Abstract-The aim of this study was to investigate, how young children learn to produce a non-native vowel embedded in a pseudo-word context after short and intensive listen-and-repeat training sessions. The trained vowel contrast was chosen so that it would generate maximal learning difficulties according to models of second language learning. The group consisted of 13 7-10 year old girls. The child subjects participated in the study twice, on two consecutive days. Both days consisted of two training and two recording sessions. The model stimuli were pseudo-words /ty:ti/ and /tʉ :ti/, synthesized using a semi-synthetic method. The primary target word was /tʉ :ti/, which contained the vowel phoneme phonologically irrelevant in Finnish. The two lowest formants and the fundamental frequency were analysed from the recorded vowels. The results revealed that children learned to produce the new vowel after only three training sessions and that the effect was also seen after the fourth session. These results suggest that children learn to produce a new vowel after a short period of phonetic training.