The way the monolingual Persian and early bilingual Azerbaijani/Persian learners of English as a foreign language conceive of the vowels of American English in terms of vowel quality and vowel duration is very much the same, but differs markedly from the perceptual representation of the American English control group. Moreover, the perceptual representation entertained by the three groups of listeners mirrors the way the participants pronounce the vowels (as examined in Experiment 1). The discrepancy between the L1 and L2 speakers is most clearly seen in the high vowel pairs /i- ɪ/ and /u-ʊ/. For the Iranian EFL learners, the short/lax and long/tense member of each pair have virtually the same centroids and almost complete overlap in their dispersion ellipses. The native listeners have different targets for /i/ vs. /ɪ/, as well as for /u/ vs. /ʊ/, so that the centroids for the vowels are more than a full Bark away from each other and with no overlap between the associated (25%) dispersion ellipses. Similarly, /o/ is not distinct from /u, ʊ/ in the L2 perceptual representation of the vowel quality, while the /o/ centroid is more than 1 Bark away from both /u/ and /ʊ/ with no overlap in the L1 listeners. EFL /ʌ/ clusters with the low-back vowels /ɑ, ɔ/, whereas /ʌ/ is clearly separated in the L1 perceptual representation.