Vocal imitation of song and speech.

We report four experiments that explored the cognitive bases of vocal imitation. Specifically, we investigated the accuracy with which normal individuals vocally imitated the pitch-time trajectories of spoken sentences and sung melodies, presented in their original form and with phonetic information removed. Overall, participants imitated melodies more accurately than sentences with respect to absolute pitch but not with respect to relative pitch or timing (overall duration). Notably, the presence of phonetic information facilitated imitation of both melodies and speech. Analyses of individual differences across studies suggested that the accuracy of imitating song predicts accuracy of imitating speech. Overall, these results do not accord with accounts of modular pitch processing that emphasize information encapsulation.