Does ultrasound training lead to improved perception of a non-native sound contrast ? : evidence from japanese learners of english *

Both the perception and production of speech sounds in a non-native language can be challenging to adult second language learners due to long-time experience with their native language. Numerous studies have investigated how the ability to perceive and produce speech sounds can be modified during adulthood. This study explores whether production training using ultrasound as visual feedback can lead to improved perception and production of a non-native speech contrast in the absence of perceptual training. To this end, Japanese learners of English who were beginning ESL students in Canada were trained to produce English /r/ and /l/. It has been well documented that native Japanese speakers are likely to have difficulty in discriminating between English /r/ and /l/ (e.g., Goto 1971, Miyawaki et al. 1975) due to the perceived similarity between these phonemes and the Japanese liquid /r/ (e.g., Best and Strange 1992). This claim is often made despite the fact that the Japanese /r/ is phonetically an apico-alveolar tap that is distinct from the English liquids (Vance 2008). The primary acoustic cue that differentiates English /r/ and /l/ is the third formant (F3), which is lower for /r/ and higher for /l/ (e.g., O’Conner et al. 1957). Those phonemes also differ in the second formant (F2), which is slightly lower for /r/ and slightly higher for /l/; however, this does not appear to be a reliable cue in discriminating the phonemes for native English speakers (O’Conner et al. 1957). Compared to native English speakers, native Japanese speakers are less sensitive to the F3 difference (Miyawaki et al. 1975, Best and Strange 1992, Iverson et al. 2003) but are more sensitive to the F2 difference, which may be crucial in identifying the Japanese tap (Iverson et al. 2003). English /r/ and /l/ can also be difficult for native Japanese speakers to produce (e.g., Goto 1971, Sheldon and Strange 1982). This could be due in part to their unfamiliarity with the accurate configurations of the articulatory gestures required for these sounds (Bradlow 2008). Lotto et al. (2004) acoustically analyzed Japanese speakers’ productions of /r/ and /l/ by plotting multiple