Author:
Provost Kaiya L.,Yang Jiaying,Carstens Bryan C.
Abstract
AbstractOne of the most obvious traits of birds is the astonishing array of songs that are produced by most species. The diversity and functions of bird vocalizations have been studied across many facets of biology, but variation in song is often assumed to be determined by sexual or social selection, rather than natural selection acting in the context of the landscape. Here we use deep machine learning to investigate multiple co-distributed species in the New World Sparrows. We leverage existing bioacoustic and genetic repositories and environmental data to identify the processes that structure variation in bird song within a community, and to determine how this variation corresponds to genetic and phenotypic patterns. Automated processing of these data at the family level has allowed us to synthesize at deep phylogenetic scales. Analysis of these data using a multiple matrix regression with randomization indicate that song variation in ∼40% of species can be explained by environmental and stochastic factors. Genetic variation is explained by similar factors in ∼30% of the species. This suggests that across a community and global scale, the action of natural selection on the evolution of song is at least as impactful as it is on other genetically-determined traits.
Publisher
Cold Spring Harbor Laboratory