Investigated the influence of a talker's face (e.g., articulatory gestures) and voice, vocalic context, and word position in the training of Japanese and Korean English as a Second Language to identify American English /r/ and /l/. Findings revealed significant effects of training type, talker, word position, and vowel. Findings are compatible with episodic models for the encoding of speech in memory. (Author/VWL)