F0 as a cue to prosodic structure in voice onset time categorization

Abstract

This study investigates the effect of F0-based cues to English prosodic structure on the categorization of voice onset time (VOT) continua. Human perceptual systems are capable of extracting relevant phonetic and phonological information in spite of pervasive variation in the speech signal, with prosodic structure being one source of variation. For example, in English VOT in voiceless stops is robustly longer when intonational phrase (IP) initial, versus IP-medial. Kim & Cho (2013) found listeners compensate for perceived IP boundaries, requiring longer VOT to categorize sounds as voiceless when IP-initial. However, because the target sound was in an utterance-medial IP, compensation could have arisen due to the lengthened duration of the preceding IP-final syllable, as in speech rate normalization, instead of due to prosodic structure itself. Mitterer et al. (2016) found categorization shifts solely on the basis of duration, by rendering F0 contours prosodically ambiguous, yet it remains unknown what effect F0 has on categorization when it unambiguously cues prosodic structure. This study tests the hypothesis that compensation is triggered by more than duration normalization, by varying F0 cues independently from duration. Results will be discussed in terms of language-general auditory processing and top-down language-specific influences on segment categorization.