Migration and the evolution of duetting in songbirds

David M. Logue , Michelle L. Hall


Many groups of animals defend shared resources with coordinated signals. The best-studied of these signals are the vocal duets produced by mated pairs of birds. Duets are believed to be more common among tropical-breeding species, but a comprehensive test of this hypothesis is lacking, and the mechanisms that generate latitudinal patterns in duetting are not known. We used a stratified sample of 372 songbird species to conduct the first broad-scale, phylogenetically explicit analysis of duet evolution. We found that duetting evolves in association with the absence of migration, but not with sexual monochromatism or tropical breeding. We conclude that the evolution of migration exerts a major influence on the evolution of duetting. The perceived association between tropical breeding and duetting may be a by-product of the migration–duetting relationship. Migration reduces the average duration of partnerships, potentially reducing the benefits of cooperative behaviour, including duetting. Ultimately, the evolution of coordinated resource-defence signals in songbirds may be driven by ecological conditions that favour sedentary lifestyles and social stability.

1. Introduction

Coordinated signals are used to defend shared resources in diverse taxa ranging from social shrimp Synalpheus [1] to lions Panthera leo [2]. The vocal duets and choruses produced by songbirds are particularly striking for the remarkable precision with which individuals coordinate their vocalizations [35]. Researchers have investigated the function of avian duetting [6,7], sexual dimorphism of the brain regions associated with song [8] and the neural basis of coordinated singing [9]. However, factors driving the evolution of duetting are poorly understood. We investigated the evolution of duetting to advance our understanding of cooperative group signalling and integrate that line of research with the rich literature on the evolution of cooperation.

Intensive experimental research has shown that avian duets function primarily in cooperative defence of territories and mates (reviewed in [6,7]). Consistent with this functional interpretation, there are positive associations between duetting and both pair-bond duration and year-round territoriality in North American birds [10]. Avian duets are thought to be more common in the tropics [11,12], and may be associated with sexual monochromatism [12,13] (but see [10,11]). Conclusions about the evolution of duetting based on existing comparative analyses are limited by the scopes and methodologies of those studies. Early comparative studies did not take phylogeny into account [11,12]. More recent studies included only partial control for phylogenetic non-independence, and were restricted in geographical and phylogenetic scope [10,13].

We hypothesize that seasonal migration may be an important factor driving the evolution of cooperative duetting. Theoretical [14] and empirical (reviewed in [15]) work suggest that long-lasting relationships favour the evolution of cooperative behaviour. Factors that facilitate long-term relationships are therefore predicted to promote the evolution and maintenance of cooperation, whereas those that disrupt relationships are likely to drive the evolutionary loss of cooperation. The cooperative functions of duetting [6,7] may be of limited benefit in migratory songbirds, which have shorter pair bonds owing to higher rates of divorce between breeding seasons [16]. Migration also has the potential to explain other previously identified correlates of duetting, since migration covaries with latitude [17] and is associated with sexual dichromatism [18]. Our hypothesis predicts a negative correlation between the evolution of duetting and migration that is stronger than correlations between duetting and either breeding latitude or sexual monochromatism.

We investigate the evolution of duetting with a broad-scale comparative analysis. We include both duets and choruses in our analysis (using the term ‘duets’ inclusively, for convenience) because the two are related both structurally (vocalization by one individual prompts a coordinated vocal response by one or more other individuals) and functionally (both function in cooperative resource defence; e.g. [19]). We sample across the entire songbird phylogeny (suborder: Passeri), and test for correlated evolution between duetting and migration, breeding latitude and sexual monochromatism. Our findings challenge long-standing hypotheses about the evolution of duetting, and reveal a previously untested factor affecting the evolution of this cooperative behaviour.

2. Material and methods

Our basic approach involved (i) sampling songbird species, (ii) acquiring phylogenies of the sampled species, (iii) scoring characters of interest and (iv) testing for the independent evolution of duetting versus each of the three other characters.

(a) Sampling

Our stratified sampling strategy was designed to maximize sampling breadth and evenness, so that we could discern major patterns in the evolution of duetting in songbirds. Samples (n = 372) were chosen by randomly selecting 20% of species (rounding up) from each of the 60 clades in Jønsson and Fjeldså's supertree of oscines [20].

(b) Characters

We used the Handbook of the birds of the world (HBW [21]) to score the following characters: duetting, migration, sexual monochromatism and breeding latitude. The HBW comprises 16 edited volumes of species and family accounts, covering all known bird species. The referenced species accounts, each of which includes sections dedicated to vocalizations, appearance and movements, as well as range maps, are written by experts in the focal taxa. We scored all traits as discrete, binary characters. Data were gathered by D.M.L. and two trained research assistants. One person (D.M.L.) reviewed every score against the source to ensure consistent scoring.

Our goal when scoring duetting was to include all species that produce loud vocalizations in temporal coordination with the pair mate or other group members (this definition is designed to exclude contact calls in foraging flocks). Duetting was scored as present when the HBW described vocalizations with the following language: ‘duet’ (n = 20), ‘chorus’ (n = 5), ‘antiphonal’ (n = 4), one mate vocally ‘responds to’ or ‘accompanies’ the other's vocalizations (n = 4), ‘loud’ vocalization ‘given by several birds at one time’ (n = 1) or pairs ‘countersing’ (n = 1). To address the concern that duetting may be under-reported in the HBW, M.L.H. searched written reports and online recording repositories for additional information on the vocal behaviour of the 333 species initially scored as non-duetters. She found additional information on 250 species, prompting us to re-score 23 species (9.2%) as positive for duetting. To account for the lingering possibility of false negatives, we conducted all analyses both with and without the putative non-duetters for which we did not have an additional source (n = 83). Removing these species did not affect the statistical significance of any tests, so we report the results of the tests that include them (results of the alternative tests are in the electronic supplementary material, table S2).

Migration was scored as present when the HBW indicated that the species made ‘medium-distance’, ‘long-distance’ or ‘seasonal’ movements, or was ‘nomadic’ over a large range. Species exhibiting ‘altitudinal’ or ‘local’ movements, ‘irruptions’, ‘local nomadism’, ‘intra-tropical migration’ and occasional sightings outside of the primary range were scored as non-migratory. Species that exhibited a mixed migratory strategy (n = 90) and those for which there were no movement data (n = 7) were excluded from the test for independent evolution of duetting and migration (n = 275) because we do not know whether vocal behaviour was characterized in migratory or non-migratory birds. We scored sexual monochromatism as absent if there were any plumage differences between the sexes that were visible to the human observer. We used range maps in the HBW together with Google Earth v. 6.0 (Google Inc., 2011, freely available at http://www.google.com/earth) to identify the maximum and minimum latitude in each species’s breeding range. We scored species as tropical if their breeding range was entirely within the tropics, and temperate if it was entirely out of the tropics. Species that breed in both temperate and tropical latitudes (n = 150) were excluded from the test for independent evolution of duetting and breeding latitude (n = 222). The complete dataset is available in the electronic supplementary material (table S3, and summarized in table S1).

(c) Phylogeny

We based each of our analyses on many trees to account for phylogenetic uncertainty. We used an online phylogeny-generating tool (http://birdtree.org) to construct sets of 1000 phylogenies for each analysis. This tool is based on a recent phylogeny of all 9993 known species of extant birds [22]. Trees are represented in the ‘tree pool’ according to their posterior probabilities (W. Jetz 2014, personal communication), so sampling many trees eliminates the need to weight results by each tree's posterior probability [23]. Tree files used for each analysis are archived in the electronic supplementary material.

(d) Analysis

If a character evolves slowly enough that closely related taxa tend to share character states, the character is said to exhibit ‘phylogenetic signal’. The probable existence of phylogenetic signal for duetting motivates our use of phylogenetically explicit analyses [24]. We quantified phylogenetic signal with Pagel's λ, which transforms the length of the internal branches of the phylogeny. If species are independent with respect to the character of interest, trait evolution can be modelled by a ‘star’ phylogeny in which all taxa evolve directly from the common ancestor. This situation is simulated by setting λ = 0. At the other extreme, if there is constant-variance (Brownian) evolution over the estimated topology, λ = 1. We used the random walk model of evolution in BayesTraits v. 2 beta (freely available at www.evolution.rdg.ac.uk) to generate maximum-likelihood estimates of λ for 1000 trees in the full dataset (n = 372). We then ran the analysis again, forcing λ = 0. Based on the hypothesis that duetting exhibits phylogenetic signal, we predicted that the likelihoods of the models with freely estimated λ would significantly exceed those with λ = 0.

We tested for independent character evolution with Pagel's [25] test for the independent evolution of discrete binary traits, as implemented by BayesTraits. This analysis tests the hypothesis that two binary characters have evolved independently of one another, and estimates evolutionary transition rates between character states. We constructed two models to describe the evolution of the two characters. In the four-parameter model, each character is assumed to evolve independently. The four parameters are the instantaneous transition rates from state 0 to state 1 and from state 1 to state 0 for each of the two characters. The eight-parameter model allows for character evolution to be correlated by estimating separate transition rates for gain and loss of one character given each state of the other character. We used the maximum-likelihood function in BayesTraits to estimate the log likelihood of each of the two models, and the transition rates (q) for each of the 1000 phylogenies in the analysis.

We used the log-likelihood values averaged over the 1000 trees to calculate a likelihood ratio (2 × difference in log-likelihood of the two models) [25]. Likelihood ratios follow the χ2-distribution with d.f. = the difference in number of parameters in the two models (test for phylogenetic signal: d.f. = 1; correlation tests: d.f. = 4). We used both the χ2-test and the difference in Akaike's information criterion (ΔAIC) to test hypotheses of phylogenetic signal and correlated trait evolution. We concluded that the more complex model was supported when p < 0.05 and ΔAIC > 2.

3. Results

We estimate that duetting occurs in 15.6% of songbird species (58 of 372 species), representing 43.8% of songbird families (28 of 64 families). We found significant phylogenetic signal for duetting (average Pagel's λ = 0.57, 95% C.I. over 1000 trees = 0.45–0.69, average log-likelihood ratio = 26.9, d.f. = 1, p < 0.0001, average ΔAIC = 24.9). The families with at least two duetting species in our dataset are the Meliphagidae (seven species), Timaliidae (six), Troglodytidae (five), Emberizidae (four), Thraupidae (three), Cisticolidae (three), Icteridae (two), Laniidae (two), Malaconotidae (two), Ploceidae (two), Sturnidae (two), Sylvidae (two) and Vangidae (two).

We found a significant negative evolutionary relationship between duetting and migration (table 1 and figure 1). The estimated transition rates away from ‘duetting and migrating’ through loss of migration (q42) and loss of duetting (q43) are high relative to the other estimated transition rates (figure 1). Only one species (scarlet tanager, Piranga olivacea) was scored as both duetting and migrating. We did not find evidence of an evolutionary association between duetting and sexual monochromatism or duetting and breeding latitude (table 1). Overall, we found high levels of migration and low levels of duetting in high North-temperate latitudes (figure 2).

View this table:
Table 1.

The results of maximum-likelihood tests for independent character evolution in a sample of 372 songbird species.

Figure 1.

A schematic of estimated transition rates (q) between character states. States are defined by the presence (+) or absence (–) of the characters duetting (D) and migration (M). Arrows represent evolutionary transitions between states, and arrow width is proportional to the log of the estimated transition rate.

Figure 2.

The relationship between duetting, migration and breeding latitude in a sample of 372 songbirds. The proportion of species that duet (black lines) and migrate (grey lines) are plotted as a function of the latitudinal centre of their breeding range. Tropical latitudes are shaded. Numbers in parentheses indicate the number of species that contribute to the duetting and migrating data, respectively. Sample sizes differ because species with a mixed migratory strategy were excluded from the estimates of migration frequency. The two most extreme bins were collapsed because they included few cases.

4. Discussion

(a) Taxonomic distribution of vocal duetting

Our analysis of a large sample of randomly selected songbird species suggests that, based on the currently available information, duetting occurs in 15.6% of songbirds, exceeding previous estimates (4.3% of all birds in [7], 7% of North American passerines in [10]). These estimates may change as increasing interest in female song and duetting improve the quality of available data. We confirmed the existence of the three major clades of duetting songbirds (wrens, honeyeaters and bush-shrikes) that were identified by Farabaugh [11]. In addition, we identified clusters of duetting songbirds in the babblers, cisticolas, New World blackbirds, New World sparrows, Old World warblers, shrikes, starlings, tanagers, vangas and weavers. It is likely that there are other important clusters of duetting songbirds that were not identified owing to the limits of our sampling. Our finding of moderate phylogenetic signal of duetting in the songbirds is consistent with the finding that ancestry influences duet structure in the taxonomic cluster of wrens formerly belonging to the genus Thryothorus [26,27]. Both results affirm both the necessity of controlling for phylogeny in comparative analyses of duetting and the importance of considering ancestry in single-species studies of duet function.

(b) Duetting and migration

We identified a strong negative correlation between duetting and migration. Duetting and migrating are an evolutionarily unstable state, with high estimated values of transition rates away from this state (q43 and q42; wide arrows in figure 1). The scarlet tanager is the only species that we scored as both duetting and migrating, and not all sources agree that the species duets [10]. The negative relationship between duetting and migration appears robust, but we interpret this finding with a degree of caution because knowledge of vocal behaviour in some species is poor. We hope that our novel finding stimulates more rigorous documentation of duetting behaviour in birds, as well as studies that examine the evolution of duetting and migration in well-studied groups. We propose three mutually compatible hypotheses that attempt to explain why cooperative maintenance of territories and partnerships by duetting is evolutionarily incompatible with migration in songbirds.

Our first hypothesis is that migration increases the likelihood of divorce between breeding seasons, limiting the duration of pair bonds and reducing the benefits of cooperation between pair mates. Consistent with this idea, comparative analyses show that divorce rates are higher in migratory than resident species [16] and higher in species that do not stay together year-round than in those with continuous partnerships [28]. Migration is negatively related to group size in the suboscine clade Tyranni, leading Boyle & Conway [29] to suggest that the evolution of migration ‘could impede the maintenance of pair and family group foraging bonds’.

Pair-bond duration could influence the evolution of cooperative duetting between pair mates in at least two ways. First, longer pair bonds will result in more repeated interactions between partners, favouring the evolution of cooperation [14,15,30]. Second, long-term partnerships may promote cooperation if pair mates that stay together for many breeding efforts have more common interests (e.g. the mate's continued well-being [31,32]) than those that do not.

Partner choice is a key component to many theoretical models of the evolution of cooperation because it can punish defectors and facilitate assortative pairing between cooperators [33]. Our second hypothesis is that a non-migratory lifestyle favours the evolution and maintenance of cooperative duetting by facilitating partner choice via strategic partner-switching. The cost–benefit ratio of partner-switching during the breeding season may be higher for migratory species if (i) breeding seasons are shorter for migratory species [34, p. 81] or (ii) migratory birds are less likely than non-migratory birds to keep their new partner through the subsequent breeding season [16].

Our third hypothesis to explain the negative relationship between duetting and migration is that migration favours divergent sex roles, such that only males invest heavily in resource defence. Evidence for less similar sex roles in migratory species includes the links between migration and high testosterone levels in males [35], and the absence of song and colourful plumage in migrating females [18,36].

(c) Duetting and sexual monochromatism

We conclude that there is currently insufficient evidence to support the long-held hypothesis that duetting evolves in correlation with sexual monochromatism in songbirds [12,37]. Interestingly, sexual dichromatism is strongly correlated with migration in New World orioles (genus Icterus), with dichromatism arising through a loss of female plumage elaboration in migratory species [18]. That report emphasized that different evolutionary pressures may promote monochromatism in which both sexes are ornamented versus monochromatism in which both sexes are plain, and that dichromatism can evolve when either sex gains or loses ornamentation. Our ‘dichromatic’ species covered the full range of dichromatism that is visible to human observers, with the sexes ranging from almost identical to radically different. Our ‘monochromatic’ species included species where the sexes were equally ornamented and equally plain. Because our scores rely on human colour perception, we have probably overlooked both subtle cases of dichromatism and dichromatism involving short wavelengths that would be apparent to avian visual systems [38]. Different results might be obtained from an analysis that acknowledges a range of variation, from ornamentation to plainness, in both sexes, and defines dichromatism according to avian visual systems [39].

(d) Duetting and breeding latitude

Our analysis suggests that the widely perceived correlation between duetting and breeding latitude [40] represents a non-causal relationship generated by the relationship between migration and latitude (figure 2). Previous studies that linked breeding latitude and duetting did not control for phylogeny or migration, nor did they include as wide a distribution of species (both phylogenetic and geographical) as this study. We encourage further research on the relationship between migration, latitude and duetting at both broader and narrower phylogenetic scales.

5. Conclusion

Our data suggest that the evolution of migration is a major driver of the evolution of duetting in songbirds. We hypothesize that migration reduces the average duration of partnerships, thereby reducing the likelihood that cooperative behaviours such as duetting will evolve and be maintained. Given that ecological variation drives the evolution of migration [34], ecological stability may indirectly influence the evolution of duetting and other cooperative behaviours between mated songbirds by promoting year-round residency (i.e. the absence of migration). Our findings help to link duetting research to the broader context of research on cooperative behaviour, and establish a tentative link between ecological stability and the evolution of cooperation in songbirds.


Discussions with Nico Franz, Walter Jetz, Daniel Krupp, Raul Machiavelli, Mark Pagel, Jordan Price and Gavin Thomas benefitted the development of this manuscript. The library at the University of Puerto Rico, Mayagüez, and the comments from the editor and referees were very helpful as well. We are grateful to Wildeby Borges, Noelia Nieves, Hans Cancel, Nardia Sierra, Maria Cruz, Crisanta Serrano and Cristina Garcia for their help gathering data.

  • Received January 16, 2014.
  • Accepted February 12, 2014.


View Abstract