Production of schwa by Japanese speakers of English : a crosslinguistic study of coarticulatory strategies

The rhythm of English is charaterized by the alternation of full and reduced vowels. Current studies on the reduced vowel /a/ seem to suggest that schwa may be targetless in F2 (Browman &; Goldstein 1992c; Recasens 1986; Bergem 1993; Bergem 1994). This implies that there may be a contrast of targeted and targetless vowels in languages like English. In other words while the full stressed vowels of English are targeted and resistant to contextual effects (Fowler 1981; Magen 1984), the reduced vowels of English may be targetless and extremely variable as a function of contexts. This pattern of coarticulation seems to be determined at the higher level of the linguistic representation and surfaces as phonetic underspecification of schwa. On the other hand, there is no phonological vowel reduction in Japanese. Accent in Japanese shows little correlation with vowel quality. Presumably, all its vowels are targeted though large extent of Vto-V effects have been reported by a number of studies (Keating &; Huffman 1984; Magen 1984). In other words, different accent types, stress accent and non-stress accent (Beckman 1986) observed in English and Japanese seem to constrain the coarticulatory patterns of the two languages in a global manner. In the present study the coarticulatory pattern of the British English schwa was first observed by using VCoCV sequences with the contextual consonants of /p, t, k/ and the vowels /i, as, u/ embedded in natural sentences. The results of the study suggest that schwa may be targetless in F2. I argue that this contrast in the extent of context dependent vowel variability is an important feature of stress-timing. In the second part, the coarticulatory pattern of the Japanese vowels was studied. In the third part, variability of the English schwa produced by Japanese speak¬ ers of English was investigated. Two groups of Japanese speakers of English pro¬ duced the VCaCV sequences described above. The two groups may be referred to as fluent and non-fluent non-native speakers of English. Similar VC-CV ut¬ terances were recorded in Japanese with the vowel /a/ as the middle vowel as it is the most likely candidate for a transfer from the LI (the first language) vowel system.