Vowel devoicing in innovative syllables in Japanese

In Japanese, the short high vowels i and u are often devoiced between voiceless consonants. This study investigates a rare topic in the literature, i.e., whether devoicing pattern differs in syllables found only in loanwords, specifically [ϕi] and [ti], compared to more established ones like [hi] and [pi]. We created made-up mountain name compounds of the form X-san “Mt. X,” where X consisted of four light syllables: [ma], {[ϕi], [hi], [ti], [pi]}, {[sV], [tV]}, and CV (e.g., /matiteno/san “Mt. Matiteno"). 15 items were created for each condition (four target syllables x [s]/[t] contexts), totaling 120 target words. Together with the additional 50 filler items of existing words (involving devoicing and non-devoicing environments), the list was randomized for each participant. 10 native speakers of Japanese (mean age: 23.9 years, 9 females) participated. Analyses using mixed-effects logistic regression models and post hoc pairwise comparisons revealed that the devoicing rates were significantly lower for the innovative syllables than for the established ones: [ϕi] (24.0%) vs. [hi] (38.9%) ( < 0.0001), [ti] (11.8 %) vs. [pi] (26.5%) ( < .0001). In existing words, the effect size appeared even larger ([ϕi] (7.0%) vs. [hi] (47.4%), [ti] (17.1 %) vs. [pi] (48.0%)).