Relationship between categorical perception of /r/ and /l/ and identification of liquid-glides and voiced plosives by Japanese and American listeners

The purpose of this study was to investigate the relationship between categorical perception of American English /r/ and /I/ and identification of American English liquid-glides and voiced plosives by Japanese and American listeners in quiet and in two noise conditions (S/N = +8 dB and S/N = +4 dB). Identification of liquid-glides by Japanese subjects showed that the higher the ability of categorical perception of /r/ and /I/, the higher the identification scores among Japanese subjects, whereas identification of the voiced plosives by the Japanese subjects was not related with the categorical ability. The identification scores by the American subjects did not show relationship with the categorical ability for either liquid-glides or voiced plosives. Among the Japanese subjects, reduction of identification scores of liquid-glides due to noise was smaller among the subjects with low categorical ability than those with high categorical ability. A possible reason was that the identification scores of"the liquid-glides was already low (near chance level) in quiet among the subjects with low categorical ability. This study also showed that the scores for the optimal synthetic /r/ and /I/ stimuli were higher than the natural speech stimuli among the Japanese subjects, contrary to results from most of the previous studies. This may be due to the peculiarity of the talker, and also due to redundant cues distracting the listeners. Implications of these results were discussed in terms of separate speech processing strategies which might have been used by the Japanese subjects. Further research was suggested for investigating identification scores in noise using synthetic stimuli using only subjects who scored high (around 90%) in quiet, so that difference scores can be compared more meaningfully.