Native and non-native perception of phonemic length contrasts in Japanese

Abstract

Segment length is distinctive in Japanese, e.g., /kado/ ‘‘corner’’ versus /ka:do/ ‘‘card,’’ unlike languages such as English. To study this property extensively and precisely, a series of perception experiments were conducted. Stimuli were pairs of unaccented 3- or 4-mora nonsense Japanese words of the basic form /ere——e/ where —— is one of /p t k m n s/. Each pair contrasted in either vowel length, e.g., /erete/ versus /ere:te/, or in consonant length, e.g., /erete/ versus /erette/. Each pair was also embedded within a carrier sentence. A synthetic continuum gradually varying in the duration of the target vowel or consonant was built for each pair using straight, a high-fidelity speech analysis, synthesis, and manipulation system [Kawahara et al., Speech Commun. 27, 187–207 (1999)]. Native Japanese and native English speakers participated in both identification and discrimination tests of each continuum. Preliminary results from stimuli containing stops indicate that the non-native discrimination scores were significantly poorer than the native scores, particularly when the continua were embedded in a sentence. The non-native identification scores, however, were similar to those of the native. Implications of these findings are discussed along with results from the fricative and nasal continua.