Identification of English /r/ and /l/ in white noise by native and non‐native listeners

Fifty‐three English word pairs, minimally contrasting in /r/–/l/, were identified by 14 native American English (AE) listeners and 42 native Japanese (J) listeners, under systematically varied signal‐to‐noise ratios (SNR) which ranged from no‐noise to 21 dB. Word identification for filler contrasts (FL: /d/–/k/, /m/–/n/, etc.) was also tested under the same SNR conditions. Results show that: (a) identification accuracy by the AE listeners was nearly 100% in the no‐noise condition compared to about 70% with a SNR of 21 dB; (b) the J listeners showed similar identification accuracy to the AE listeners for the FL words, whereas they performed much poorer on the /r/–/l/ contrasted words (70% accuracy without noise and 55% accuracy with a SNR of 21 dB); and (c) /r/–/l/ identification accuracy was affected by an interaction between the consonant position in a word and the listeners’ native languages. The role of perceptual assimilation in the interaction will be discussed. [Work supported by TAO and JSPS, Japan.]