Global macroevolution and macroecology of passerine song

Abstract

Studying the macroevolution of the songs of Passeriformes (perching birds) has proved challenging. The complexity of the task stems not just from the macroevolutionary and macroecological challenge of modeling so many species, but also from the difficulty in collecting and quantifying birdsong itself. Using machine learning techniques, we extracted songs from a large citizen science dataset, and then analyzed the evolution, and biotic and abiotic predictors of variation in birdsong across 578 passerine species. Contrary to expectations, we found few links between life-history traits (monogamy and sexual dimorphism) and the evolution of song pitch (peak frequency) or song complexity (standard deviation of frequency). However, we found significant support for morphological constraints on birdsong, as reflected in a negative correlation between bird size and song pitch. We also found that broad-scale biogeographical and climate factors such as net primary productivity, temperature, and regional species richness were significantly associated with both the evolution and present-day distribution of bird song features. Our analysis integrates comparative and spatial modeling with newly developed data cleaning and curation tools, and suggests that evolutionary history, morphology, and present-day ecological processes shape the distribution of song diversity in these charismatic and important birds.

Song complexity—no correlation between standard deviation of frequency and traditionally used song complexity metrics in passerines: A comment on Pearse et al. (2018)

“…Surprisingly, Pearse et al. () did not validate their novel song complexity scores against previously reported data on song complexity derived from single‐species repertoire size estimates.…”

“…While we appreciate the novel approach used to analyze such a large birdsong database, as well as the attempt to find simple metrics for quantifying birdsong complexity across large spatial and taxonomic scales, we believe a likely reason for the discrepancy between the results of previously published works and those presented by Pearse et al. () has been missed.…”

“…In a recent paper, Pearse et al () used a large citizen science dataset, Xeno‐canto (http://www.xeno‐canto.org), and machine learning techniques to characterise avian song characteristics, including song complexity. While they used several metrics to capture this parameter in passerines, all were intercorrelated; hence, they used a single metric as the main surrogate for song complexity, defined as SD of frequency (hereafter SDF).…”

“…Pearse et al. () found only weak support for correlation between life‐history traits such as monogamy and sexual dimorphism and the evolution of song complexity. Surprisingly, using comparative and assemblage‐based analyses, they found a significant positive link between climate factors such as net primary productivity (NPP) and NPP and temperature, respectively, and song complexity, i.e., that birds with more complex songs occupied more productive (and presumably more complex) environments and warmer areas than species with less elaborate songs.…”

“…Pearse et al. () attempted to use simple automatically derived metrics, estimated as SDF, as a substitute for these birdsong complexity metrics. Surprisingly, Pearse et al.…”