Vocal Imitation of Human Speech, Synthetic Sounds and Beluga Sounds, by aBeluga (Delphinapterus leucas)

Abstract

We tested the ability of a beluga (Delphinapterus leucas) to imitate sounds presented to it. During the training session, we presented the subject three recorded sounds that were emitted by the subject, and the subject was trained to imitate them. The subject learned to correctly imitate the sounds. During the test session, two novel computer-generated artificial sounds were presented through an audio speaker. In addition, nine arbitrary vocal sounds produced by the experimenter were presented to the subject, and the subject was required to imitate them. Seven persons, who were not involved in the experiment, were presented the sample sounds and imitated calls; subsequently, they judged whether both sounds were similar to each other. In addition, sound spectrums of the sample sounds and imitated calls were analyzed. As a result, some components of the sound spectrums were similar, and most of imitated calls possess spectral features similar to the sample sounds. These results demonstrated that the beluga was able to correctly imitate novel sounds and spontaneously displayed aptitude for imitation.