The role of learned song in the evolution and speciation of Eastern and Spotted towhees

TL;DR

This paper explores how song differences in Eastern and Spotted towhees may contribute to their speciation and reproductive isolation.

Contribution

It uses machine learning to show that combinations of song features can distinguish these recently diverged species with high accuracy.

Findings

No single song feature reliably distinguishes Eastern and Spotted towhees.

Machine learning models classified songs with high accuracy (88-90%).

Song classification accuracy was lower in the zone of overlap between species.

Abstract

Oscine songbirds learn vocalizations that function in mate attraction and territory defense; sexual selection pressures on these learned songs could thus accelerate speciation. The Eastern and Spotted towhees are recently diverged sister species that now have partially overlapping ranges with evidence of some hybridization. Widespread community-science recordings of these species, including songs within their zone of overlap and from potential hybrids, enable us to investigate whether song differentiation might facilitate their reproductive isolation. Here, we quantify 16 song features to analyze geographic variation in Spotted and Eastern towhee songs and assess species-level differences. We then use several machine learning models to measure how accurately their songs can be classified by species. While no single song feature reliably distinguishes the two species, machine learning models classified songs with relatively high accuracy (random forest: 89.5%, deep learning: 90%, gradient boosting machine: 88%, convolutional neural network: 88%); interestingly, species classification was less accurate in their zone of overlap. Finally, our analysis of the limited publicly available genetic data from each species supports the hypothesis that the species are reproductively isolated. Together, our results suggest that small variations in multiple features may contribute to these sister species’ ability to recognize their species-specific songs.

Songbirds learn their songs through imitation; these songs are important in mate selection and thus could affect evolution. Multiple factors, including genetics, cultural transmission, and environmental conditions, could influence the song differences between individuals, and some song variations might be more attractive than others. These variations in song can accumulate over time, leading to the emergence of distinct dialects and regional variations, which can potentially contribute to reproductive isolation and speciation. Thus, in order to shed light onto the incredible diversity of songbird species, we aim to uncover evolutionary patterns in birdsong to improve our understanding of its role in speciation. To this end, we used statistical analyses to investigate the patterns of song variation that exists in the Spotted and Eastern towhees, a sister-species pair whose ranges overlap, and we use various machine learning algorithms trained on 16 song features to assess the degree of song distinguishability between the two species. Our results show that no single song feature reliably distinguishes the two species; however, combinations of these features could classify songs with high accuracy. These findings suggest that song could play some role in the ability of these birds to recognize their own species’ songs.

Figures