Detecting cultural evolution in a songbird species using community-science data and computational modeling

Abstract

Song in oscine birds is learned across generations, and aspects of the song-learning process parallel genetic transmission: variation can be introduced into both cultural and genetic traits via copy-error, and both types of traits are subject to drift and selective pressure. Similarly to allele frequencies in population genetics, observing frequencies of birdsong features can improve our understanding of cultural transmission and evolution. Uniquely, community-science databases of birdsong provide rich spatiotemporal data with untapped potential to evaluate cultural evolution in songbirds. Here we use both community-science and field-study recordings of chipping sparrows to examine trends across nearly seven decades of song. We find that some syllable types tend to persist in the population for much longer than others. Persistent songs tend to contain more syllables of shorter duration than songs which were observed across fewer years. To draw inferences about the effects of learning biases on chipping sparrow syllables, we construct a spatially explicit agent-based model of song learning. By comparing our empirical analysis to simulated song distributions using three different song-learning strategies--neutral transmission, conformity bias, and directional selection--we suggest that chipping sparrows are unlikely to select tutors neutrally or with a conformity bias and that they learn their songs with a remarkably low copy-error rate.