Cooperative breeding influences the number and type of vocalizations in avian lineages

Abstract

Although communicative complexity is often predicted to correlate with social complexity in animal societies, few studies have employed large-scale comparative analyses to test whether socially complex species have more complex systems of communication. I tested this social complexity hypothesis in birds (Class: Aves) using the large amount of natural history information that describes both vocal repertoire and social system in these species. To do so, I marshalled data from primary and secondary records of avian vocal repertoires (n = 253), and for each of the species in the dataset I recorded the reported repertoire size and associated species information. Using phylogenetic comparative methods, I found that cooperative breeding was a strong and repeatable predictor of vocal repertoire size, while other social variables, e.g. group size and group stability, had little or no influence on repertoire size. Importantly, repertoire sizes expanded concurrently with the evolution of cooperative breeding, suggesting a direct link between these two traits. Cooperatively breeding species devoted significantly more of their repertoire to contact calls and alarm calls. Overall, these results therefore lend support to the hypothesis that social complexity via behavioural coordination leads to increases in vocal complexity.

1. Introduction

The evolution of sociality begins when previously solitary individuals form a cohesive group [1], and considerable research has focused on the ecological and demographic conditions necessary for individuals to form groups and cooperate with group members [2–6]. As these groups form and stabilize, individuals within these groups must communicate to achieve tasks and resolve conflict [7–9]. However, the dynamics of communication systems in social species has received relatively little empirical attention despite the importance of communication within groups (though see [9]). The general hypothesis is that as species become more socially complex, increases in communicative complexity evolve to navigate novel social dilemmas [10]. Importantly, targeted studies have found increases in vocal complexity in species with complex social systems. For example, the social African mole-rat, Fukomys micklemi, has an expanded vocal repertoire relative to the repertoires of other less social subterranean mammals [11], and social mongooses show more variation in how signals are used than less social species [12]. Within humans, a larger population of people speaking a specific language is associated with increased phonological complexity in that language [13]. Bouchet et al. [14] compared three species of monkeys and found that the most socially complex species both vocalized the most and had the largest repertoire. Support for the social complexity hypothesis has also been found in birds, in both single-species [15,16] and interspecific [17,18] studies. Specifically, Krams et al. [18] demonstrated that series of social factors influence the ‘chick-a-dee’ call in parids. Within parids, Freeberg [15] showed that in larger flocks of Carolina chickadees (Poecile carolinensis) individuals include more zero-order and first-order uncertainty (i.e. complexity) in their calls. These previous studies demonstrate that social complexity can lead to changes in vocal complexity in birds, and suggests that the social complexity hypothesis is probably relevant in other species or groups.

Interspecific support for the social complexity hypothesis has been found in other groups. For example, a study on non-human primates found that group size and social interactions increased the size of vocal repertoires [19]. A recent and comparative study on Halictid bees demonstrated that social species increased investment in physiological structures necessary for communication, and that, within species, the cuticular chemical profile differed between solitary individuals and individuals living in groups [9]. Taken together, these results suggest that increased social complexity requires increased complexity of chemical signalling, and that this complexity requires increased investment in sensory structures. Apart from these results in Halictid bees, however, there has been limited comparative evidence showing that communication increases in complexity in social species. For instance, one comparative study on avian species found that the breeding system did not predict vocal complexity [20], though this was possibly owing to limited sampling of species or differences in definition of vocal complexity.

To resolve whether social complexity leads to vocal complexity across birds, I amassed a dataset of adult repertoires from 253 bird species that represent 59 families of birds in the world. Of the 253 species, 97 repertoires were determined using estimates from the primary literature. There have been several attempts to define vocal complexity, often with mixed results [21]. I therefore consider vocal complexity as the ‘functional repertoire size’, as defined by Bradbury & Vehrencamp [22], which is the number of vocalizations that are used in specific behavioural contexts. For example, in species that sing multiple song variants, the song variants are all collapsed into one single vocalization for broadcast song. The remaining 156 species repertoires were estimated using accounts in Birds of North America (BNA) [23] and the Handbook of Australian, New Zealand, and Antarctic Birds (HANZAB) [24]. For these secondary sources the entire repertoire of adult vocalizations was used as the repertoire size. Using two independent sets of data allowed me to identify social variables that consistently affected repertoire size. I also predicted that the composition of the functional repertoire will differ between social and non-social species owing to different selective forces in social environments. Specifically, I predicted that the number of vocalizations associated with sociality (e.g. contact calls that facilitate group cohesion) would expand in social species versus non-social species.

2. Methods

(a). Species sampling: primary literature

I performed a literature search of primary accounts to identify natural history studies that describe the vocal repertoires of avian species. I searched Web of Science© with the search terms ‘vocal repertoire’, ‘vocal repertoires’ and ‘vocalizations’ from 1864 to 2016 and retrieved any articles that assigned vocalizations to specific contexts. I included vocal studies that met the following criteria: (i) assigned vocalizations to specific behavioural contexts; (ii) provided spectrograms for more than 80% of the vocalizations; (iii) recorded vocalizations from at least three individuals. This yielded 97 species from the primary literature.

As the number of vocalizations assigned to behavioural contexts could increase with the effort of the study, three variables associated with effort were recorded from each study: the length of the study, the number of individuals studied (i.e. sample size) and the number of recording hours involved. I only considered vocalizations from adults; I therefore did not count begging calls emitted by nestlings, nor did I count begging calls from juveniles unless it was specified that adults also emitted these vocalizations. I did not count vocalizations only produced because of observer interference (e.g. distress vocalizations during handling). I recorded the vocalization length and any information on the frequency of vocalizations when provided to determine if sociality influenced the acoustic structure of vocalizations. In addition to these measurements, I also categorized vocalizations into one of seven categories: aggression, alarm, contact, flight, vocalizations directed towards a mate, territorial, and vocalizations that did not fit any of these categories were classified as ‘other’. Three other researchers were asked to assess the repertoire size of species from the primary literature to determine inter-observer reliability of assessing repertoire size.

(b). Species sampling: secondary literature

To expand the dataset and validate findings, I marshalled data from two secondary sources: BNA [23] and HANZAB [24]. Both of these sources are curated volumes of species information for their respective geographical locations. In both volumes, individuals compile species’ repertoires, although not all vocalizations are described in detail, and in some cases the behavioural context of vocalizations is not confirmed. I amassed the adult sound repertoire of all cooperatively breeding species from BNA and HANZAB following the list of cooperative breeders listed in Riehl [25]. In HANZAB, there were several congeneric species of cooperative breeders that are listed as cooperative breeders but are not included in the list offered by Riehl [25]; in those cases, I included these species as cooperative breeders if the Handbook of Birds of the World confirmed cooperative breeding. Although there are slightly different definitions of cooperative breeding, the large majority (more than 95%) of cooperatively breeding species in this study meet the requirement of having breeders with helpers, when compared with joint nesting of distinct pairs (e.g. Crotophaga spp.). The expanded the list of cooperative breeding is included in the electronic supplementary material. To facilitate comparisons between cooperative breeders and non-cooperative breeders, I included the vocal repertoire of all congeneric species of cooperatively breeding species in BNA or HANZAB. This resulted in the addition of 156 species' repertoires from the two handbooks.

All comparative analyses were performed on the set of species from the primary literature, the set of species from the handbooks and a third time on the combined set of species. The results were then compared to determine if the findings were consistent across datasets. In the combined analysis the source of the species' account (BNA, HANZAB, primary literature) was added as a fixed factor in the analyses to control for differences in repertoire size.

(c). Phylogeny and comparative analysis

I acquired phylogenies (Ericson backbone) for included species from the ‘Global Phylogeny of Birds’ database (birdtree.org) published by Jetz et al. [26]. These phylogenies are supported by several backbones derived from genetic data, but in certain families and genera the distal placements of species are not fully resolved. I therefore downloaded 1000 trees that represent combinations of distal taxa placements and ran the comparative analysis (see the description of statistical models below) on each of these trees. This allowed me to determine whether the results are robust to slight changes in the topology between trees. Repertoire size was skewed; I therefore square-root transformed repertoire size to achieve normality (electronic supplementary material, figure S1). To assess the effects of group size and categorical variables I built phylogenetic generalized least-squares (PGLS) models for each phylogeny. I controlled for phylogeny by estimating Pagel's λ for each phylogenetic tree and used this estimate as a correlational structure in the model [27]. Estimation of Pagel's λ was conducted using the package ‘ape’ [28]. Phylogenies were plotted with the package diversitree [29]. PGLS models were performed using the package nlme [30]. All analyses were carried out in R v. 3.2.3 [31].

(d). Analysis of evolutionary order

For the analysis that examined whether large repertoires were concordant with the evolution of certain categorical traits, I developed custom scripts that employ functions within phytools to estimate ancestral values and then extract relevant ancestral values of categorical and repertoire traits [32]. Specifically, I isolated branches in the phylogenetic trees where an ancestral node was estimated as a solitary breeder, and the descendent node was a cooperative breeder. These branches were not necessarily just branches leading to extant tips on the tree, but could be internal branches. It was on these branches that cooperative breeding was estimated as evolving from a solitary ancestor. Using these branches on the tree, I determined how repertoire size changes when there is a change in the type of cooperative breeding (methods similar to [4]). To estimate how repertoire size changes with the evolution of cooperative breeding, I used the branches that were isolated in the previous ancestral state estimation of cooperative breeding. Using these branches where cooperative breeding evolves, I then estimate how repertoire size changes from a solitary ancestor to a cooperative breeding descendant. However, the causality may be reversed, where ancestors with large repertoires may facilitate the evolution of cooperative breeding. To test the idea that ancestral species with larger repertoires more often give rise to cooperative breeders, I used methods developed by Blackmon et al. [33]. Briefly, these methods isolate nodes within the phylogeny that generate descendent nodes with a specific value of a categorical variable. In this analysis, the repertoire size of ancestral nodes that give rise to cooperative breeding is compared to the repertoire size of a random selection of nodes within the phylogeny. This analysis therefore determines whether nodes that generate cooperatively breeding descendent species have larger repertoires than would be expected if randomly sampling nodes from the phylogeny.

(e). Sociality variables

Bergman & Beehner [34] argue that social complexity should be quantified by counting the number of different types of individuals that a focal individual will interact with; unfortunately, this definition requires highly detailed information that is not readily available for most species. I therefore recorded several variables that represent distinct aspects of sociality. First, I recorded the maximum group size individuals may experience during the course of a year, as this measure was typically more available than the average group size. There are several pelagic species in the dataset that occur in large breeding colonies, though massive colony size (approx. 30 000 breeding pairs in thick-billed murre, Uria lomvia) was not considered a ‘group size’ in this analysis because individuals in these colonies typically function as independent monogamous pairs rather than cohesive groups [35]. In this research, a group is considered a set of individuals (more than a breeding pair) that interact and are stable through space and time (though not necessarily permanent). Second, I recorded the type of cooperative breeding following the list in Riehl [25]: kin cooperative breeders, non-kin cooperative breeders, cooperative breeders of unknown status or non-cooperative breeders. Third, I recorded whether species have stable relationships versus species where individuals do not interact with the same conspecifics. These classifications were compiled from the primary literature (see the electronic supplementary material), Handbook of Birds of the World [36], and a recent study that also looked at the stability of social bonds [37]. For the latter, I combined variables from Tobias et al. [37] to estimate the cohesiveness of individuals within species. Specifically, species that were classified as ‘long-term social bonds’ and ‘seasonal territoriality’ (e.g. many seabirds) were classified as ‘moderate cohesiveness’. Any species labelled as ‘solitary’ was classified as ‘minimal cohesiveness'. Species that maintained ‘long-term social bonds’ and ‘year-round territoriality’ were therefore classified as having ‘high cohesiveness’. Group size measures were taken from the published literature and often not from the same publication that reported the repertoire size (electronic supplementary material). In certain species, a range of group sizes was reported and the averages of the minimum and maximum were taken. In one instance (Amazona aestiva), group size was reported as a ‘few individuals’ and I therefore assigned this species a group size of 3 as this was the most conservative estimate. Given that this is an imprecise measure of group size in this species, I repeat the analysis where A. aestiva group size is set to 30, and 300. This did not qualitatively change the results regarding group size.

(f). Other variables

Along with the variables associated with sociality, I classified species into one of several habitat types (forest, woodland, grassland, wetland and pelagic). Species' breeding system, both presence and type of dimorphism (size versus colour), and species' body mass were recorded for analyses.

3. Results

(a). Species database and functional repertoire size

The range of functional repertoire size across the 97 species from the primary literature was 2–25 vocalization types (median = 9.0, mean = 9.64, standard deviation = 4.64; figure 1), and the range of 156 species from the secondary literature had a range of 1–22 vocalization types (median = 6.0, mean = 6.91, standard deviation = 3.64; electronic supplementary material, figure S2). Within the studies of repertoire size from the primary literature, there was no evidence that study effort influenced functional repertoire size (p > 0.25 in all linear models; electronic supplementary material). Inter-observer reliability of repertoire size was high as the adjusted R² of the linear relationship among my assessment and the assessment of three others was 0.96. The slope of this relationship was not significantly different from 1 (β = 1.07, p = 0.88; electronic supplementary material, figure S3). All PGLS analyses were conducted on a set of 1000 trees from the Jetz et al. [26] phylogeny and the median test statistics from the 1000 analyses are reported for analysis. There was a significant difference in estimated repertoire size between estimates from the primary literature and estimates from secondary sources (median F_2,250 = 8.57, median p < 0.001, all p-values < 0.001), with BNA (n = 42, mean = 8.02, standard deviation = 4.12) and HANZAB (n = 114, mean = 6.50, standard deviation = 3.38) having significantly lower repertoire sizes than reports from the primary literature (n = 97, mean = 9.64, standard deviation = 4.64). In any following analysis that combines all repertoire sizes the effect of source is controlled for by adding source as a fixed effect.

Figure 1.

Phylogeny of the species from the primary literature with cooperative breeding and repertoire size on the tips of the phylogeny. The interior circle represents different tiers of repertoire size. Darker colours in the interior circle represent larger repertoire size (see legend for specific tiers). The outer circle represents whether species are a cooperative breeder or not (red represents cooperative breeding, while white represents solitary breeding). Species names are listed on the exterior of the phylogeny.

(b). Effect of habitat and other variables

Using PGLS, and after controlling for the literature source, I found no difference in functional repertoire size between passerines and non-passerines (median F_2,248 = 0.70, median p-value = 0.52). There was no significant effect of sexual dimorphism (either size or plumage) on functional repertoire size (median F_1,249 = 1.54, median p = 0.22). PGLS was used to determine whether there was an effect of habitat, and after controlling for repertoire source and cooperative breeding, there was a significant effect of habitat on functional repertoire size (median F_4,246 = 4.07, median p = 0.003, range of p-values: 0.002–0.005). Specifically, species that live in grasslands (mean = 5.86) and pelagic species (mean = 4.86) had significantly lower repertoire sizes than species living in forests (mean = 8.35), wetlands (mean = 7.62) or woodlands (mean = 8.37).

(c). Effect of social variables

Neither group size (median F_1,246 = 2.41, median p-value = 0.12) nor cohesiveness of social bonds (median F_2,245 = 2.29, median p-value = 0.13) significantly influenced repertoire size. However, cooperative breeding had a positive effect on repertoire size (median F_1,246 = 13.5, median p-value < 0.001, range of p-values: < 0.001–0.001; figure 2a) in the combined analysis. This overall effect of cooperative breeding was not driven by one of the data sources as cooperative breeding was associated with larger repertoires in both the primary literature (median F_1,96 = 5.11, median p-value = 0.02, range of p-values: 0.001–0.05) and the repertoires acquired from the handbooks (median F_1,155 = 11.9, median p-value < 0.001, range of p-values: < 0.001–0.001). Within cooperative breeders, there was no effect of group size on repertoire size (median F-value = 1.02, median p-value = 0.32).

Figure 2.

(a) Mean repertoire size and standard error of the mean (s.e.m.) for repertoire sizes in cooperative breeders derived from the primary literature. (b) Mean repertoire size and standard error of the mean (s.e.m.) for repertoire sizes in cooperative breeders derived from the secondary literature. Cooperative breeders have significantly larger functional repertoire sizes than species without cooperative breeding in both analyses. (c) Transition analysis of repertoire size from the primary literature showing that transitions to cooperative breeding are associated with an increase in repertoire size. (d) Transition analysis of repertoire size from the secondary literature showing that transitions to cooperative breeding are associated with a significant increase in repertoire size. Numbers above the bars in (c) and (d) are the sample size for each group, i.e. the number of transitions in each phylogenetic analysis. Error bars represent s.e.m. around the mean for panels (c) and (d).

(d). Evolutionary order of cooperative breeding and large repertoire size

Not only is cooperative breeding associated with larger repertoire sizes, but the evolution of cooperative breeding is also directly associated with increases in repertoire size. For the repertoires from the handbooks, transitions to cooperative breeding are associated with a significant increase in repertoire size (median t₄₁ = 2.95, median p-value = 0.005, range of p-values: 0.001–0.12). The median Pagel's λ for this set of species was 0.31 (Pagel's λ median p-value = 0.002), suggesting phylogenetic signal among species in repertoire size. Although transitions to cooperative breeding were followed by an increase in repertoire size in species from the primary literature, this increase was not significant (median t₁₁ = 1.33, median p-value = 0.21). To determine if large repertoires facilitates the evolution of cooperative breeding, I employed methods developed by Blackmon et al. [33] to determine if nodes that generate cooperatively breeding species had larger repertoires than expected, given the other nodes within the phylogeny. Neither the primary literature dataset (p = 0.083) nor the secondary literature dataset suggest that large repertoires precede the evolution of cooperative breeding (p = 0.31; electronic supplementary material, figure S9).

(e). Repertoire composition

To determine how repertoires expand in cooperatively breeding species, the vocalizations of each species from the primary literature were classified into functional categories for analysis. Cooperatively breeding species had significantly more alarm calls (PGLS; median F_1,84 = 4.10, median p = 0.04, range of p-values = 0.02–0.05) and contact calls (median F_1,84 = 12.7, median p = < 0.001, range of p-values all < 0.001). Cooperatively breeding species on average had 0.90 ± 0.43 more alarm calls and 1.42 ± 0.43 more contact calls than non-cooperatively breeding species (figure 3a). Cooperatively breeding species devoted a significantly larger proportion of their repertoire to contact calls (PGLS; median F_1,84 = 4.02, median p = 0.04, range of p-values = 0.03–0.06). There was a non-significant trend where cooperatively breeding species had a smaller proportion of calls associated with aggression (median p = 0.08). Social variables did not influence the vocalization length nor the range of vocalization lengths (p > 0.05, F < 3.50 for both comparisons); however, there was a non-significant trend (F_1,84 = 3.41, p = 0.07) whereby species in larger groups had shorter vocalizations.

Figure 3.

(a) Comparison of the total number of calls devoted to each category across studies. The vertical axis represents the number of calls in each of the categories in cooperative breeders versus non-cooperative breeders. After controlling for phylogeny, cooperatively breeding species have significantly more vocalizations associated with alarm calling (p < 0.05) and contact calling (p < 0.001). Error bars represent s.e.m. around the mean. (b) Comparison of repertoire proportions across several categories. The vertical axis represents the proportion of the repertoire dedicated to each of the categories in cooperative breeders versus non-cooperative breeders.

4. Discussion

Cooperative breeding is considered to entail an increase in both social and behavioural complexity; consequently, increased social complexity is theorized to drive an increased communicative repertoire [34]. Consistent with the social complexity hypothesis, in this study I found that species which breed cooperatively have significantly larger functional repertoires than species that are not cooperative breeders (figure 2a). In the analysis of repertoire sizes from the secondary literature, the evolution of cooperative breeding is concordant with the evolution of significantly larger repertoire sizes in general (figure 2b). Although the effect of cooperative breeding was similar in the primary literature, the increase in repertoire size associated with the evolution of cooperative breeding was not significantly different from zero. However, this may be owing to sample size, as in both datasets the effect of evolving cooperative breeding entails an increase in repertoire size of approximately one vocalization (figure 2b,d). I also found no evidence that larger repertoires drive the evolution of cooperative breeding (electronic supplementary material, figure S9). In summary, I found that an increase in social complexity via cooperative breeding is concomitant with an increase in vocal complexity.

How do vocal repertoires expand? Cooperatively breeding species have significantly more contact calls (approx. 1.4 more contact calls; median p < 0.001) and alarm calls (approx. 0.9 more alarm calls; median p = 0.04) than species that do not breed cooperatively (figure 3a). The functional repertoires of cooperative breeders therefore increase largely through the addition of contact calls and potentially alarm calls (possibly at the expense of aggressive signals, p = 0.08). As contact calls facilitate group cohesion [38,39], it is unsurprising that this type of vocalization would expand in cooperative breeders. For instance, the addition of contact calls by cooperative breeders could be owing to the increased necessity of group cohesion while performing daily behaviours, e.g. foraging [40]. There is also considerable evidence that vocal complexity can increase in social species owing to increased complexity in alarm calling. A study on dwarf mongooses (Helogale parvula) provided early evidence that sociality could influence vocal complexity as individuals in these groups encode information about predator type, distance and relative danger in alarm calls [41]. Similar results on alarm calls have been found in birds. For instance, both Carolina (Poecile carolinensis) chickadees and black-capped chickadees (Poecile atricapillus) encode information about predator size, and therefore the level of threat, by increasingly adding a specific note to the end of a vocalization [42,43]. However, these chickadee species are not cooperative breeders, and there are seemingly more cases where cooperative breeding species of birds employ more complex alarm calling. For instance, a study on Siberian jays (Periosoreus infaustus) showed that individuals employ different alarm calls depending on the location of the predator [44]. In the pale-winged trumpeter (Psophia leucoptera) Seddon et al. [45] found that alarm calls vary depending on whether they are in response to aerial or terrestrial predators. In addition, studies on other cooperatively breeding birds and mammals have demonstrated that alarm calls can be predator- or distance-dependent [46–49]. Predation of a member of a cooperatively breeding group may entail large fitness costs for other group members. It is therefore possible that selection has increased the number of alarm calls within the functional repertoire of cooperatively breeding species (figure 3). This may also explain why repertoire size expands in cooperative breeding species rather than with group size or the stability of social bonds. Given the considerable threat predation poses, it is possible that predation selects for both social complexity [50] and subsequent modification of alarm calls [51].

Although the results presented here provide considerable evidence for the social complexity hypothesis, not all studies find consistent effects of sociality on vocal complexity. For example, one study found that song and syllable repertoire size in several species of birds was not associated with breeding variables such as extra-pair paternity and mating system [20]. Additionally, a recent comparative study of new world jays found that cooperative breeding was associated with reduced repertoire size [52]. One potential explanation for the discrepancy in results is that cooperative breeding may have clade-specific effects, and the evolution of functional repertoire size may change depending on the group of cooperative breeding species being analysed. For instance, the filial relationships (i.e. kin versus non-kin cooperative breeders) in cooperative breeders could influence functional repertoire size. Specifically, alarm calls may be more likely to evolve and expand as a component of repertoire in kin groups [53]. Additionally, repertoire size may decrease, while information in certain signals increase [54,55]. For instance, signals may diversify to include information about individual identity, sex or breeding status [56].

This study also detected an effect of a general habitat variable on repertoire size, where pelagic species and species living in grasslands had significantly lower average repertoire sizes than species in forests, woodlands or wetlands (median p = 0.003, electronic supplementary material, figure S7). Although these are general habitat classifications, it is noticeable that the two habitats which are the most open (pelagic and grasslands) have the lowest average repertoire sizes. One potential explanation is that in open habitats visual signals are more useful and species in these habitats are therefore less likely to evolve vocal signals. Further work that investigates repertoire size across differing ecologies would illuminate the influence of habitat on vocal repertoires.

One important aspect of this study was the specific definition of repertoire size that was employed. Although I employ a definition of repertoire size that estimates the number of vocalizations given within specific social contexts, this contrasts with other definitions of repertoire size, e.g. syllable repertoire size or song-type repertoire size, which may be larger in solitary species. For instance, the famously large sound repertoire of mimics, e.g. Mimus polyglottos, have considerably more notes and motifs than any repertoire found in this study [57]. In addition to mimics, solitary species often have larger song repertoires than the functional repertoires; specifically, white-eared ground sparrows (Melozone leucotis) have up to 40 male song types in a single population [58], which is larger than any functional repertoire reported here. Additionally, an interspecific study on North American wrens showed that polygynous wren species had larger song repertoires than non-polygynous species [17]. Sexual selection therefore can successfully drive increases in the overall number of sounds produced by birds [59,60]. While the overall number of sounds produced by a bird is influenced by numerous selective pressures, the results presented here suggest that the number of functionally different vocalizations is associated with sociality, and specifically cooperative breeding.

Despite considerable variation in complexity of bird vocalizations across species, the variation has often been difficult to explain [61]. The results presented here suggest that cooperative breeding imposes selection for an increased number of context-dependent vocalizations. As signals may diversify with increased social complexity, the major modality of communication in other groups, e.g. chemical signalling in eusocial insects, should be tested for similar responses in these other socially complex groups. Indeed, a recent study on Halictid bees found increased communicative complexity with increased social complexity [9]. Paired with the results here, there is now strong comparative evidence from eusocial insects and birds that sociality influences the complexity of communication systems. In summary, these results therefore provide considerable comparative support for the hypothesis that complex sociality selects for increased communicative complexity.

Data accessibility

All data used in this project will be deposited into dryad at the time of publication (http://dx.doi.org/10.5061/dryad.53924) [62].

Competing interests

The author declares no competing of interest.

Funding

This work was funded by a National Science Foundation Postdoctoral Fellowship in Biology awarded to G.M.L. (no. 1523748).