More acoustic traces of “deleted” vowels in Japanese.
| Number | 1196 |
|---|---|
| Year | 2000 |
| Drawer | 22 |
| Entry Date | 03/15/2001 |
| Authors | Faber, A. & Vance, T.J. |
| Contact | |
| Publication | In Nakayama, M. and Quinn, C.J. Jnr. (eds.) Japanese/Korean Linguistics, v.9, (pp.100-113). California, CLSI Publications, 2000. |
| url | http://www.haskins.yale.edu/Reprints/HL1196.pdf |
| Abstract | [Introduction] A number of well-known languages exhibit vowel devoicing, including Korean, but the phenomenon has been studied most intensively in Japanese (Jun, Beckman, Niimi, & Tiede, 1998:1). The standard allophonic account for Japanese says that a short high vowel, /i/ or /¨/, is devoiced when it is flanked by voiceless consonants or, word-finally, when it is preceded by a voiceless consonant (Nihon, Onsei Gakkai, 1976: 748; Alfonso, 1971: xxviii), but this is just a crude first approximation. For one thing, non-high vowels occasionally exhibit devoicing as well, although not nearly as often as high vowels (Han, 1962: 84-85; Vance, 1987: 48-49). Several additional factors may also affect the probability of devoicing in high vowels. For example, it has long been claimed that devoicing is inhibited in an accented syllable (Han, 1962: 81-82; Kuriyagawa & Sawashima, 1989; Imai, 1997: 48), although some studies have challenged this conventional wisdom (Hsraguchi, 1984; Tsuchida, 1997; Kitahara, 1998), and it may be that there has been a diachronic weakening of the inhibiting power of accent in recent decades (Nihon Onsei Gakkai, 1976: 748; Kondo, 1997: 83-84). It has also been claimed that certain neighboring consonants favor devoicing more than others (Martin, 1952: 14; Han, 1962: 88-90), although the evidence for such differences is not very clear (Imai, 1997: 53). A third common claim is that devoicing in consecutive syllables is avoided (Martin, 1952: 14; Han, 1962: 91), but it can and does occur (Haraguchi, 1984; Imai, 1997: 50). When each of two or more consecutive syllables contains a short high vowel in the devoicing environment, the probability of devoicing in a particular syllable seems to depend on both accent and morphological structure (Vance, 1992; Kondo, 1997: 100-147). There is no controversy about the relevance of tempo and style to the probability of devoicing in Japanese. Devoicing is less frequent in slower and more careful speech (Sakuma, 1929; 232-233; Kuriyagawa & Sawashima, 1989; Jun, Beckman, & Lee, 1998:43). For example, professional teachers of hearing-impaired children devoice less in speech directed to such children than in speech directed to normal-hearing listeners (Imaizumi, Hayashi, & Deguchi, 1995). There is also considerable variability in the phonetic realizations of devoiced vowels. Spectrograms sometimes show energy at the appropriate formant frequencies even in the absence of any voicing (Han, 1962: 83; and Jun, Beckman & Lee, 1998:50) sort tokens into three categories: voiced, partially devoiced, and completely devoiced. They define partial devoicing as “shorter than 30ms and show[ing] weak energy, at low frequencies only.” Generally speaking, if the consonant preceding a devoiced vowel is a stop, and acoustic interval corresponding to the devoiced vowel is easily identifiable on waveforms and spectrograms, but if the preceding consonant is a fricative, such an interval is typically absent. |
| Notes |
Search Publications