Cues in auditory discrimination and identification of time of day for natural soundscapes

Background Previous work suggests that human listeners can detect structure in natural soundscapes, and that spectral information is involved in the discrimination of these ecological scenes (e.g., Apoux et al., 2023). The present study extended this work to the perception of time of day, and assessed which acoustic cues (fine spectral detail, temporal envelope, temporal fine structure) mediate performance. Methods Normal-hearing adults were asked to perform time of day identification and discrimination for four broadly defined moments: Dawn, Midday, Dusk, Night. Identification used a single-interval 4AFC task; discrimination used a three-interval oddity task. Soundscapes were scenes recorded in three natural habitats (two temperate forests, one tropical) with minimal anthropogenic noise. To probe cue use, scenes were processed with a noise vocoder to alter temporal fine structure (TFS) cues and selectively degrade spectral or temporal information by manipulating the number of bands and the cutoff of the envelope filter. Results Participants showed poor identification of time of day, even when provided with feedback and after several hours of familiarization. In contrast, soundscape discrimination based solely on the time of recording was always very reliable. In fact, all the manipulations performed to degrade the signals generally resulted in limited effects. Overall, discrimination was especially robust to manipulations of envelope and TFS. Selective degradations of spectral information reduced performance in some extreme conditions, suggesting that decisions relied on gross spectral cues (e.g., ambient-noise). The impact of spectral degradation varied by habitat, indicating that the effective spectral resolution used by listeners depends on scene statistics specific to each forest type. Conclusions Human listeners struggled to identify time of day based on 1-s samples of natural soundscapes. However, they could discriminate these short samples very reliably. This was found to be true for soundscapes recorded from three different habitats. The soundscape recorded closer to the equator led to the best discrimination performance, perhaps partly because of greater biodiversity. The identification-discrimination dichotomy suggests that while human listeners are sensitive to acoustic variations related to differences in time of day, they cannot consistently label these variations. Despite inconclusive data, it is suggested that biophony – the collective sound produced by animal vocalizations – did not play a critical role in time of day discrimination, probably because this cue was often inconsistent in the short duration of the samples used here. Consistent with previous work, the present results suggest that human listeners can use spectral and temporal cues, but that they base their decisions primarily on spectral cues, the resolution of which depends on the habitat. However, none of these cues were essential for performance of the task. Future work should test individuals with hearing loss and evaluate device algorithms that maintain spectral integrity in complex, naturalistic scenes.