There are three obvious diachronic levels in the pronunciations of the ancient voiced initial consonants in Min dialects nowadays: the unvoiced aspirated level, the unvoiced unaspirated level and the voiced level.

Having the phonetic experience of aspiration in the allophones of voiceless stops in their native language, the learners could identify the aspirated SC stops very well, but their identification on unaspirated stops is poor.

It is found that adaptive perceptual training improved categorical perception of [pa] and [pha],and the effect transferred to the perception of [ta] and [tha],which indicates that perceptual training improves categorical perception of aspirated/unaspirated contrast as a whole.