Articulatory and acoustic characteristics of devoiced high vowels in Tokyo Japanese

High vowels /i/ and /u/ are often devoiced between two voiceless obstruents in Tokyo Japanese, raising the question whether their contrastive lingual articulation is preserved in this context. Here, we used ultrasound imaging and acoustic formants to compare voiced and devoiced vowels. Eleven native speakers of Tokyo Japanese produced eight potentially devoiceable and non-devoiceable nonce word pairs. High vowels were preceded by /k/ or /p/ and followed by /t/, creating a “devoiceable” environment, or /d/, a “non-devoiceable” one. Tongue contours and formants were measured at two-thirds of an interval delimited by the C1 burst and onset of C2 closure. Articulatory results from two speakers showed that tongue position differed between voiced and devoiced vowels of the same target type, but that the distinction between vowel targets was maintained regardless of voicing. F2 results showed similar patterns: both /i/ and especially /u/ contexts differed slightly by vowel voicing, but retained the /i/-/u/ contrast when devoiced. These acoustic results did not differ by the place of C1, indicating that similar cavity shapes were excited. Overall, results provide both articulatory and acoustic evidence that devoiced vowels retain contrastive lingual articulation. [Work supported by NIH.]