Nonphonetic variability and speech categorization by native and non-native learners

Abstract

Natural speech stimuli used in studies of phonological learning typically include several in talkers and phonetic environments because nonphonemic variability usually facilitates learning [e.g., Lively et al., J. Acoust. Soc. Am. (1993)]. In contrast, our previous work demonstrated that perceived talker variability in a synthetic test set markedly decreased the ability of monolingual American English subjects to distinguish between the voiced dental stop consonant and the voiced alveolar stop [d] [Jantzen and Tuller, J. Acoust. Soc. Am. (2006)]. Only the latter is phonemic in American English. There are several possible explanations for this that are currently being evaluated. Here we examine the effect of perceived talker variability on Malayalam listeners’ identification of three dental to alveolar continua that differ in F0. In Malayalam, both the dental and the alveolar stop consonants are phonemic. The results shed light on whether talker information and phonetic information share a representational structure. [Work supported by NSF and ONR.]