Vowel devoicing in Tokyo Japanese: its lexical status

In Tokyo Japanese, high vowels can be devoiced typically between voiceless elements. In the classic view, devoiced vowels are not phonemes but allophones of corresponding vowels. Consequently, the devoicing process belongs to the phonetic domain. However, many scholars have claimed that devoicing actually belongs to the phonological process and some of them even proposed that devoicing may partly be specified in the lexicon. To investigate the lexical status of vowel devoicing, we conducted a lexical decision task of natural and edited mismatch stimuli where high vowels were devoiced before voiced consonant or voiced before voiceless consonant in real words and nonce words. The stimuli were presented with white noise to avoid the ceiling effect and RT was measured. The results suggest that the mismatch condition affected RT selectively according to real/nonce status of words and voicing. Though the idea of lexically specified devoicing is not directly supported, we can give a new insight to the issue of phonetics-phonology-lexicon interface.

Co-authored with Natsuya Yoshida and Ayako Shirose, presented at International Congress of Phonetic Sciences, Melbourne, Australia, Aug. 5-9, 2019, supported by JSPS KAKENHI Grant Number 18K00601.