In a precursor study (Afshar, Van Heuven, 2022), we examined how the eleven American English (AE) monophthongs were assimilated to the six vowels of Persian or nine vowels of Azerbaijani by 22 monolingual Persian and by 27 (early) bilingual Azerbaijani/Persian EFL learners. The monolinguals and bilinguals assimilated the AE vowels to the six vowels of Persian in virtually the same way. When responding in Azerbaijani mode, however, the bilinguals selected their native vowel /y/ as the modal (i.e., most frequent) response for AE /u/ (showing that they perceived the fronting of AE /u/) and their own /œ/ for AE /o/ and /ʊ/, again indicating that the bilinguals were sensitive to the centralized nature of these AE vowels. The AE mid-low central vowel /ʌ/, however, could not be matched with any vowel in either Persian or Azerbaijani, giving it the status ‘unclassified’.

In our first experiment, the same participants produced the eleven monophthongs of American English in everyday keywords and in /hVd/ words. The acoustic correlates of jaw aperture (also called vowel openness or vowel height) and constriction place (along the front-back dimension), i.e., F1 and F2 respectively, and vowel duration were measured and statistically analyzed. The results of the acoustic analysis of the EFL learners’ AE vowel production (/hVd/ words only) could be compared in detail with similar data collected for native speakers of AE English, made available to us. The monolingual and bilingual Iranian EFL learners pronounce the AE vowels in the same manner. Both learner groups depart grossly from the L1 norms for AE vowels. All the EFL vowels have much shorter durations, while at the same time the EFL vowel space is much smaller than the native space. Spectral differences between members of tense/lax vowel pairs are small (if they can be found at all), while relative vowel durations mirror those of the L1 speakers, and differentiate between spectrally adjacent lax and tense vowels.

In experiment 2, the same monolingual and bilingual EFL learners listened to 86 artificial vowel sounds in /mVf/ nonsense items differing in degree of jaw aperture (7 perceptually equal steps of one bark unit along the F1 resonance dimension) and constriction place (9 equal steps of 1 Bark along the F2 dimension). A set of 43 vowels were generated in a short (200 ms) and a long (300 ms) version. Listeners had to decide for each synthetic vowel token, which one of the 11 AE monophthongs was most similar to it. This procedure allows us to map out the perceptual representation of the AE vowel system in the mind of the listener. Specifically, the results reveal which location in the vowel space the listener considers as the most typical realization of the vowel (the vowel centroid) and how this location may shift under the influence of vowel duration. The results obtained were compared with similar data collected from native listeners of American English. Our data indicate that there are hardly any differences between the monolingual and bilingual EFL learners in the way they identify the artificial vowels as tokens of the eleven AE vowel types. However, for both groups of learners alike, their perceptual representation differs markedly from what is found for the native control listeners, the most conspicuous difference being that the native listeners hardly rely on duration to differentiate between the tense and lax subsystems in the AE vowel inventory, while the learners ignore small differences in vowel quality and overestimate the importance of vowel duration.

The perceptual representation of the AE vowel system is highly isomorphic with the acoustic realization found in experiment 1. The isomorphism between acoustic vowel realization and perceptual representation is observed in like manner for American L1 participants and for both groups of Iranian EFL participants alike.