The grain size of auditory mismatch response in speech perception

Abstract

This phonetic study examined neural encoding of within-and cross- category information as a function of language experience. Behavioral and magnetoencephalography (MEG) measures for synthetic /ba–wa/ and /ra–la/ stimuli were obtained from ten American and ten Japanese subjects. The MEG experiments employed the oddball paradigm in two conditions. One condition used single exemplars to represent the phonetic categories, and the other introduced within-category variations for both the standard and deviant stimuli. Behavioral results showed three major findings: (a) a robust phonetic boundary effect was observed only in the native listeners; (b) all listeners were able to detect within-category differences on an acoustic basis; and (c) both within- and cross- category discriminations were strongly influenced by language experience. Consistent with behavioral findings, American listeners had larger mismatch field (MMF) responses for /ra–la/ in both conditions but not for /ba–wa/ in either. Moreover, American listeners showed a significant MMF reduction in encoding within-category variations for /ba–wa/ but not for /ra–la/, and Japanese listeners had MMF reductions for both. These results strongly suggest that the grain size of auditory mismatch response is determined not only by experience-dependent phonetic knowledge, but also by the specific characteristics of speech stimuli. [Work supported by NIH.]