Breeding season length predicts duet coordination and consistency in Neotropical wrens (Troglodytidae)

Abstract

Many animals produce coordinated signals, but few are more striking than the elaborate male–female vocal duets produced by some tropical songbirds. Yet, little is known about the factors driving the extreme levels of vocal coordination between mated pairs in these taxa. We examined evolutionary patterns of duet coordination and their potential evolutionary drivers in Neotropical wrens (Troglodytidae), a songbird family well known for highly coordinated duets. Across 23 wren species, we show that the degree of coordination and precision with which pairs combine their songs into duets varies by species. This includes some species that alternate their song phrases with exceptional coordination to produce rapidly alternating duets that are highly consistent across renditions. These highly coordinated, consistent duets evolved independently in multiple wren species. Duet coordination and consistency are greatest in species with especially long breeding seasons, but neither duet coordination nor consistency are correlated with clutch size, conspecific abundance or vegetation density. These results suggest that tightly coordinated duets play an important role in mediating breeding behaviour, possibly by signalling commitment or coalition of the pair to mates and other conspecifics.

1. Background

Animals often use complex, coordinated displays to defend territories and attract mates [1,2]. Such signals range from coordinated courtship displays [3,4] to vocal duets [5–7]. In some tropical antiphonally duetting bird species, males and females combine their songs with such precision that they sound like a single bird singing [8–10]. While a number of studies have investigated the drivers responsible for duet evolution generally, few studies have investigated how and why especially coordinated duets evolved [5,7,11–14]. The remarkable level of duet coordination in some species suggests that precisely timed, highly coordinated duets have evolved for reasons beyond those that have been proposed to explain duet evolution in general [15–18].

Leading hypotheses for the evolution of duets include year-round territory defence, maintaining contact between mates and maintenance of long-term pair bonds [17–21]. The territory defence hypothesis posits that duets signal the ability of a mated pair to jointly defend a territory [11,12], whereas the maintaining contact hypothesis posits that duets allow mated pairs to stay in contact in dense habitats where visual cues would be obscured [11,21]. By contrast, the pair bond maintenance hypothesis suggests that duets help to maintain long-term partnerships and signal commitment, which may aid in synchronizing reproduction and coordinating reproductive behaviours [22,23]. The pair bond maintenance hypothesis has some overlap with the other two hypotheses, as pairs may communicate the strength of their pair bond to conspecifics during territory defence and maintaining pair bonds may be aided by maintaining contact [15,22,24]. Indeed, duets may serve a variety of functions [14]. In Neotropical songbirds, evidence suggests that duets function in territory defence [25–30], contact maintenance [21] and long-term pair bond maintenance [31–33].

Duet evolution is often tied to tropical natural history [23,34]. The seasonal stability and high adult survival in tropical environments enables species to be sedentary, long-term monogamous and territorial year-round [23,35]. Competition with long-term neighbouring conspecifics in turn promotes the evolution of behaviours such as duets that help pairs defend year-round resources and territories [16,34,35]. In addition, many tropical species can breed over extended periods of the year, which may favour the evolution of behaviours that facilitate asynchronous reproduction by enabling pairs to coordinate reproductive activities, including onset of reproduction, parental care and ‘bet hedging’ strategies for when to reproduce based on environmental and partner condition [23,36,37]. The importance of maintaining a social partner, coordinating breeding activities and territory defence over longer periods of the year may lead to the evolution of particularly elaborate signals that allow pair members to signal their commitment to each other or the strength of their pair bond to others [12,15,32,33].

Neotropical wrens (Troglodytidae) offer an exceptional system to study the evolution of duet coordination. Species in this family exhibit diverse singing patterns ranging from long, complex solos to highly coordinated male–female duets and choruses ([7], figure 1). Neotropical wren duets are particularly notable for their temporal coordination, which varies from antiphonal to polyphonal [7–9]. The extent to which polyphonal duet phrases overlap varies among both species and conspecific pairs within species [7,39].

Figure 1.

Phylogenetic relationships and spectrograms of duets for four wren species, showing variation in coordination across the four wren genera in the study. Male songs are highlighted in blue and female songs are highlighted in red. Photos by Daniel Mennill (rufous-and-white wren), Hector Bottai (buff-breasted wren), Christopher Templeton (happy wren) and T.C. Davis (Carolina wren, conspecific to white-browed wren in some taxonomies). Phylogeny based on [38]. (Online version in colour.)

We examined several possible factors responsible for the evolution of highly coordinated duets in the Neotropical wrens [12,13]: (i) breeding season length, (ii) clutch size, (iii) conspecific abundance, and (iv) vegetation density. We hypothesized that longer breeding seasons may select for more coordinated duets because synchronizing breeding for a large part of the year may be benefited by coordinated communication between mated pairs, including partners' ability to signal their commitment to one another or to signal coalition (e.g. strength of the pair bond) to conspecifics [15]. We hypothesized that larger clutch sizes would favour more coordinated duets, assuming that clutch sizes may lead to greater fecundity in long-lived, year-round territorial species. We hypothesized that conspecific abundance would select for increased duet coordination, assuming that more abundant species experience higher levels of conspecific competition for resources and territories. Species with high conspecific abundance or that use duets in intrasexual competition might also develop highly coordinated duets to avoid masking of mates during territory defence [40]. Lastly, following the acoustic-contact hypothesis, we hypothesized that species in particularly dense forest habitats would have higher levels of duet coordination.

We assessed duet coordination and consistency across renditions of duets in 23 species of Neotropical wrens within a phylogenetic framework. We then compared duet coordination and consistency to the above natural history and environmental parameters. Our goal was to evaluate selective pressures that might be responsible for the evolution of highly coordinated duets.

2. Methods

(a). Field recordings

We measured and compared songs and duets for 23 species of wren, representing four closely related genera (figure 1). The recordings were gathered during previous fieldwork by N.I.M. and colleagues to quantify variation in wren duet structure, as described by Mann et al. [7]. To summarize, Mann et al. visited a variety of sites throughout Central and South America, banding and recording 1–6 pairs for each of the 23 wren species between January 2001 and May 2003. Each individual was marked with a unique combination of coloured leg bands and a blood sample was collected to later genetically confirm sex. For each pair, Mann et al. took four separate 90 min recordings on different days, typically in the morning, and noted each song that was sung by each sex. Each recording was paired with notes and spectrograms, which we used to establish when each sex sang within a duet. Mann et al.'s recording sessions sometimes resulted in hundreds of recorded songs per pair. We measured a subset of 10 randomly chosen duets per pair for each pair that was recorded. For the few species without 10 duets available, we measured all instances of duets. There were also a few species for which there were no instances of duets in the original data, so we selected solo vocalizations for our dataset. This resulted in a final dataset of 162 solos and 457 duets measured for the 23 species in the current study (see the electronic supplementary material, table S1). We defined a duet as any instance in which a pair member performed its song immediately following singing by its mate, either in an overlapping or alternating pattern [5,12,13].

(b). Sound analysis

Neotropical wren songs and duets typically consist of phrases, which are stereotyped groupings of 2–15 notes that each singer repeats throughout its song [1]. We used Raven Sound Analysis Software v. 1.5 [41] to select individual notes that comprised each phrase, as well as every phrase for each song within a duet. From these selections, we measured the beginning/end times for each note and phrase, the lengths of notes and phrases, the overlap between notes and between phrases, and the lengths of gaps between phrases (inter-phrase intervals). We extracted these values for males and females separately, so that we could compare timing of male and female songs within a duet.

(c). Coordination score

To statistically evaluate duet coordination, we used the Monte Carlo randomization test implemented in the coor_test function from the R package warbleR [42,43]. The function used the start and end times of vocalizations within a vocal bout (duet) to differentiate among song components and the intervals in between the songs. It then compared the observed duration of overlap between notes (and/or phrases) within a duet (i.e. the sum of the durations from all overlaps in a duet) to a null distribution of overlap durations calculated by repeatedly shuffling both signals and gaps (the silent space between notes or phrases). Null distributions were created using 10 000 iterations. A p-value was calculated as the proportion of random overlap durations that were equal to or more extreme than the observed value. We also calculated a duet coordination score, defined as the proportional difference between the observed overlap durations and those expected by chance according to our randomization procedure ((obs − exp)/exp). A coordination score of zero represented no coordination while larger integers (positive or negative) represented increasing coordination. Negative values corresponded to coordinated–alternating while positive values corresponded to coordinated–overlapping duets [44]. We calculated a duet coordination score for each duet and then averaged coordination scores by species. We measured duet consistency as the 95% confidence interval surrounding the mean coordination score across each species. Species without duets (soloing species) were assigned a score of zero, indicating that they had uncoordinated singing. We calculated coordination scores for both phrases and notes within duets, with similar results. Here, we present the results for male–female note placement within duets. These species-level mean note coordination and consistency scores were used for subsequent phylogenetic comparative analyses (electronic supplementary material, table S2).

(d). Natural history data collection

(i). Breeding season length and clutch size

We gathered information on breeding season lengths and clutch sizes from Handbook of the birds of the world [45]. Breeding season length was counted as the total number of months listed for which each species has been observed breeding, and clutch size was counted as the number of eggs per clutch reported for each species. When more than one breeding season length or clutch size was provided for different parts of a species range, we took the median value for breeding season length and the average value for clutch size from the values listed.

(ii). Conspecific abundance

We calculated species prevalence as a proxy for conspecific abundance throughout each species' range using eBird data. Prevalence was calculated as the relative frequency of occurrence on eBird checklists within forested areas. To calculate this, we extracted all complete eBird checklists within each species range from 2010 to 2019 that were less than 5 km in length and 5 h in duration. We used BirdLife International's range polygons to define each species' range. To reduce the impact of spatial and temporal bias in the eBird data, we overlaid an equal area hexagonal grid with 5 km spacing between cell centres and assigned each checklist to a cell. Within each hexagonal cell for each year, we calculated the frequency of detection of the species on eBird checklists, then calculated the mean frequency across all hexagons to get a single prevalence value for each species. We restricted our estimates to hexagons that contained at least 500 m² of forest within a 2 × 2 km neighbourhood around an eBird observation. Forested habitat was determined based on land cover data from the NASA MODIS satellite (product names: MCD12Q1 v. 6).

(iii). Vegetation density

Vegetation density was calculated from Enhanced Vegetation Index (EVI) data extracted from the NASA MODIS satellite (product: MOD12A3 v6). We used EVI over Normalized Difference Vegetation Index (NDVI), as EVI has higher resolution for habitats with dense vegetation (e.g. rainforest). We calculated both mean EVI and the standard deviation of EVI for each species based on annual 2 km resolution MODIS data. Using the same set of eBird checklists assigned to hexagonal cells as described for calculating species prevalence, we extracted EVI values for the location of each checklist using the MODIS data from the year of the checklist. The annual MODIS EVI data for 2019 were unavailable at the time of analysis, so 2018 EVI data were used for 2019 eBird checklists. To reduce spatial and temporal bias, for each hexagonal cell and each year, we calculated the mean EVI for checklist locations in which the species occurred. Finally, we calculated the mean and standard deviation of EVI across all hexagonal cells and years. We used the current best practices for both prevalence and vegetation density estimates [46].

(e). Phylogenetic reconstruction

We examined evolutionary patterns of duet coordination and consistency using the phytools and geiger packages in R [43,47,48]. We evaluated the robustness of ancestral state reconstructions of duet coordination and consistency by comparing several reconstruction methods, including squared-change parsimony, maximum likelihood (ML) and Bayesian reconstructions (using the ace and anc.bayes functions in phytools) [47]. We also assessed the fit of alternative underlying evolutionary models to evolutionary patterns of duet coordination and consistency, including Brownian motion (BM), Ornstein–Uhlenbeck, early burst models and Pagel's lambda, kappa and delta models with the fitContinuous function in geiger [48]. This verified that our data met assumptions of most reconstruction methods of an underlying distribution similar to BM (electronic supplementary material, table S3). Reconstructed ancestral states were visualized and compared among methods using both the phenogram function in phytools to focus on the state of the root node and the contMap heat map function in phytools to compare how states changed over time [47]. All reconstruction and comparative analyses were conducted on a phylogenetic tree for the wrens with good coverage for species in this study [38].

(f). Phylogenetic correlative analyses

We compared evolutionary patterns of duet coordination and consistency to natural history predictors (breeding season length, clutch size, conspecific abundance and vegetation density). Comparative correlational analyses were conducted using phylogenetically controlled Bayesian models with the R package MCMCglmm [49]. For all models, we used a non-informative, parameter-extended prior to improve mixing with R-structure V = 1, nu = 0.002 and G-structure V = 1, nu = 1, alpha.mu = 0, and alpha.V = 25². We also ran all comparative analyses as phylogenetic least squares (PGLS), including selection and evaluation of the best underlying model of evolution (electronic supplementary material). Duet coordination, consistency and all predictor variables were compared visually on the phylogeny with the contMap function using the ML reconstruction in phytools [47]. The results were qualitatively equivalent for both sets of correlative analyses. Phylogenetic signal was calculated using both Blomberg's K [50] and Pagel's Lambda [51] with the phylosig function in phytools [47].

3. Results

Most Neotropical wren species produced duets classified as coordinated–alternating, meaning that males and females alternated their song phrases and overlapped each other significantly less than expected by chance (figure 2; electronic supplementary material, table S2). However, there was considerable variation among species in the extent to which song phrases overlapped within duets. Interestingly, species with the most coordinated duets were not each other's closest relatives (figure 2). The highest levels of coordination in alternating duets (i.e. the most negative coordination scores) were observed in Cantorchilus modestus, Cantorchilus leucotis, Cantorchilus nigricapillus and Pheugopedius euophrys. The lowest levels of coordination (i.e. scores closest to zero) were observed in Thryothorus albinucha, Cantorchilus leucopogon, Thryophilus pleurostictus, Thryophilus sinaloa and Thryophilus rufalbus, all of which are solo singers except T. rufalbus. Solos were all scored as uncoordinated, but T. rufalbus duets were just as uncoordinated on average as the songs of soloist species (figure 2; electronic supplementary material, table S2).

Figure 2.

(a) Maximum-likelihood ancestral state reconstruction of duet coordination and (b) means and standard deviations for duet coordination scores by species. Coordination scores closer to −1 (red) indicate coordinated duets in which partners alternate their vocalizations and values closer to 0 (blue) indicate uncoordinated duets. Thryophilus pleurostictus, Thryophilus sinaloa, Cantorchilus leucopogon and Thryothorus albinucha are primarily soloists and, therefore, were assigned coordination scores of zero. Phylogeny based on [38]. (Online version in colour.)

Most Neotropical wrens also showed high levels of within-species consistency in duet coordination, indicated by low confidence intervals for the coordination scores (figure 2; electronic supplementary material, figure S1 and table S2). As with the coordination scores, consistency varied across species, with Cantorchilus guarayanus displaying notably inconsistent duets.

Ancestral state reconstruction indicated that the common ancestor of these Neotropical wrens probably produced duets that were slightly coordinated and alternating (figure 2; electronic supplementary material, figure S2). However, the strength of this result varied with reconstruction method: ML and Bayesian approaches reconstructed the ancestor as having an intermediately coordinated duetting ancestor, whereas parsimony suggested an uncoordinated or non-duetting ancestor (figure 2; electronic supplementary material, figure S2 and table S3). As ML and Bayesian approaches take branch lengths into account and maximize parameters to reflect the distribution of the data, we have greater confidence in these approaches.

Phylogenetically controlled comparative analyses revealed that duet coordination and duet consistency were both positively correlated with breeding season length (p = 0.019 and p = 0.032, respectively; table 1), with more coordinated and consistent duets occurring in species with longer breeding seasons (figure 3). However, neither coordination nor consistency were correlated with other natural history traits, including clutch size (p = 0.328 and 0.269), species abundance (p = 0.575 and 0.826), vegetation density (p = 0.886 and 0.730) or variation in vegetation density (p = 0.639 and 0.333; table 1; electronic supplementary material, figure S3). These results were also supported by phylogenetic least squares using a model selection approach (electronic supplementary material, tables S5 and S6) and were robust to removal of species without duets (electronic supplementary material, table S7).

Table 1.

Results from phylogenetically controlled Bayesian models used to evaluate correlated evolution of duet coordination and consistency versus predictor variables. (Values in italics are statistically significant according to a Bayesian p-value < 0.05.)

explanatory variable	predictor variable	posterior mean	lower 95% CI	upper 95% CI	effective sample size	Bayesian p-value
coordination	breeding season length	−0.21	−0.39	−0.04	49 177	0.021
	clutch size	0.08	−0.09	0.26	48 988	0.326
	species abundance	−0.04	−0.18	0.11	49 984	0.571
	EVI mean	0.01	−0.13	0.15	49 984	0.900
	EVI s.d.	−0.03	−0.17	0.10	49 984	0.637
consistency	breeding season length	−0.12	−0.23	−0.01	49 984	0.031
	clutch size	−0.16	−0.44	0.13	43 541	0.263
	species abundance	0.03	−0.22	0.28	41 342	0.829
	EVI mean	−0.03	−0.24	0.17	49 984	0.724
	EVI s.d.	−0.10	−0.30	0.11	47 518	0.333

Figure 3.

Duet coordination and consistency are both greater in species with longer breeding seasons. More negative coordination scores represent species with more strongly coordinated, alternating duets. Species with duet consistency values closer to zero have more consistently coordinated duets (lower 95% confidence intervals). Note that the y-axes of both graphs are flipped to more readily show that coordination and consistency increase with breeding season length. Black points in each plot represent species with coordination and consistency score data. (Online version in colour.)

Phylogenetic signal for duet coordination was high (K = 1.23, lambda = 1.28), whereas phylogenetic signal for duet consistency was relatively low (K = 0.77, lambda < 0.001). This is consistent with findings by Mann et al. [7] that there are similarities in duet coordination and overlap within genera.

4. Discussion

While most Neotropical wrens we examined produced coordinated and alternating duets, the degrees of coordination, consistency and amount of overlap of song phrases varied considerably among species. Some species produced duets with exceptionally high precision and consistent coordination, and the species with the most coordinated duets (e.g. C. leucotis, C. nigricapillus, and P. euophrys) were not each other's closest relatives. In fact, we demonstrated that exceptionally coordinated duets have independently evolved multiple times within Neotropical wrens and appear to be a derived state exhibited by only some species. Of all the natural history features we examined, only breeding season length explained a species' degree of duet coordination and consistency. This result suggests that highly coordinated duets might benefit species that breed over longer parts of the year, perhaps by helping them coordinate breeding activities or communicate strength of the pair bond.

(a). Duet coordination and breeding season length

At least two hypotheses have been proposed for how duet coordination could reflect pair bond strength: signalling commitment to a partner and signalling pair coalition to conspecifics [12–14]. The rationale for both hypotheses is that temporal coordination of duets is challenging to coordinate, requiring individuals to continually listen to their partner's song and modify their own output [32,52,53]. The addition of male and female-specific song phrases [7], as well as pair-specific answering rules (duet codes) [54], both commonly found in Neotropical wrens, create further challenges for coordinating duet songs [55]. Recent evidence also indicates that duet coordination is learned over time in Neotropical wrens [33,55–57]. By taking the time and effort to learn to coordinate songs, both partners are signalling their commitment or willingness to invest in territory defence, reproductive efforts and parental care [12,22,32,55].

Longer breeding seasons could also promote commitment to a pair bond if reproductive success increases with pair bond length and throughout the season. In some monogamous species, reproductive success does increase with subsequent broods and length of the pair bond [58]. While many Neotropical wren species form long-term pair bonds with especially low levels of extra-pair mating [59], pair bond length varies across pairs [29,31,57,60]. In buff-breasted wrens (C. leucotis), a species with especially coordinated duets and long breeding seasons, pairs that have bred together show high mate and territory fidelity, whereas individuals newly entering the population are much more likely to divorce and often do so within the first several months of a partnership [61]. There may be true reproductive benefits for an individual to accurately assess its partner's commitment to the pair bond.

Alternatively, the fact that duets are learned and become more coordinated over time could signal to other conspecifics the length or strength of the pair bond and, therefore, the ability of a pair to jointly defend a territory [15,18]. In at least some Neotropical wren species, more coordinated duets are not perceived as more threatening signals [39]; however, it seems plausible that conspecifics could obtain additional information about the quality or strength of a pair from the fine structure of the duets [25–30]. Furthermore, signalling commitment and coalition are not necessarily mutually exclusive functions. Duets are known to be multifunctional, and it seems that duet coordination could simultaneously benefit pairs and individuals as a signal both within and between pairs [14].

Duet coordination could also be indicative of a pair's ability or willingness to coordinate reproductive activities, including reproductive synchrony [12,14]. Yet, some species duet well outside of the breeding season, suggesting synchronizing reproductive behaviour is not the sole function of duets in those species [62,63]. Nevertheless, both male and female songs are known to stimulate gonadal growth and bring birds into breeding condition in some species [64,65], and female song can be used to coordinate breeding activities such as nest relief and care for young [66–68].

Another alternative is that high levels of duet coordination arise in species with longer breeding seasons because of ‘masking avoidance’. Under this scenario, the efforts of each pair member to ensure that its own song is heard leads to duets with increasing coordination [40,69,70]. This phenomenon could be particularly pronounced in species with longer breeding seasons because these individuals may duet over longer periods of the year and, therefore, achieve greater coordination through enhanced masking avoidance. In addition, species with longer breeding seasons may face more intraspecific and intrasexual competition for resources, creating a heightened need for each pair member to most effectively compete with same or opposite sex intruders. However, we did not find evidence that duet coordination varies with intraspecific competition (e.g. high population densities) and other metrics of intrasexual conflict (high extra-pair paternity or divorce rates [37,40,57]), which seem to be low for these wrens [59,71]. Nevertheless, territory defence appears to be a primary function of duets in general [11,12,69] and our coarse metric for assessing population density could have missed the importance of intraspecific competition on duet structure. Indeed, we think masking avoidance is a compelling and parsimonious hypothesis that could influence duet coordination in these and other duetting species and we encourage more studies that directly test the contributions of masking avoidance and signalling commitment to duet coordination in a range of species [40].

Similarly, Neotropical wren species with longer breeding seasons may have more coordinated duets simply because they have practised their duets over a longer period of the year. Wrens learn to coordinate both duet codes and response timing [33,57]. However, in the few species studied, wrens achieve a high level of duet coordination within the first few days of a new partnership [55,56]. Therefore, it seems unlikely that the additional practise during longer breeding seasons is a primary factor leading to more coordinated duets in species with long breeding seasons. In addition, duets are not tied exclusively to the breeding season for many duetting species, so most wren species would have opportunities to practise duet coordination throughout the year. Lastly, if practising was the only factor influencing duet coordination, we would expect more species-level variation in general, as we presumably sampled individuals of each species that had been together for different lengths of time.

(b). Duet coordination and tropical life history

Prevailing theory and empirical data support duet evolution being tied to natural history traits that are particularly common in the tropics, including long-term monogamy and year-round territoriality [17–20,34,35]; however, to our knowledge, all species in the current study are year-round territorial. Therefore, we suggest that there are selection pressures that select for duet evolution generally (e.g. a sedentary lifestyle) and additional ecological factors that favour particular duet structures or levels of coordination [13,14,16,17].

We found no relationship between duet coordination or consistency versus conspecific abundance or vegetation density. Associations between year-round territoriality and duets have been supported by other studies [17–19], although a global analysis of duet evolution did not find an association with open versus closed habitat [18]. However, it is important to point out that both of these variables were measured relatively coarsely in our analysis compared to the fine scale at which we measured duets. Therefore, we might not have been able to detect biologically relevant patterns owing to the scale at which both of these metrics were calculated. Thus, it would be worth revisiting these hypotheses in a comparative context with improved resolution for these variables or through field studies that directly test these hypotheses in individual species.

To further tease apart existing hypotheses for the evolution of highly coordinated duets, more natural history knowledge is needed for these wrens and other tropical species, such as territory sizes, conspecific densities, pair bond lengths, number of nesting attempts per year, and resource availability. Breeding season length data could come from one or a few locations for each species [45], and, therefore, could be more variable than estimated here. Moreover, our coordination scores are only a snapshot of Neotropical wren behaviour gathered during a narrow window of time and in specific locations. There may be more variability to these behaviours among all these species; for instance, duets may be more or less coordinated during certain times of year, and this warrants future study incorporating seasonal and geographical variation in duet coordination and structure, particularly for species for which we had small sample sizes (e.g. fawn-breasted wren, C. guarayanus). Studies like these will benefit greatly from additional data as we continue to compile natural history data for tropical species from various locations.

(c). Duet structure in Neotropical wrens

Our findings extend the comprehensive descriptions of wren duet form by Mann et al. [7] by quantifying levels of coordination in these same species. Mann et al. [7] demonstrated that members within each genus—Cantorchilus, Thryophilus, Pheugopedius or Thryothorus—tend to sing in a similar style. We found similar patterns in duet coordination levels within genera, with the exception of a few species. For example, three of the four most highly coordinated duets occur within the Cantorchilus genus, while Thryophilus, Pheugopedius and Thryothorus were generally less coordinated, with the exception of P. euophrys. As in Mann et al.'s [7] description of singing styles, the phylogenetic signal we observed for duet coordination was high. Duets by species within the same genus tend to be similarly structured—with Pheugopedius notable for their overlapping duets and Cantorchilus their introductory phrases and alternating duets—which may also be reflected by the high phylogenetic signal. Along these lines, coordination scores were lower in many Pheugopedius species compared to Cantorchilus, which is probably owing to the overlap often seen between the beginnings and ends of phrases within Pheugopedius. Nevertheless, while phylogenetic signal is clearly important, the species with the most highly coordinated duets were not closely related to each other, indicating that precisely coordinated duets evolved multiple times independently within the Neotropical wrens. Ultimately, our study demonstrates that other ecological and natural history factors—especially breeding season length and the large degree of within-pair commitment that must be necessary in these species—are important factors in the evolution of highly coordinated duets.

Data accessibility

Code and data for these analyses are archived through Dryad Digital Repository: https://doi.org/10.5061/dryad.pk0p2ngm6 [72].

Authors' contributions

C.N.T., J.J.P. and K.J.O. conceived of the project. N.I.M. collected and organized the field recordings with help from P.J.B.S. E.L.K. and M.A.M. extracted the acoustic measurements. M.S.-M. and K.G.H. computed the prevalence and vegetation density metrics. K.J.O., M.A.-S. and E.L.K. analysed the data. E.L.K., K.J.O., C.N.T., J.J.P. and M.A.-S. wrote the manuscript and all authors helped edit the manuscript.

Competing interests

We declare we have no competing interests.

Funding

Song recordings were made largely by N.I.M., funded by a grant to P.J.B.S. from the Leverhulme Trust. We thank the Kristen Rupert and John Foote Undergraduate Research Fund at the Cornell Laboratory of Ornithology and the M.J. Murdock Charitable Trust (grant no. 2017312) and Pacific University for research funding. K.J.O. was supported by a US National Science Foundation Postdoctoral Research Fellowship in Biology (grant no. 1612861). K.G.H., K.J.O. and M.A.-S. were supported by Cornell Laboratory of Ornithology Rose Postdoctoral Fellowships.