Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Distribution of Devoiced High Vowels in Korean

Sun-Ah Jun (1), Mary E. Beckman (2)

(1) Linguistics, University of California, Los Angeles, CA, USA
(2) Linguistics, Ohio State University, Columbus, OH, USA

Korean has a process much like the well-studied "vowel devoicing" of Japanese. Evidence from Japanese suggests that the process is a phonetic undershoot of the vowel's gestures due to overlap with a preceding voiceless consonant. The similar Korean phenomenon would offer further support to this interpretation if vowel devoicing occurs differentially in the context of the three contrasting types of voiceless consonants, which have very different glottal gestures. The phonetic-undershoot account predicts that the vowel devoicing should be most common after plain fricatives or aspirated stops, where the glottal-opening gesture is largest and most likely to overweigh the vowel's voicing gesture. In syllables with lenis-stop initials, the vowel devoicing should occur less often in an accentual phrase medial position. The distribution of devoiced vowels was examined in a controlled corpus of dialogues containing CVCV target words, where the two consonants were voiceless and the first vowel was high. Each target word was produced either in phrase initial or in phrase medial position. The predictions were borne out. In addition, distributional differences involving the following consonant type suggest further testable differences about gestural overlap across syllable boundaries.

Full Paper

Bibliographic reference.  Jun, Sun-Ah / Beckman, Mary E. (1994): "Distribution of devoiced high vowels in Korean", In ICSLP-1994, 479-482.