Vocal repertoire and information encoding in African Penguin calls: A decade of research

The African Penguin (Spheniscus demersus) is a colonial and critically endangered seabird. Over the past decade, we have identified four distinct vocal types in the vocal repertoire of adult African Penguins mediating a variety of social interactions. Within the framework of source–filter theory, we combined bioacoustics recordings with anatomical dissections, computed tomography, and computational modeling to disentangle the roles of the vocal source (the syrinx, determining fundamental frequency) and the vocal filter (the vocal tract, shaping formants) in producing different vocalizations and encoding species-specific, size-related, and individual identity cues across these call types. Furthermore, we investigated the high-frequency food solicitation signals (begging calls) that African Penguin chicks use to request food from parents returning from foraging trips. When characterizing these calls, we demonstrated that nonlinear vocal phenomena (NLP), including deterministic chaos, subharmonics, and sidebands, are often present but increase with respiratory illness, suggesting both communicative function and diagnostic value. Expanding upon this body of knowledge, our recent work has explored the use of Convolutional Neural Networks to detect and classify African Penguin vocalizations from soundscape recordings and to identify NLP in begging calls, highlighting the potential of this approach to support behavioral research, health monitoring, and population censusing of extant colonies.