Listeners cope with speaker and accent variation differently : evidence from the Go/No-go task

The present study tests the hypothesis that speaker and accent normalization are mediated by distinct mechanisms, using a Go/No-go paradigm. Listeners naive to Dutch vowels and to Dutch and Flemish accents were trained to discriminate isolated /ɪ/ and /ɛ/ vowel tokens produced by a female Dutch speaker, and then tested on their categorization of /ɪ/ and /ɛ/ vowels from a different female Dutch speaker, a male Dutch speaker, a female Flemish speaker, and a male Flemish speaker. Our results demonstrate that listeners can correctly categorize vowels in the context of a speaker and gender change, but are unable to do so in the context of an accent or an accent and gender change. This supports our hypothesis that human listeners have separate mechanisms to cope with speaker versus accent variation in vowel productions: a mechanism that is intrinsic for speaker and gender versus a learned mechanism for accent. These results also demonstrate that the XAB task and Go/No-go task produce comparable results, which enables comparisons of Go/No-go responses in humans to non-human listeners.