Pressure for rapid and accurate mate recognition promotes avian‐perceived plumage sexual dichromatism in true thrushes (genus: Turdus)

Abstract Ecological conditions limiting the time to find a compatible mate or increasing the difficulty in doing so likely promote the evolution of traits used for species and mate recognition. In addition to interspecific character displacement signalling species identity, intraspecific traits that signal an individual's sex and breeding status reduce the challenge of identifying a compatible conspecific mate and should be more common in migratory rather than sedentary species, species with shorter breeding seasons and species breeding under high sympatry with many closely related heterospecifics. Here, we tested this recognition hypothesis for promoting plumage sexual dichromatism in the true thrushes (Turdus spp.), a large and diverse genus of passerine birds. We used receptor‐noise limited models of avian vision to quantify avian‐perceived chromatic and achromatic visual contrasts between male and female plumage patches and tested the influence of breeding season length, spatial distribution and sympatry with other Turdus species on plumage dichromatism. As predicted, we found that (1) true thrush species with migratory behaviour have greater plumage sexual dichromatism than non‐migratory species, (2) species with longer breeding seasons have less plumage sexual dichromatism, and (3) greater numbers of Turdus thrush species breeding in sympatry is associated with more plumage sexual dichromatism. These results suggest that social recognition systems, including species and mate recognition, play a prominent role in the evolution of plumage sexual dichromatism in true thrushes.

recognition, and making it swifter, should be more likely to arise and be maintained under conditions that increase both the difficulty of finding a compatible mate and degree of resource competition among conspecifics and closely related species. Conditions likely to favour traits signalling individuals' species, sex and breeding status include higher sympatry with many closely related species with ancestrally shared phenotypes, limited time to find compatible breeding mates during the reproductive season and lower rates of encounter with potential breeding mates due to reduced densities (Andersson, 1994).
In birds, plumage colour can be a visually perceived trait that signals species and (often) sex identity (Bitton & Doucet, 2016;Martin et al., 2015). Plumage sexual dichromatism, or the distinct set of differences in the appearance of male and female feather colours and patterns, is common in birds and is usually attributed to different natural and sexual selection pressures on males and females (Badyaev & Hill, 2003;Burns, 1998;Dale et al., 2015;Dunn et al., 2015;Martin & Badyaev, 1996). Plumage sexual dichromatism results in a visibly perceivable trait useful for recognizing an individual's species, sex and breeding status (e.g. in species with sex-specific delayed plumage maturation, see Hawkins et al., 2012), reducing the time and effort expended to identify a suitable mate (Hamilton, 1961;Saetre & Slagsvold, 1992), both from the perspective of females and males and irrespective whether in the context of mate choice or intrasexual competition.
Evidence in favour of this sex dichromatism-based species recognition hypothesis in birds includes a positive comparative association of greater plumage sexual dichromatism with migratory behaviour and shorter breeding seasons across taxa (Badyaev & Hill, 2003), both of which reduce the amount of time available to search and find suitable mates and successfully breed. Additional support for this recognition hypothesis includes a consistent pattern of greater plumage sexual dichromatism and plumage colour elaboration in avian species that reside on mainland continents and have large geographic ranges in comparison with species that do not migrate, reside on islands and have limited breeding ranges (Badyaev & Ghalambor, 1998;Dale et al., 2015;Doutrelant et al., 2016;Figuerola & Green, 2000;Friedman et al., 2009;Kearns et al., 2020;Matysioková et al., 2017;Simpson et al., 2015;Tobias & Seddon, 2009).
Moreover, plumage sexual dichromatism likely plays a role in hybridization avoidance via reproductive character displacement to facilitate species and mate recognition, especially among closely related species. For example, in Ficedula flycatchers, female choice selects for divergent male plumage colouration, leading to this now well-known example of the male's character displacement across species and between populations of the same species, thus reducing rates of interspecific hybridization (Alatalo et al., 1994;Laaksonen et al., 2015;Saetre et al., 1997).
More broadly and across taxa, however, greater plumage dichromatism is also positively associated with higher breeding sympatry with closely related heterospecifics. For example, among a large sample of passerine sister species pairs, transitions from allopatry to parapatry and increases in geographic range overlaps are positively correlated with greater plumage dichromatism (Cooney et al., 2017). Greater plumage sexual dichromatism has also been found to be positively associated with greater avian species divergence and richness (Cooney et al., 2019;Seddon et al., 2013).
Among passerine sister species pairs, more pronounced changes in male rather than female plumage colouration in sexually dichromatic species suggest that female choice and male-male competition often lead to concurrent increases in sexual dichromatism and speciation events (Seddon et al., 2013). Therefore, plumage sexual dichromatism may be a selected trait, in addition to interspecific character displacement (see above), for facilitating more rapid species and mate recognition when closely related species breed in sympatry (Martin et al., 2010(Martin et al., , 2015. We here test the latter concept in the focal lineage of Turdus thrushes. True thrushes (genus Turdus) are a plumage-wise exceptionally diverse and monophyletic genus of passerine birds consisting of about ~86 species distributed across the globe (Figure 1). These thrushes are an ideal passerine clade for examining the recognition hypothesis for plumage sexual dichromatism because plumage sexual dichromatism and migratory behaviour vary substantially between true thrush species, and sexual dichromatism has evolved multiple times in this genus across their worldwide distribution (Clement & Hathway, 2000;Nagy et al., 2019). Hybridization also occurs in some, but not all, Turdus species, indicating that some sympatric Turdus species can successfully interbreed. A particularly well-documented example of hybridization in true thrushes occurs at large hybrid zone between four Turdus species (T. atrogularis, T. eunomus, T. naumanni, T. ruficollis) in north-central Asia (McCarthy, 2006). Further, plumage sexual dichromatism in true thrushes often coincides with age and breeding status in male thrushes and both plumage and integument traits, including colouration, are used in mate-choice decisions by female thrushes where studied (De La Torre et al., 2020;Faivre et al., 2001;Jarska et al., 2015;Préault et al., 2005;Rowe & Weatherhead, 2011).
Delayed plumage maturation in males is common among true thrushes (Escalona-Segura & Peterson, 1997;Ligon & Hill, 2013;Peterson et al., 2003), where males have 'female-like' plumage colouration during their first breeding season and develop typical breeding adult male plumage for subsequent breeding seasons.
The presence of delayed plumage maturation and distinct juvenal plumage may serve as a signal of a young male's sexual immaturity in order to reduce levels conspecific aggression from older adults (Ligon & Hill, 2013). Delayed plumage maturation prevalence also suggests that female thrushes prefer older males with distinct older adult plumage as breeding mates.
Overall, ecological conditions that increase the time and degree of difficulty in finding a suitable conspecific mate should select for phenotypic traits that reliably signal species and sex identity. Across various bird lineages, greater plumage dichromatism is present in species that are (i) migratory rather than nonmigratory, (ii) have shorter breeding seasons, (iii) live on mainlands rather than islands, (iv) have larger breeding ranges (distributions) and (v) breed in sympatry with more closely related species. These patterns suggest that ecological circumstances where rapid and accurate mate recognition is challenging strongly favour the evolution and maintenance of prominent plumage sexual dichromatism in birds. Here, we test these predictions of the recognition hypothesis for plumage sexual dichromatism by evaluating the potential influences of breeding timing, spacing and sympatry on plumage dichromatism in Turdus thrushes ( Figure 2). We acknowledge that biologically and, thus, statistically, these predictions do not all represent independent life history phenomena; for example, (i) migratory behaviour may be related to (ii) reduced length of the breeding season across Turdus and other avian lineages (García-Peña et al., 2009). Nonetheless, we conduct all these tests separately, even if our overall conclusions are not driven by the full number of these five traits independently.

| Plumage sexual dichromatism
A total of N = 77 Turdus thrush species (approximately ~89% of all known true thrush species) were sampled for plumage spectral reflectance using prepared bird skin specimens at the American Museum of Natural History in New York City and the Field Museum in Chicago, USA. Museum specimens can accurately reflect the extent and variability of live birds' plumage colouration as measured by reflectance spectrometry (e.g. Doucet & Hill, 2009).
We used breeding season month ranges listed in the Thrushes guidebook (Clement & Hathway, 2000) and specimen collection tag dates to choose specimens collected during their approximate breeding seasons. Reflectance measurements spanning 300-700 nm F I G U R E 1 Breeding ranges of all recognized Turdus species from BirdLife International, with representative species' males and females shown for species with plumage sexual dichromatism. The colour scale indicates the number of Turdus thrush species in sympatry with overlapping breeding ranges. Illustrations used with permission from HBW Alive/ Lynx Edicions F I G U R E 2 Hypotheses and predictions for each model (large yellow boxes). Arrow colours indicate predicted correlation, positive (blue) and negative (red) were taken in triplicate from the belly, breast, throat, crown and mantle plumage patches (Andersson & Prager, 2006) of each individual. N = 3 male and N = 3 female individuals were measured for most species (exceptions: T. lawrencii, N = 2 males and N = 2 females; T. swalesi, N = 1 male and N = 1 female). Because of the lack of knowledge which plumage traits are relevant for mate-choice decisions by most Turdus species, we collected data from several patches across the body and treated them equitably with respect to perceptual and statistical analyses (see below). Our approach yielded continuous data to be applied in analyses of sexual dimorphism, similar to some other sex dichromatism studies in birds (e.g. Webb et al., 2016).
In turn, the visible area of these corresponding patches was not quantified in this study between the sexes, even though colourful patch sizes are known to be under sexual selection in several other species (e.g. Qvarnstrom et al., 2006). Similarly, because of the fading of integument tissues with longer museum storage (Kennedy, 2010), we did not quantify colour metrics for beaks and skin patches in our subjects.
Reflectance spectra were measured using a 400 μm fibre optic reflection probe fitted with a rubber stopper to maintain a consistent measuring distance of 3 mm and area of 2 mm2 at a 90° angle to the surface of the feather patch. This approach did not allow us to measure iridescence in plumage colouration, which is not known or documented to be present in the Turdus genus. Measurements were taken using a JAZ spectrometer with a pulsed-xenon light source (Ocean Optics, Dunedin, USA), and we used a diffuse 99% reflectance white standard (Spectralon WS-1-SL, Labsphere).
We applied a receptor-noise limited visual model (Vorobyev & Osorio, 1998) of the European Blackbird (T. merula) visual system (Hart et al., 2000) in the pavo (Maia et al., 2019) package in R v4.0.0 (R Core Team, 2020) to calculate avian-perceived chromatic and achromatic visual contrast (in units of 'Just-Noticeable Differences', or JNDs) of male vs. female plumage patches for all sampled Turdus species. Chromatic and achromatic JNDs were calculated for malefemale pairs within each species (i.e. N = 9 JND values calculated per patch for each species where N = 3 males and N = 3 females sampled), and then JND values were averaged for each species' respective plumage patches. Under ideal laboratory conditions, 1 JND is generally considered to be the discriminable threshold past which an observer is predicted to be able to perceive the two colours as different. However, natural light environments vary both spatially and temporally (Endler, 1993), bringing into question the accuracy of a 1 JND threshold for generalizing visual contrast under natural conditions. Therefore, we calculated the total number of sexually dichromatic plumage patches per species (out of N = 5 measured patches) as the number of plumage patches with average JND values >1, 2 or 3 to account for uncertainty in visual discrimination thresholds due to variation in psychophysical and ambient lighting conditions affecting the strength of between-sex plumage visual contrast (Kemp et al., 2015).
Additionally, we modelled the number of divergent plumage patches (at the three different JND thresholds listed above) within sexes and between different sympatric species under different levels of breeding range overlap (10% increments between 0 and 90%; Figure S1).

| Breeding timing model
We collected data on migration behaviour and breeding season length from Thrushes (Clement & Hathway, 2000) and the Handbook of the Birds of the World (del Hoyo et al., 2017). We assigned three different kinds of migratory behaviour: (1) full migration when a species description clearly stated that a species 'migrates', (2) partial migration when a species was described to have 'altitudinal migration', 'latitudinal migration' or 'movement during non-breeding season' or (3) sedentary when a species was described as 'resident' or 'sedentary'. Breeding season length was defined as the number of consecutive months the species breeds each year.

| Breeding sympatry model
Species' breeding ranges were acquired from BirdLife International (BirdLife International and Handbook of the Birds of the World, 2018). We calculated congener breeding range overlaps (as percentages) using the letsR package in R (Vilela & Villalobos, 2015).
We then calculated the number of sympatric species as the number of congeners with breeding ranges that overlap >30% with the focal species' breeding range (Cooney et al., 2019). Comparisons of the number of sexually dimorphic plumage patches vs. the number of sympatric species among different breeding range overlap thresholds are provided in Figure S2.

| Breeding spacing model
Species' breeding range sizes (in km 2 ) were acquired using the BirdLife International breeding range maps. Species' island vs. mainland residence was also determined using breeding ranges from BirdLife International. Mainland residence was assigned if the species had a breeding range on any continent and Japan. Island residence was assigned to species having a breeding range limited to a noncontinental landmass entirely surrounded by a marine body of water.

| Statistical modelling
We used phylogenetically corrected Bayesian multilevel logistic regression models using the brms v2.13.0 package (Bürkner, 2019) in R v4.0.0 (R Core Team, 2020). We modelled plumage sexual dichromatism responses as the number of sexually dichromatic patches >1, 2 or 3 chromatic and achromatic JNDs. Plumage dichromatism responses were modelled as binomial trials (N = 5 plumage patch 'trials') to test for associations with breeding timing, breeding sympatry and breeding spacing. For all phylogenetically corrected models, we used the Turdus molecular phylogeny from Nylander et al. (2008) to create a covariance matrix of species' phylogenetic relationships. All models used a data set of N = 67 out of the Turdus species for which all the types of data (see above) were available.
Our breeding timing models included the following predictors: zscores of breeding season length (mean-centred by = 5.4 months and scaled by one standard deviation = 2.3 months), migratory behaviour (no migration as the reference category versus partial or full migration) and their interaction. Breeding sympatry models included the number of sympatric species with greater than 30% breeding range overlap as the only predictor of the probability of having a sexually dichromatic plumage patch. Breeding spacing models included log e transformed breeding range size (km 2 ) and breeding landmass (mainland as the reference category versus island). We also ran null models (intercept only) for all responses.
All models' intercepts and response standard deviations were assigned a weakly informative prior (Student T: df = 3, location = 0, scale = 10) (Gelman et al., 2013), and predictor coefficients were assigned flat uninformative priors. We ran each model for 6000 iterations across 6 chains and assessed Markov Chain Monte Carlo (MCMC) convergence using the Gelman-Rubin diagnostic (Rhat) (Gelman et al., 2013). We then performed k-fold cross-validation (Vehtari et al., 2017) to assess each model's accuracy in predicting plumage sexual dichromatism of randomly selected samples of Turdus thrush species, refitting each model K = 16 times. For each k-fold, the training dataset included a randomly selected set of N − N 1 K or N ≈ 63 species, and the testing data set included N 1 K or N ≈ 4 species not included in the training data set. Finally, we compared differences between the models' expected log pointwise predictive densities (ELPD) to assess which model(s) best predicted the probability of having a sexually dichromatic plumage patch. (Vehtari et al., 2017).
Models' predictor effects were assessed using 90% highestdensity intervals of the posterior distributions and probability of direction, the proportion of the posterior distribution that shares the same sign (positive or negative) as the posterior median (Makowski et al., 2019), to provide estimates of the probability of that a predictor has an entirely positive or negative effect on the presence of sexually dimorphic plumage patches. We assume predictor estimates with a probability of direction ≥0.90 to be indicative of a reliable existence of a predictor's effect on sexually dimorphic plumage patches (Makowski et al., 2019).

| Avian visual modelling
Among N = 77 Turdus species, the following proportion have sexually monomorphic plumage (combined achromatic and chromatic JND thresholds): 1.3% (n = 1 species) have no sexually dimorphic patches >1 JND, 44% (n = 34 species) have no dimorphic patches >2 JND, and 63% (n = 49 species) have no dimorphic patches >3 JND (Table S1). Additional proportions of Turdus species with sexually dimorphic achromatic or chromatic plumage patches are available in Table S2. When comparing within sexes between sympatric species (i.e. following Cooney et al., 2017) at least a 30% overlap in breeding ranges: n = 39 species with at least one sympatric species and a median of n = 6 sympatric species per focal species, the median number of avian-discriminable plumage patches between species is 1 or greater for all three achromatic and chromatic JND thresholds except for sympatric females at a chromatic JND threshold >3 ( Figure S1).

| Model comparisons
Breeding sympatry, breeding timing and breeding spacing models performed considerably better than intercept only (null models) in predicting the probability of a species having a sexually dimorphic plumage patch. We obtained N ≥ 4000 effective posterior samples for each model parameter and all models' Markov Chains (MCMC) successfully converged (Rhat = 1 for all models' parameters). All breeding sympatry, breeding timing and breeding spacing models performed similarly well and substantially better than intercept only models in predicting the probability of having a sexually dimorphic plumage patch with achromatic JND values >1, 2, or 3 (Table 1;  Among models predicting the probability of having a sexually dichromatic plumage patch with chromatic JND values >1, 2, or 3, all breeding sympatry, breeding timing and breeding spacing models performed much better than intercept only models, and breeding sympatry models had the top predictive performance (Table 1; breeding sympatry models all have ELPD = 0, only the breeding spacing models predicting dichromatic plumage patches had similar predictive performance).

| Achromatic plumage sexual dichromatism predictors
Migratory behaviour and shorter breeding season lengths were strongly associated with greater odds of a species having achromatic plumage sexual dichromatism. All model predictors' effect estimates are provided as the posterior median odds ratio (OR) and 90% highest-density interval (HDI) in Table 2

| Chromatic plumage sexual dichromatism predictors
Migratory behaviour, shorter breeding season lengths and larger numbers of sympatric Turdus species were strongly associated with greater odds of a species having chromatic plumage sexual dichromatism. Among predictors of breeding timing models predicting chromatic sexually dimorphic plumage patches, longer breeding season length was associated with lower odds of a species having a plum-

| DISCUSS ION
Our results provide comparative correlative evidence in support of predictions of the recognition hypothesis for plumage sexual dichromatism in true thrushes. We used a receptor-noise limited model of Turdus merula vision (Hart et al., 2000;Vorobyev & Osorio, 1998) to measure avian-perceivable visual contrast of plumage colours and found that the odds of plumage sexual dichromatism are much TA B L E 1 Expected log pointwise predictive densities (ELPD) differences and k-fold information criterion values of models (ELPD difference ± standard error [k-fold IC ± standard error])  0.01, 1.42 greater for Turdus thrush species that have full or partial migration rather than no migration, have relatively short breeding seasons and are in sympatry with many other true thrush species (Tables 1 and   2). Our results align with prior comparative studies of avian plumage sexual dichromatism where strong associations of sexual dichromatism with greater migratory behaviour (Dale et al., 2015) and more sympatric taxa (Cooney et al., 2017) were found among many species of different passerine families.

Model
Further, we determined that sympatric Turdus species have distinguishable plumage colouration differences from one another when measuring plumage appearance from the avian visual perspective ( Figure S1).

| CON CLUS IONS
Patterns of plumage sexual dichromatism in true thrushes (Turdus) are consistent with select predictions of the recognition hypothesis for plumage sexual dichromatism. Migratory behaviour and limited breeding seasons reduce the amount of time available to find a mate, and greater plumage sexual dichromatism may help migratory species find compatible mates more rapidly. Greater plumage sexual dichromatism in Turdus species under sympatry with other true thrush species also supports the possibility that increased plumage sexual dichromatism may be the result of reproductive character displacement. Therefore, greater plumage sexual dichromatism likely increases the speed and accuracy of finding a compatible breeding mate, reduces species and mate recognition errors and decreases hybridization.

AUTH O R CO NTR I B UTI O N S
ABL and MEH conceived the study and wrote drafts of the manuscript. ABL completed the study design, data collection, and data analysis. MEH provided funding and advisory support for the study. MEH was supported by a Humboldt Foundation Prize held at the University of Bielefeld, Germany. ABL was supported by the Department of Evolution, Ecology and Behaviour at the University of Illinois.

ACK N OWLED G EM ENTS
We thank the American Museum of Natural History in New York, NY, USA and the Field Museum in Chicago, IL, USA for access to bird skin specimens used in this study. MEH was supported by a Humboldt Foundation Prize held at the University of Bielefeld, Germany.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare.