Ecological drivers of song evolution in birds: Disentangling the effects of habitat and morphology

Abstract

Environmental differences influence the evolutionary divergence of mating signals through selection acting either directly on signal transmission ("sensory drive") or because morphological adaptation to different foraging niches causes divergence in "magic traits" associated with signal production, thus indirectly driving signal evolution. Sensory drive and magic traits both contribute to variation in signal structure, yet we have limited understanding of the relative role of these direct and indirect processes during signal evolution. Using phylogenetic analyses across 276 species of ovenbirds (Aves: Furnariidae), we compared the extent to which song evolution was related to the direct influence of habitat characteristics and the indirect effect of body size and beak size, two potential magic traits in birds. We find that indirect ecological selection, via diversification in putative magic traits, explains variation in temporal, spectral, and performance features of song. Body size influences song frequency, whereas beak size limits temporal and performance components of song. In comparison, direct ecological selection has weaker and more limited effects on song structure. Our results illustrate the importance of considering multiple deterministic processes in the evolution of mating signals.

Keywords: Furnariidae; acoustic adaptation; biomechanical constraints; bird song; speciation; stochasticity; trade‐offs.

Figure 1

Phenotypic traits of ovenbirds. Exemplar data used in this study, illustrated for a single species (Brown cacholote, Pseudoseisura lophotes). (A) Morphological measurements collected from museum specimens, including beak depth (a), width (b) and length (c), tarsus length (d to e), and body mass (f). (B) Spectrogram of song segment indicating acoustic traits measured, including duration (g–h), pace (song duration/number of notes), peak frequency (i), maximum frequency (j), minimum frequency (k), and frequency bandwidth (j–k). (C) Frequency bandwidth plotted as a function of pace with the upper‐bound regression for the Furnariidae (y = −79.374x + 5066.2) and the orthogonal distance (vocal deviation) for a song of P. lophotes (l), which has comparatively lower vocal performance than song of many other ovenbird species, for example, Schizoeaca fuliginosa (m). Photograph by Mario Fiorucci; song file downloaded from www.xeno-canto.org (XC151258)

Figure 2

Phylogenetic hypothesis and habitat preferences for the ovenbird radiation. Colored bars show two different types of habitat data associated with tree tips. Height of bars indicates value of Environmental PC1 extracted from geographical range polygons; color‐coding of bars reflects habitat type categories generated from the literature (closed habitats = green; semi‐open habitats = blue; open habitats = yellow)

Figure 3

Diversification of song traits is differentially impacted by habitat and morphology. Plotted values indicate the region of predicted song trait values (Song PC1–3 and vocal performance) for 95% of the observed morphological measurements (beak size and body size) within each habitat type (closed, semi‐open, and open). All song traits increase along the y‐axis (i.e., larger y‐values indicate higher frequency, longer duration, faster pace, higher performance). Labels of column pairs indicate song traits that loaded most strongly onto Song PCs (e.g., frequency traits on Song PC1). Heat maps indicate variation in beak length as a third axis to convey information about predicted beak moment (larger values are more yellow). Informative relationships (ΔAIC < 2 of top model; parameter weights >30%) are indicated in black boxes. Left to right: (Frequency) Larger birds sing lower frequency songs. (Duration) No informative relationships. (Pace) Songs in open habitats are faster, and birds with greater beak moment sing slower songs. (Vocal Performance) Birds with greater beak moment sing lower performance songs, especially in open habitats. AIC, Akaike Information Criterion