Sixth International Conference on Spoken Language Processing
This paper investigated how Japanese learners of English discriminated the American English vowel /I/ from /E/ in /CVt/ monosyllables with 23 different initial consonants, and how the differing discrimination was related to the acoustic characteristics of the English vowels compared with those of Japanese vowels /i/ and /e/ in disyllables with a final syllable of /to/. The results showed that the overall error rate of /I/ was significantly higher than that of /E/. This difference between the two vowels in discrimination could be accounted for by the acoustically closer relations of /I/ to the Japanese vowel /e/. It was found, however, that the discrimination of the two English vowels varied across the consonantal contexts. The error rate was significantly higher when the initial consonant was /tS/, /D/ or /j/ for /I/, and when it was /r/ for /E/. An acoustical analysis indicated that the greater decrease of the ratio of the F2 frequency to the F1 frequency of /I/ through the vowel, rather than the acoustical closeness of /I/ to the Japanese /e/, tended to be more closely related with the poorer discrimination of this vowel in the particular consonantal contexts.
Bibliographic reference. Joto, Akiyo (2000): "The effect of consonantal context and acoustic characteristics on the discrimination between the English vowel /i/ and /e/ by Japanese learners", In ICSLP-2000, vol.3, 957-960.