Japanese and Korean listeners’ identification and discrimination of English /r/ and /l/ were compared using a common set of minimal pair stimuli. The effects of speakers (two native speakers of Australian English), position of the contrast within the word (word initial, initial consonant cluster, and medial positions), and listening task (forced choice identification versus oddball discrimination) were examined, with a view to assessing the relative importance of language-specific and language-independent factors operating at the acoustic–phonetic and phonological levels of signal processing in “foreign sound” speech perception. Both prior phonological learning and the relative acoustic discriminability of the items affected subjects’ performance on the identification test. Where both factors were engaged, phonological learning effects predominated over the effects of acoustic discriminability. The extent to which a speaker encoded critical acoustic cues for the /r–l/ distinction was found to affect /r–l/ identification. Dynamic spectral features known to be relevant for the /r–l/ contrast were effective in predicting (in a linear regression analysis) speaker-dependent differences in identification scores. Although the discrimination test may have been influenced by ceiling effects, the performance profiles on the identification and discrimination tests were quite different, indicating that the identification and discrimination tests imposed quite different task demands upon listeners and that phonological processing of the signal was more engaged by the former task.