We present the results of an acoustic investigation of English vowels as produced by Serbian speakers, students in the English Department, Faculty of Philology, University of Belgrade. The number of participants was 26 (13 first-year students, and 13 fourth-year/ MA students), and measured were F1 and F2 of stressed vowels — 11 monophthongs (KIT, DRESS, TRAP, FOOT, STRUT, LOT, FLEECE, PALM, GOOSE, THOUGHT and NURSE) and 4 diphthongs (GOAT, PRICE, MOUTH and FACE). Measurements were also made of the participants’ L1 vowels, with which their L2 vowels were compared. Participants were recorded reading BBC news copy in English, and two very short stories in Serbian. The number of tokens analyzed was 7534 for English (around 305 per speaker), and 4266 for Serbian. The results show that Serbian-speaking learners, at the proficiency level of our informants, do not on the whole substitute their L2 vowels with L1 vowels; on the other hand, they also do not quite reach the qualities characteristic of native speakers, but rather reach compromise values. Exceptions, regarding substitution, are DRESS, and for some informants at least, TRAP (both are substituted by the Serbian short /e/). The vowels that exhibited the largest intergroup differences were GOAT, GOOSE, MOUTH, PRICE, and to a lesser degree THOUGHT, with older students showing more native-like qualities.
