Understanding sources of variance and correlation among features of Bornean gibbon (Hylobates muelleri) female calls.

Acoustic signals serve important functions in mate choice, resource defense, and species recognition. Quantifying patterns and sources of variation in acoustic signals can advance understanding of the evolutionary processes that shape behavioral diversity more broadly. Animal vocalization datasets are inherently multivariate and hierarchical, wherein multiple features are estimated from calls of many individuals across different recording locations. Patterns of variation within different hierarchical levels-notwithstanding the challenges they present for modeling and inference-can provide insight into processes shaping vocal variation. The current work presents a multivariate, variance components model to investigate three levels of variance (within-female, between-female, and between-site) in Bornean gibbon calls. For six of the eight features estimated from call spectrograms, between-female variance was the most important contributor to total variance. For one feature, trill rate, there were site-level differences, which may be related to geographic isolation of certain gibbon populations. There was also a negative relationship between trill rate and duration of the introduction, suggesting trade-offs in the production of gibbon calls. Given substantial inter-individual variation in gibbon calls, it seems likely that there has been selection to confer information regarding caller identity, but mechanisms leading to site-level variation in trill rate remain to be determined.