A harbor seal (Phoca vitulina) shows extensive respiratory control in sound production

Abstract

The duration of animal vocalizations varies between and within species. Which mammals can learn to control this duration? Such respiratory production learning is a scarcely studied subcomponent of vocal learning. Here, we test the hypothesis that harbor seals (Phoca vitulina) are capable of respiratory production learning by testing whether a harbor seal can be trained to i) actively control its vocalization’s duration in two directions (short and long), and ii) exceed the pre-experimental vocalization’s duration (min = 0.202 s, max = 2.621 s). The seal learned to produce uninterrupted vocalizations spanning more than two orders of magnitude in duration, from 79 ms to 9.23 s. Our findings demonstrate a remarkable level of respiratory control in a harbor seal: this respiratory production learning encompasses an extensive range of sound durations and arises at a young age. Producing durations that span such a magnitude is hardly reported in the non-human animal literature; this capacity may be orthogonal to other vocal learning modules and should be tested in more species, both vocal production learners and non-learners.

Supplementary Information: The online version contains supplementary material available at 10.1186/s12862-025-02404-9.

Keywords: Animal training; Bioacoustics; Breathing control; Respiratory production learning; Vocal learning.

Fig. 1

The seal is positioned on its station in front of the microphone and recorder, which records the seal’s vocal output. The experimenter stands in front of the laptop on which our custom-made application runs to assess the recorded response (see Fig. S1) of the seal

Fig. 2

The experimental results show that the seal successfully learned to gradually increase and decrease the duration of its vocalizations. A) Grouped by session threshold, the sessions_long durations (orange) increase between successive thresholds, and vice versa for sessions_short (green). These distributions become increasingly distinct from the pre-experimental vocalizations’ durations (grey). The dashed lines indicate a distribution’s median, the dotted lines the first and third quartiles. Note the logarithmic scaling of the y-axis, spanning over two orders of magnitude. B) At the end of the experiment, there is virtually no overlap between the final session_short and session_long distributions and the pre-experimental one. C) Similarly, the median and inter-quartile range of the responses per session display the diverging vocal durations. D) The learning curve of the sessions_long shows the seal’s respiratory production learning progress over time. E) The learning curve of the sessions_short demonstrates the seemingly higher difficulty in consistently producing shorter vocalizations as the threshold lowers, especially for the final threshold of 0.161 s