Machine learning reveals cryptic dialects that explain mate choice in a songbird

• DOI: 10.1038/s41467-022-28881-w
• Publication Date: 2022-03-28T10:22:53Z
• Authors: Daiping Wang; Wolfgang Forstmeier; Damien R. Farine; Adriana A. Maldon...; Katrin Martin; Yifan Pei; Gustavo Alarcón-N...; James A. Klarevas...; Shouwen Ma; Lucy M. Aplin; Bart Kempenaers
• Keywords: info:eu-repo/classification/ddc/570; Male; 0106 biological sciences; Science; Q; General Physics and Astronomy; Genetics and Molecular Biology; General Chemistry; 01 natural sciences; Article; Machine Learning; 10127 Institute of Evolutionary Biology and Environmental Studies; General Biochemistry; 570 Life sciences; biology; 590 Animals (Zoology); Animals; Female; Finches; Vocalization, Animal; Language

AbstractCulturally transmitted communication signals – such as human language or bird song – can change over time through cultural drift, and the resulting dialects may consequently enhance the separation of populations. However, the emergence of song dialects has been considered unlikely when songs are highly individual-specific, as in the zebra finch (Taeniopygia guttata). Here we show that machine learning can nevertheless distinguish the songs from multiple captive zebra finch populations with remarkable precision, and that ‘cryptic song dialects’ predict strong assortative mating in this species. We examine mating patterns across three consecutive generations using captive populations that have evolved in isolation for about 100 generations. We cross-fostered eggs within and between these populations and used an automated barcode tracking system to quantify social interactions. We find that females preferentially pair with males whose song resembles that of the females’ adolescent peers. Our study shows evidence that in zebra finches, a model species for song learning, individuals are sensitive to differences in song that have hitherto remained unnoticed by researchers.