Spectro-temporal modulation templates for human perception of animal vocalisations in natural soundscapesa)

The human ability to hear animal vocalisations in natural settings was studied by asking naive participants to detect vocalisations in 11 520 samples from a large database of soundscapes varying in terms of habitat (latitude), moment of the day, and precipitation period. Results show that agreement was moderate/good across participants, and the number of vocalisation detections varied systematically as a function of the three ecological factors: Vocalisations were more often detected in the most equatorial habitats, at sunrise and sunset, and for the highest precipitation period. These results are consistent with the latitudinal gradient of biodiversity and diel and seasonal cycles of animal vocal activity. Analysis of modulation power spectra for sounds categorised as containing vocalisations indicates that listeners may base their decisions upon a variety of spectro-temporal modulation cues that are mostly influenced by the openness of the habitat (savannah and desert versus forests). This finding may reflect either adaptation of animal vocal production to the specific sound-propagation characteristics of their native environment, and/or the filtering and distortion effects caused by sound-propagation within natural habitats. This study reveals important ecological constraints on sensory mechanisms involved in detection of vocalisations and provides modulation templates involved in this process for further research.