations had strong effects both on the rate of modifications that English speakers produced, and on the type of modifications (e.g. prothesis, epenthesis, C1 change, C1 deletion). For the high-variability condition, the effects of the acoustic manipulations were either eliminated or considerably attenuated.
The results for the low-variability condition demonstrate that under such circumstances, speakers may ‘over-interpret’ non-contrastive acoustic details in determining what structure to produce. These findings are discussed in terms of both the language-specific phonetics and general acoustic characteristics that contribute to the production findings. The attenuation of the sensitivity to phonetic detail in the high-variability condition suggests that listeners generalize over highly variable acoustic information, and that more stable production targets in the high-variability condition result in blending of the information in the multiple talker stimuli.