Vowel identification and production performance for an experimental group of 35 Japanese subjects was measured before and after a 6-week period within which subjects received identification training with feedback. A control group of 9 Japanese subjects was administered the same identification and pre- and post-test, separated by the same 6-week interval during which they did not receive the experimental vowel identification training that provided feedback regarding which vowel sound had been produced by the native American English talkers. The results showed that the experimental groups identification performance (as measured by d) for each of the five American English (AE) vowels /æ/, /A/, /2/, /O/, /Ç/ improved more than the control groups did. Both before and after the 6-week identification- training period, recordings were made of the experimental groups productions of a list of words in a varied [CVC] context, each containing one of the five target AE vowels, which were evaluated by 3 AE native listeners using a 2-interval, forced-choice identification task. The results showed the AE listeners preferred the post-test to pretest productions for four out of the five target AE vowels. Overall, the results indicate the feedback-based identification training had a positive effect on both the experimental groups identification and production performance.