The present study explored whether listening to songs and singing can improve second language pronunciation and vocabulary learning at beginning stages of language acquisition. One hundred and eight Chinese students underwent a 4-min training session to learn 14 words from a meaningful French song about the parts of the body in either one of two conditions: in Experiment 1, listening to rhythmic speech vs listening to the same words but in a song (50 participants), and in Experiment 2, listening to vs singing a song (58 participants). Accentedness ratings of pretest and posttest recordings revealed that (a) the song listening group reduced accentedness significantly more than the rhythmic speech listening group (Experiment 1); and (b) singing and listening to a song yielded similar significant improvements after training (Experiment 2). No advantage of song listening compared to rhythmic speech listening nor of singing compared to song listening was found for word recall. Individual measures of working memory and imitation ability significantly influenced our results.