Overlap at the interface: High vowel devoicing in Japanese

Abstract

High vowel devoicing (HVD) in Japanese has been characterized as both phonological [J. D. McCawley, The Phonological Component of a Grammar of Japanese (1968)], phonetic due to gestural overlap [S.-A. Jun and M. E. Beckman, LSA Paper (1993)], and as a result of overall vowel reduction [M. Kondo, Ph.D. thesis, University of Edinburgh (1997)]. Electromyographical and laryngeal spread data from one speaker [A. Tsuchida, Ph.D. dissertation, Cornell University (1997)] supports the existence of both phonological and phonetic devoicing in complementary environments, while data from ten speakers [J. K. Varden, Ph.D. dissertation, University of Washington (1998)] indicates the presence of both types of devoicing in overlapping environments. Results of Varden (1998) also indicate high-speaker variability in rate of devoicing, duration of voicing for voiced vowels, and context-dependency. A synthesis of these studies supports a characterization of HVD as the result of varying processes, each having the same acoustic end: the devoicing and overall reduction of high vowels in devoicing environments. The data in Varden (1998) also indicate a high incidence of devoicing of /u/ in the mora [tsu] before /n/ for some speakers. This indicates relaxation of the requirement of a following underlyingly voiceless segment.