Perception of English /i/ and /I/ by Japanese and Spanish Listeners: Longitudinal Results

Flege’s Speech Learning Model predicts that if an L2 learner perceives an L2 speech sound as similar to an L1 speech sound, the two sounds will be combined as a diaphone category, the properties of which will eventually be intermediate between the properties of the L1 and L2 sound. In contrast if the L2 sound is perceived as new, then a new category will be established with properties which may eventually match the properties of the L2 sound. Canadian English has two high front vowels: tense /i/ and lax /I/ differing in spectral and duration properties. Japanese has two high front vowels: long /i:/ and short /i/ differing in duration only. English /i/ and /I/ are expected to be perceived as similar to Japanese /i:/ and /i/, and Japanese learners of English are predicted to diaphone categories. Their identification of English /i/ and /I/ is predicted to initially match their perception of Japanese /i:/ and /i/, but eventually be intermediate between the native norms for the L1 and L2 categories. Spanish has one high front. Spanish learners of English are predicted to perceive English /I/ as less similar Spanish /i/ than English /i/, and are predicted to eventually establish a new /i:/. Their identification of English /i/ and /I/ is predicted to initially be poor but match that of English listeners. These predictions were tested using a edited-speech continuum covering the English words /bIt bit bId bid/. which varied in the continuum included vowel spectral properties and vowel. A longitudinal study was conducted testing Japanese and Spanish speaking learners of English one month and six months after their arrival in Canada. Japanese listeners were found to have a primarily duration-based categorical boundary between English /i/ and /I/ which did not change between the initial and final tests. Spanish listeners did not have a categorical identification pattern in the initial test, but they did establish duration-based or spectrally-based categorical boundaries by the time of the test. Results were therefore consistent with the theoretical predictions.