Distributional properties of voice onset time in child speech and adult listener judgments

Abstract

Voice onset time (VOT) is well-recognized as the primary cue for perception of plosive voicing distinctions although perceptual judgments can also be affected by secondary cues, such as fundamental frequency (f0) and burst amplitude. Further, perceptual judgments may be influenced by distributional properties of the dataset, i.e., how VOT values are distributed along the continuum, and how many productions fall into each category in a listening test. This study assessed the impact of distributional characteristics of naturally produced child speech stimuli to determine their influence on listeners’ labeling behavior. The dataset comprised bilabial and alveolar CV words produced by 2–3-year-old English-speaking children. Six exemplars were chosen from six children for four groups: short-lag /b d/ and /p t/, long-lag /b d/ and /p t/. Within-category VOT distributions per POA were bimodal, separated by a 5 ms gap. Listening data were obtained from 20 adults. Preliminary data are consistent with our past experiments: Listeners were highly accurate (i > 90%) at labeling the production as intended by the child when the production matched the VOT category expected for the target word. Measures of f0 and burst amplitude suggest that secondary cues in the speech signal contributed to adults’ perception of children's stops.