Temporal stability in songs across the breeding range of <i>Geothlypis philadelphia</i> (Mourning Warbler) may be due to learning fidelity and transmission biases

ABSTRACT We found a stable pattern of geographic variation in songs across the breeding range of the Geothlypis philadelphia (Mourning Warbler) over a 36-year period. The Western, Eastern, Nova Scotia, and Newfoundland regiolects found in 2005 to 2009 also existed in 1983 to 1988 and 2017 to 2019. Each regiolect contained a pool of syllables that were unique and different from the other regiolects. The primary syllable types that defined each regiolect were present throughout the study, but there were changes in the frequencies of variants of these syllable types in each regiolect. We developed an agent-based model of birdsong learning within each regiolect to explore whether these frequency changes were consistent with unbiased copying or 2 forms of transmission bias: frequency bias and content bias. Strong content bias, possibly for more complex syllables, best models the temporal dynamics across regiolects. In combination with a high estimated learning fidelity, this may explain why regiolects and syllable types were stable for 36 years. We also examined whether variation in physical parameters of song over time could be attributed to acoustic adaptation to breeding habitat, using Landsat variables as a proxy for vegetation characteristics of each male’s breeding territory. The physical parameters of the songs, which changed little over time, revealed no coherent relationships with the Landsat variables and therefore little evidence for acoustic adaptation.