Rômulo Mendonça Machado Carleial

Transcription

1 Universidade Federal de Minas Gerais Programa de Pós-Graduação em Ecologia, Conservação e Manejo da Vida Silvestre Rômulo Mendonça Machado Carleial Evolution of plumage coloration and sexual dichromatism in the family Pittidae (Passeriformes:Tyranni) Belo Horizonte Minas Gerais Brasil 2015

2 Rômulo Mendonça Machado Carleial Evolution of plumage coloration and sexual dichromatism in the family Pittidae (Passeriformes:Tyranni) Trabalho de dissertação apresentado ao Programa de Pós-Graduação em Ecologia, Conservação e Manejo da Vida Silvestre da Universidade Federal de Minas Gerais como requisito para a obtenção do título de Mestre. Orientador: Prof. Adriano Paglia Coorientadora: Prof.ª Marina Anciães Coorientador: Dr. Richard Prum Belo Horizonte Minas Gerais - Brasil 2015

3 Abstract The role of different selective pressures underpinning plumage evolution has occupied biologists for the past century. However, only recently have scientists incorporated avian visual models into the study of plumage colors. Pittas are very colorful passerines and offer a great opportunity to investigate the mechanisms behind plumage diversification and intensity in sexual dichomatism. We used a tetrahedral color space model to quantify plumage coloration and a color distance model to access the degree of sexual dichromatism in this family. We tested whether plumage metrics are influenced by ecological variables using Ornstein-Uhlenbeck models with different selective regimes. We also performed an ancestral reconstruction analysis to investigate the evolution of sexual dichromatism in the family. We found that plumage diversity in this group is better explained under a Brownian motion model but that migratory behavior has affected plumage evolution in the genus Pitta. Sexual dichromatism is the plesimorphic condition in pittas and more wider spread than previously reported. Ultra violet reflectance is more pronounced in males of sexually dichromatic species, although conspicuity has increased in both sexes. These findings suggest that reciprocal sexual selection has likely taken place in this group, and that evolution of plumage colors in pittas is complex and has involved more than one evolutionary mechanism. Key words: sexual selection, ultraviolet, tetrahedral color space, avian vision, pittas, migratory Introduction The importance of natural and sexual selection on the evolution of plumage coloration and sexual dichromatism has been part of intense debates among evolutionary biologists (Darwin 1871; Wallace 1889; Anderson 1994). Sexual dichromatism the difference in coloration between males and females of the same species - occurs when both males and females experience different selective pressures, in many cases because of differences in parental investment (Cunningham and Birkhead 1998; Kimball and Ligon 1999; Amundsen 2000; Badyaev and Hill 2003). Sexual dichromatism has been more strongly associated with sexual selection on males for increased plumage elaboration, presumably because plumage ornaments are either attractive to females (intersexual selection) or increase male competitiveness against rivals (intra-sexual selection) (Darwin 1871; Anderson 1994). However, many ecological variables have been invoked to explain variation in sexual dichromatism: increase in predation risk at the nest has been associated with loss in female conspicuity (e.g.wallace 1868; Soler and Moreno 2012); colder micro-climate and spatial separation of food resources in high altitudes has been associated with increase in male care and reduction in sexual dichromatism (Badyaev, 1997a; Badyaev and Ghalambor, 2001); migratory behavior is widely reported to affect sexual dichromatism, but with mixed results (e.g. Badyaev and Hill 2003; Friedman et al 2009; Dunn et al 2015); to cite a few.

4 With an increase of studies evaluating plumage evolution, we are now aware that shifts towards sexual dichromatism or monochromatism can occur by either loss or gain in plumage ornamentation in both sexes (e.g. Burns 1998; Amundsen 2000; Siefferman and Hill 2005; Hoffman et al 2008, Dunn et al 2015). It is important then, to analyze male and female plumage separately (Badyaev and Hill 2003). Many of the earlier studies on plumage coloration failed to fully evaluate the extent of the complex interactions mentioned above, because they relied on methods that were based solely on human perspective (reviewed in Bennet et al 1994). However, avian visual system is very different from ours in many ways. Birds have tetrachromatic vision that is sensitive to ultraviolet and violet light (Chen and Goldsmith 1986; Hart 2001) and oil droplets in their cone receptors that narrows the absorption spectra, which is believed to reduce the overlap between pigments, increasing the number of hues they can discern (Govardovskii 1983; Vorobyev et al 1998; Vorobyev 2003). Birds are also believed to process achromatic signals independently (Vorobyev and Osorio 1998; Kelber et al. 2003; Jones and Osorio 2004; Endler and Mielke 2005). It is of summary relevance then, that birds signaling should be analyzed under an avian visual model. The tetrahedral color space model (Goldsmith 1990; Endler and Mielke 2005; Endler et al. 2005; Stoddard and Prum 2008) was developed in order to fulfill this predicament, by incorporating avian cone type sensitivities into the equivalent of a human chromaticity space. Each vertex of the tetrahedral represents one of the four birds retinal cone types namely uv or violet sensitive (uv/v), short-wavelengthsensitive or blue (s), medium-wavelength-sensitive or green (m), and long-wavelength-sensitive or red (l). Each color has a unique set of relative stimulation values for these cones and therefore a unique position in the avian color space. In this study we examine plumage evolution in the family Pittidae. Pittas are passerines of the suborder Tyranni, mainly distributed in tropical Australasian and Asiatic forests, with two species occurring in tropical Africa. They are currently divided in three genera (Pitta, Erythropitta and Hydrornis) with around 30 species. However, new studies have been raising some subspecies to species level, which would increase this number to more than 40 (Rheindt and Eaton 2010; Irestedt et al 2013). Pittas are small to medium size, usually terrestrial, birds with its diet being constituted mainly by worms and other invertebrates (Erritzoe 2003; Chowdhury et al 2013). The species altitudinal distribution ranges from lowlands to up to 2200m, with altitudinal shift believed to occur in some species (Erritzoe 2003). In all species investigated so far, the mating system is monogamy with bi-parental care, and both sexual

5 dichromatism and monochromatism are present in the family. Migratory behavior occurs in some species of the genus Pitta, although this number may be underestimated, since pittas movements are poorly understood (Erritzoe 2003). It has been suggested that migrant species in the genus Pitta are duller than their sedentary conspecifics, because either migratory behavior constrained plumage evolution or genetic drift in small, sedentary populations promoted plumage diversification (Irestedt et al 2006). However, this hypothesis has never been objectively tested. Pittas are astonishingly colorful birds (Figure 1), which makes them a promising group for the study of plumage evolution. Although they are known as jewel thrushes and are popular among bird watchers (Erritozoe 2003), little is known about their ecology and behavior (but see Donald et al 2009). It remains to be investigated whether the great color diversity in this group is a product of adaptive evolution to different ecological conditions or if it has been shaped by other evolutionary mechanisms, such as sexual selection. Therefore, our objectives in this study were to: (1) describe male and female pittas plumage coloration under an avian visual model; (2) quantify sexual dichromatism for each species; (3) test several existing hypotheses regarding evolution of plumage coloration and sexual dichromatism in relation to ecological variables; (4) investigate the correlation between overall plumage contrast and degree of sexual dichromatism and (5) investigate differences in plumage conspicousness between males and females. We hypothesized that birds occupying higher elevations would be duller than birds which are lowland specialists, because of scarcity of resources at higher elevations (Badyaev 1997a; Badyaev and Ghalambor 2001). For an instance, certain plumage hues can be environmentally constrained, such as the majority of carotenoid-based colors, since its metabolic precursors can be only obtained through diet (reviewed in Hill and McGraw 2006b). We also expect ground nesting birds to be more cryptic than birds that preferentially nest on trees, because predation risk is likely to be higher at ground level (e.g.wallace 1868; Soler and Moreno 2012). It is also likely that species inhabiting islands should have less color diversity because islands tend to hold smaller populations than the mainland and small populations are less able to respond to changes in conditions, due to its smaller genetic variation (Kimura and Ohta 1969). Finally, we also expect that overall plumage contrast would increase with sexual dichromatism, because if the intensity of the latter is a good estimator of sexual selection, it would be logical to assume that selection would favor conspicuousness if it helps males attract females. A strong correlation between males and females color span would suggest that plumage coloration is somewhat driven by similar selective pressures in both sexes.

6 Since all pittas are mainly found in forests within the tropical range, we are already controlling for habitat type and latitude, two variables that have been reported to strongly affect plumage evolution and sexual dichromatism (e.g. Badyaev and Hill 2003; Gomez and Théry 2004; Gomez and Théry 2007; Shultz and Burns 2013). a) b) c) d) e) f) Figure 1 Some pitta species and their exuberant plumage. a) Blue-winged pitta Pitta moluccensis; b) Male banded pitta Hydrornis guajana; c) Graceful pitta Erythropitta venusta; d) Blue-rumped pitta Hydrornis soror; e) Rainbow pitta Pitta iris f) Male blue pitta Hydrornis cyanea.

7 Methods Color measurements We measured the reflectance spectra of plumage patches of Pittidae museum skins from the Yale Peabody Museum and American Museum of Natural History. Whenever the sample allowed, we used three males and three females of the 30 pitta species listed in Erritzoe (2003). Because plumage coloration in museum skins can change with time, we chose the best-looking skins from birds of the same subspecies and locality whenever it was possible. We took plumage measurements using an S2000 Ocean Optics spectrometer with an Ocean Optics DH BAL deuterium halogen light source (Ocean Optics, Dunedin, FL) held in an aluminum block to minimize incident light. We establish six standard patches (throat, chest, belly, crown, back and wing) which were measured at constant distance with the block held perpendicular to the plumage. Additional plumage patches were measured if the species had colors which were distinct to the human eye within a patch greater than 4mm 2 (the resolution limit of the spectrometer). We took three measurements per patch per individual to guarantee repeatability, which were then averaged to create the individual s spectra. Since we were interested in investigating the evolution of plumage coloration at the species level, we used the averaged spectra of males and females per species. In order to objectively quantify plumage coloration, we used the tetrahedral color space model based on avian-cone type sensitivities (Goldsmith 1990; Endler and Mielke 2005; Stoddard and Prum 2008). The averaged spectra of males and females of each species were plotted in the tetrahedral using the program TetraColorSpace (Stoddard and Prum 2008) implemented in MATLAB 7 (MathWorks, Natick, MA) and the package pavo (Maia et al 2013) implemented in the software R (R Core Development Team 2015). Since pittas inhabit forest, usually with dense understory (Erritzoe 2003), we modelled avian plumage colors using both the ideal (homogeneous illuminance of 1 accross wavelengths) and forest shade illuminance implemented in pavo. We did this in order to compare the differences in plumage conspicuity under both methods, and evaluate how much information is lost or misinterpret by using an artificial illumination model. Many color metrics can be calculated when using the tetrahedral color space. For analyses on plumage evolution and sexual dichromatism we used the following metrics: average color span, which is a measure of overall contrast among plumage patches and it is calculated based on the Euclidean distance between color points; average chroma, which is a measure of overall plumage saturation and is the

8 distance of the color point from the achromatic origin; color volume, which is a measure of color diversity and it represents the minimum convex polygon containing all plumage colors; average phi, which accounts for the ultra-violet contribution to the overall plumage coloration. We also used average brilliance, which is calculated independently from the tetrahedral and quantifies overall plumage brightness. We decided that it would be informative to investigate the extent in which pittas are colorful compared to other birds. To do so, we compared pittas color variables with those from Stoddard and Prum (2011), in which they estimated the avian color plumage gamut by sampling 965 plumage patches from 111 species of birds. Fortunately, pittas were not part of their sample and their study species were chosen in order to comprehensible sample the variety of colors produced by birds. Therefore, a direct comparison between their results with ours can provide us with an estimation of pittas color diversity. Since we lack description for the majority of color production mechanisms in pitta, the mechanisms quoted in this study are assumptions based on the literature. General color patterns are usually attributed to specific mechanisms namely red, orange and yellows (carotenoids), blue, violet and iridescent colors (structural) black, brown and buff colors (melanin) and greens (combination of both carotenoid-based and structural colors) (Reviewed in Hill and McGraw 2006a). The reader should keep in mind that this is a rough generalization, and there are many complex colors that are not covered in this summary since determining mechanisms behind color production was out of the scope of this study. Plumage evolution and sexual dichromatism In order to quantify sexual dichromatism we compared color contrast between homologous patches of males and females. We performed pairwise analyses using the just-noticible-differences (jnd) model developed by Vorobyev and Osorio (1998), which calculates the distance between pairs of patches in the avian color space using receptor noise, based on the relative abundance of cone receptors in the retina. The model assumes that for a given pair of colors the threshold for discrimination is 1 jnd, with values below this threshold being indistinguishable for birds. However, although this threshold has been supported by behavioral data (Vorobyev et al 1998), other studies have stated that values below 3 jnd are difficult to distinguish (Siddiqi et al 2004; Cassey et al 2009). All these studies assumed a Weber s fraction of 0.05 which is inconsistent with more recent studies, which have assumed a more conservative value (w=0.1), which effectively reduces jnd results by half (Lind et al 2013; Lind et al 2014; Olsson et al 2015). Because of such different approaches, we analyzed sexual dichromatism considering all the

9 above methodologies and compared the results with previous interpretations of sexual dichromatism based solely on human observation. To better visualize differences between males and females under the avian visual system, we overlapped the male and female volume in the avian color space. A small overlap would indicate that different hues contribute to the overall plumage coloration of females and males, while a big overlap would indicate that similar hues are being exploited by both sexes. We performed comparative analyses of plumage metrics using the 50% majority rule consensus tree from Irestedt et al (2006), which is currently the most comprehensible phylogeny of the Pittidae. We investigated whether regularly reported ecological and behavioral variables affected the evolution of plumage coloration in males and females pittas and if it also affected the degree of sexual dichromatism among species. We did so, by fitting Ornstein-Uhlenbeck (OU) models of evolution (Hansen 1997; Butler and King 2004) in plumage metrics and mean plumage jnd. OU models simulate stabilizing selection by assuming different selective optima along the phylogeny, which are defined by painting the tree branches with the correspondent selective regime (Figure 2). We compared an OU model with three selective regimes based on altitudinal distribution ( lowland, mid-montane and uppermontane ), two selective regimes reflecting nesting behavior ( ground and tree ) and two selective regimes based on insular or continental distribution ( island and continent ) (Figure 2). We analyzed regimes based on migratory behavior ( migratory and resident ) separately for the genus Pitta only (Figure 2), because Irestedt et al (2006) hypothesis contemplated only this genus. We defined the regimes based on published information available for pittas. For species that are flexible in their behavior or habitat use e.g. nest both on trees or on the ground we chose the one in which they are more associated with. All OU models were ranked against a pure Brownian motion model (Felsenstein, 1985) and an OU model with a single optimum (Butler and King 2004), using the AIC criterion (Akaike 1973) corrected for small sample sizes (Sugiura 1978). BM is a pure drift model whereas a OU with a single optimum simulates constant, stabilizing selection, along the whole phylogeny. We evaluated model fitting using the ΔAIC criteria in which a difference of less than 2 represents no differences of fit between models, a difference of 2 indicating difference of fit, a difference of 4 indicating considerable difference of fit and a difference greater than 10 strong preference for one model over the other. Preference for the BM or OU model with single optimum would indicate that plumage coloration

10 is unconstrained by the chosen selective regimes. We also quantified the intensity of phylogenetic signal on plumage metrics using the Pagel s λ method (Pagel 1999) to evaluate how much pittas phylogenetic affinities can predict plumage metrics. Figure 2 Selective regimes used in Ornstein-Uhlenbeck models analyses. Each color represents a selective regime fitted to the phylogeny, and the species names are abbreviated at the terminal nodes. Models from left to right: Single optimum; elevational gradient; nesting behavior; geographic distribution; migratory behavior. We investigated the evolution of sexual dichromatism by performing an ancestral reconstruction analysis of the mean plumage jnd along the Pittidae phylogenetic tree, using the square-change parsimony method. Since sexual dichromatism at the UV spectra is invisible to humans and has been reported to be

11 important in mate choice (e.g. Johnsen et al 1998; Hunt et al 1999; Hausmann et al 2003), we also performed ancestral reconstructions for mean plumage phi for both males and females. To investigate the correlation between overall plumage contrast and degree of sexual dichromatism, and between male and female plumage contrasts, we used the spearman correlation test on phylogenetic independent contrasts (Felsenstein 1985) of males and females color span and mean plumage jnd. We used independent contrasts instead of raw metrics because species are not independent samples so we need to correct for phylogenetic dependency (Felsenstein 1985). All analyses in this section were performed in the R environment (R Core Development Team 2015) using the packages pavo (Maia et al, 2013), ouch (Butler and King, 2004), ape (Paradis et al 2004) and phytools (Revell 2012), except ancestral reconstructions analyses, which were conducted using the software Mesquite (Maddison and Maddison 2015). Results Color space Metrics derived from the avian color space are reported in the supplementary materials. Pittas color span varies from as low as in the cryptic female Eared Pitta (H. phayrei) to as high as in the conspicuously colored male Garnet Pitta (E. granatina). On average, the color span of females was lower (0.159) than those of males (0.184). In general, basal species of the genus Hydrornis, which have many patches of a buff color, showed the smallest color span values. On the other hand, species of the genus Erythropitta and males of the most dichromatic species of the genus Hydrornis showed the highest color span values, mainly because of their structural violet patches and carotenoid-based red colors. The mean plumage chroma varies from in the male Blue Pitta (H. cyanea) to in the male Garnet Pitta. Males average chroma was slightly greater (0.164) than those of females (0.159). Color volume varies from a very small 2.20E-06 in the male Eared Pitta to as high as 2.97E-03 in the male Banded Pitta (H. guajana). The average color volume for males is 5.30E-04 while in females is 3.16E-04. Normalized brilliance ranged from in the male Elegant Pitta (P. elegans) to in the male Graceful Pitta (E. venusta). It is worth noticing that the greatest plumage brightness values among pittas are found in the genus Pitta, are due to a very bright greenish-blue wing patch, which was sometimes

12 even brighter than the white-standard used as a calibration point in the spectrometer. When we controlled for this patch (by removing it from the calculations) we found that the sexually dimorphic males of the genus Hydrornis are the brightest pittas. The average plumage phi ranged from as low as in the female Eared Pitta to as high as an impressive 0.35 in the male Garnet Pitta. As a comparison, the second largest value for this metric was in the male Blue-Headed Pitta (H. baudii). This disproportional value in the Garnet Pitta is due to a number of violet patches in this species. Overall conspicuousness was lower, mainly at the ultraviolet range of the spectra, when plumage coloration was modelled using forest shade as the illuminant. However, evolutionary analyses yielded similar results with both approaches so we report metrics under ideal illumination, so that our results can be compared with previous studies (e.g. Stoddard and Prum 2008; Stoddard and Prum 2011; Shultz and Burns 2013). Reflectance spectra of pitta patches revealed many different hues, ranging from highly saturated red colors to uv-blue and green colors (Figure 3). Many of these patches have at least some reflectance at the U.V range of the spectrum, which is invisible to humans.

13 Figure 3 Different reflectance spectra of plumage patches in the family Pittidae, and their respective position in the avian color space (x and y axes represent wavelength and reflectance, respectively). Color mechanisms are assumptions based on the literature (Review in Hill and McGraw 2006a).-a) Male H. guajana, structural violet breast; b) P. moluccensis, structural ultraviolet blue wing; c) P. brachyura, structural ultraviolet bluish-green wing; d) P. iris, black breast; e) E. kochi, gray back; f) P. maxima, unpigmented white breast; g) Male H. elliotti, ultraviolet green crown; h) P. elegans, green back (combined carotenoid and structural); i) Male H. gurneyi, carotenoid-based ultraviolet yellow breast; j) Male H. guajana, carotenoid-based ultraviolet orange side breast; k) E. erythrogaster, melanin brown nape; l) E. granatina, carotenoid-based red crown. When we compared pittas color diversity with that of birds in general, we found that the Pittidae color volume represents ~26% of the color volume estimated to birds (Table 1). Like other birds, pittas lack saturated blue, purple, green and uv-red colors, but also lack highly saturated violet hues, but in general they have occupied the main regions of the available avian color space (Figure 4). Out of the three pittas genera, Hydrornis has the greatest color diversity, followed by Pitta and Erythropitta, respectively. The latter, however, has the highest overall plumage contrast and saturation among pittas, due to the very contrasting red and violet patches present in some species.

14 Birds (Stoddard and Prum 2011) Pittidae A) B) S V C) L M Figure 4 Pittidae color gamut compared with that of other birds. a) Tetrahedral color space in frontal view, with all plumages patches plotted; b) The same tetrahedral plot, but seen from above; c) Robson projection of color points, which is a visual representation of the color space in two dimensions. Avian cone receptors are labeled as v, s, m and l.

15 Table 1 - Summary color statistics comparing the distribution of plumage colors between birds and pittas Plumage evolution Number % Avian % Avian Color Color Max Average Maximum Average of patches color space color gamut volume span span hue disparity hue disparity chroma Birds E E E E E E-01 Pittidae E E E E E E-01 Hydrornis E E E E E E-01 Pitta E E E E E E-01 Erythropitta E E E E E E-01 Birds' data from Stoddard and Prum (2011) Model fitting results suggest that color volume, brilliance and overall sexual dichromatism are unaffected by ecological and behavioral selective regimes, and are better explained by a Brownian motion model of evolution (Table 2). For these metrics, all other models had worst fit than the BM model. The best fitting model for plumage chroma in both sexes was the OU with single optimum, which was comparable in fit to the altitudinal model in males and to distributional model in females. However, the algorithm estimated unrealistic high values for α, which accounts for the deterministic parameter of the model, for both altitudinal (16.03) and distributional (183.87) models respectively, so we decided it would they should be treated as unreliable. Male and female color span on the other hand, seems to be affected by insular and continental distribution, with the algorithm estimating a higher color span optimum for islands than for the continent, which is contrary to what we had hypothesized. This model performed much better than the elevational gradient, nest placement and single optimum models. However, the ΔAIC between the BM and distribution model was smaller than 2, indicating no differences of fit between them. Table 2 - ΔAIC results of OU models comparisons. "M" and "F" stands for males and females and "JND" stands for just-noticible-differences. Brownian motion OU - single optimum OU - nest placement OU - elevational gradient OU - geographic distribution M Color span F Color span M Color volume F Color volume M Chroma F Chroma M Brilliance F Brilliance Mean JND

16 When analyzing migratory behavior on the genus Pitta, we found support for this model over the BM and single optimum models for color span and color volume in males and mean chroma for both males and females. For all these metrics, the algorithm estimated lower values for the sedentary regime. Table 3 - ΔAIC results of the OU model based on migratory behavior in the genus Pitta. For information on interpretation of results see table 2. Brownian motion OU - single optimum OU - migratory behavior M Color span F Color span M Color volume F Color volume M Chroma F Chroma M Brilliance F Brilliance Mean JND Phylogenetic signal was significant for all metrics evaluated, although it was somewhat weaker for mean plumage chroma in both males and females (Table 4). Table 4 - Pagel's lambda results. Values for λ close to 1 indicate strong phylogenetic signal. Associated log likelihood and p values are reported for each color metric. Male Female Brilliance Color span Chroma Log(volume) jnd λ logl p λ logl p λ logl p λ logl p λ logl p < < < < < < < < < Sexual dichromatism The number of sexually monochromatic species ranged from eight to zero, depending on the methodology used (Figure 4). With Weber s fraction at 0.05, all pittas can be considered sexually dichromatic if the threshold is set to 1 jnd. This percentage decreases drastically if the threshold is set to 3 jnd, which highlights that different parameters should be considered carefully, since they can yield very different results. When we increased the Weber s fraction to 0.1, but kept the threshold as 1 jnd, the number of monochromatic species was somewhat between both the previous results. Regardless of which parameters we used, we found less sexually monochromatic species than categories based in human vision (Figure 4). Below we report results with Weber s fraction set to 0.1, because it has been used in more recent literature (Lind et al 2013; Lind et al 2014; Olsson et al 2015).

17 Individual patches jnd ranged from 0.1 jnd in the E. erythrogaster belly to in the H. gurneyi crown. Mean plumage jnd ranged from in P. anerythra to in H. baudii. Figure 5 Bar plot representing the percentage of sexually monochromatic (Mono) and dichromatic (Di) species using different visual model parameters. Just-noticible-differences (jnd) and Weber s fraction (w) are reported. Out of the 28 species we were able to compare, only P. anerythra, P. maxima and P. superba are truly monochromatic. In general, species of the genus Pitta and Erythropitta were only slightly dichromatic, with only one or two patches having substantial differences between males and females. The exception is E. granatina, which has an intermediate plumage jnd, although to the human eye, females and males cannot be tell apart. Some basal species with cryptic plumage of the genus Hydrornis are also only slightly dichromatic, but the majority of species have the highest jnd values among pittas, with H. baudii and H. gurneyi being the most dichromatic of all. The high values for both species were mainly due to the crown patch, which are of a greenish uv-blue and deep uv-blue in males while in females the crown is of a brownish red and orange (Figure 6). In fact, in all highly dimorphic species of the genus Hydrornis, there is a preference for males to exhibit structural-based colors in patches that are of melanin or carotenoid-based colors in females. For an instance, males H. caerulea, H. scheneideri and H. cyanea all have structural blue backs while females of the same species have melanin brown backs.

18 Figure 6 Plot of mean smoothed spectra for the H. gurneyi (left) and H. baudii (right) crown patches. Line colors represent gender (blue: males and red: females) and shaded areas indicate standard deviation of the spectral data. When we overlapped color volumes between males and females we found that the latter is entirely contained inside the former (Figure 7) and that males volume is almost 2x bigger than that of females ( and respectively). This difference is mainly due to increased hue saturation at the v and l vertices on males. Figure 7 - The figure on the left represents the Pittidae s total volume in the avian color space while the one on the right highlights the male (blue) and female (red) volumes, and the overlap between them (gray).

19 However, it is important to notice that the increased expansion towards the l vertex in males, is mainly caused by two patches only: a saturated yellow patch in H. guajana and a saturated red patch in E. granatina. This pattern changed dramatically when we observed differences between males and females at the species level, as it is shown in Figure 8. Males and females of dimorphic species, mainly from the Hydrornis genus, have small volume overlap, with males normally having the biggest volume between the two. Males also exhibited more patches with high reflectance in the UV spectra if compared to females. Those are chiefly the belly, crown and tail patches, while in females high UV reflectance is restricted to the tail patch. The differences in UV reflectance between males and females can be visualized in the ancestral reconstruction of mean plumage phi (Figure 9). The basal species in the genus Hydrornis have low UV reflectance in both males and females while in the other species, the difference is considerable. In the genus Erythropitta, however, both males and females of E. granatina and E. arquata have high reflectance on the UV. In the genus Pitta UV reflectance is generally low, but it is slightly greater in the male P. steerii, female P. sordida e P. iris and in both male and female P. angolensis. Figure 8 - Color volume overlap between males and females of the most sexually dichromatic species H. baudii, H. elliotti and H. gurneyi. Males volume are represented in blue, females volume in red and the area of overlap in gray. Volumes are spatially oriented to match the tetrahedral color space image (left), with their overall location marked by an X. Notice that the males volume are bigger in these species and more expanded towards the violet vertex than those of females. When reconstructing the ancestral condition of color sexual dimorphism in the family, we found that sexual dichromatism is the ancestral condition in pittas, and that sexual monochromatism has evolved three times in the genus Pitta. The algorithm estimated mean jnd at the tree root, which is an intermediate value of sexual dichromatism for this family. As a comparison, H. nipalensis mean jnd is 1.464, and this species is considered only slightly dichromatic to human observers (Erritzoe 2003).

20 Plumage phi Figure 9 Ancestral reconstruction of overall UV reflectance (phi) in both males and females pittas using the square-change parsimony method. Highly dichromatic males of the genus Hydrornis have more reflectance in the UV than females but, in the genus Erythropitta, both males and females E. granatina and E. arquata have high UV reflectance. Overall, sexual dichromatism is consistent among the genus Pitta and the genus Erythopitta, with the exception of E. granatina, which is considerably dichromatic. The genus Hydrornis is the most sexually dichromatic, which is not surprising, since it is the only genus in which sexual dichromatism can be easily spotted with the naked eye.

21 When the differences between males and females plumage phi is plotted in the phylogeny and place side by side with the mean plumage jnd ancestral reconstruction, we can see that both metrics are visually correlated (Figure 10). This is yet another evidence that intensity in sexual dichromatism is associated with exploitation at the UV spectra. Mean plumage jnd Differences in plumage phi Figure 10. Ancestral state reconstruction using the squared-change parsimony. Mean species JND and difference in mean phi value for males and females. Species with high levels of dichromatism have more differences in the UV between the sexes than other groups. We found a marginally significant correlation between sexual dichromatism and male plumage conspicuousness, but we did not find the same correlation in females. However, we found strong positive correlation between male and female conspicuousness (Figure 11).

22 Figure 11 Spearman correlation test and associated values. A) Correlation between male and female color span; B) Correlation between male color span and mean species plumage jnd; C) Correlation between female color span and mean species plumage jnd. PIC stands for phylogenetically independent contrasts. Discussion Plumage colors Our analyses of reflectance spectra of plumage patches revealed that the variety of plumage colors in pittas is underestimated. Many colors differed in the amount of U.V light they reflect, even if to human eyes they look similar. To our knowledge, only one study investigated the mechanisms behind color production in pittas (Auber 1964), and concluded that green and some greenish blue colors in the genus Pitta are produced by a combination of structural and carotenoid-based mechanisms. However, it is clear that pittas rely on more color mechanisms, both structural and pigmentary, to produce the range of colors which compose their plumage. This has allowed pittas to achieve the equivalent of 26% of the avian color volume previously reported in Stoddard and Prum (2011), and the colors they generally lack have been shown to be very difficult to produce, such as saturated green hues (Stoddard and Prum 2011). Individual genera in the family are also impressively colorful, with the 11 species of the genus Hydrornis occupying around 14% of the avian color space. Since this is the first study comparing the color gamut of a single avian family with that of the birds as a whole, it will be interesting if future studies do something similar to other bird clades, which will enable us to compare the color diversity of these groups under an avian visual model. Studies investigating the mechanisms behind color production in pittas may also provide answers for interesting questions, such as if apparently homologous patches, such as the red bellies in the genus Pitta and Erythropitta, are really shared characters or if they are produce by totally different carotenoids and therefore homoplastic.

23 Plumage evolution and sexual dichromatism Previous reports on pittas sexual dichromatism were based on human observation only, which considered only nine species as being truly sexually dichromatic and seven having very small differences between males and females (Erritzoe 2003). Our results show a different scenario, in which sexual dichromatism is much wider spread in the family, regardless of the threshold used in our analyzes. These findings reinforce the necessity of relying on avian visual models when investigating sexual dichromatism in birds (Cuthill et al 1999; Eaton 2005; Burns and Shultz 2012). However, to a certain extent the degree of dichromatism is not too discrepant from human categorizations, since the great majority of previously reported sexually monochromatic species are only slightly dichromatic under an avian visual model. This is consistent with Seddon et al (2010) study, which concluded that human vision categories are a good estimate of sexual dichromatism, chiefly in birds with the VS cone type, which is the case in pittas (Ödeen et al 2011). In many species, dichromatic patches are restricted to one or two and, although it could be expected only specific patches to be dichromatic, such as those commonly used in courtship behavior, this is not true in pittas. Dichromatism is present in virtually all patches in the family, regardless of the color mechanism present in that patch. For an instance, in many species of the genus Pitta and Erythropitta, carotenoidbased red bellies are dichromatic between males and females, and the same is true for structural greenishblue wings in some species of the genus Pitta. Although the vast majority of studies have focused on sexual dichromatism at the ultra-violet range of the spectra, it has been shown that intersexual differences commonly occur within the human visual capabilities (Eaton 2005). Pittas are therefore a striking example, with sexual dichromatism occurring across the whole avian visual spectra. Regarding the evolutionary pressures underpinning the great color diversity in pittas, our results suggest that plumage coloration is unconstrained by many ecological and behavioral regimes. The BM model outranked OU models based on ecological variables for almost every metric tested. This result by itself does not necessarily implies that pittas plumage coloration evolved under a BM model, since other ecological variables could have played a role in plumage evolution and were not tested here. However, we did find strong phylogenetic signal for all plumage metrics and mean plumage jnd, which gives support to the BM model. However, plumage color span in both males and females, fitted the distributional model ranked better, although it was still comparable to the BM model. Interestingly, our

24 results suggests that high color span is more favorable in islands than in the continent, since the algorithm estimated a highest color span optimum for islands. We had expected that island species would be less conspicuous due to loss in genetic variation aroused by stochastic mechanisms in small populations. It seems then, that either some insular species have a greater population than we had imagined, which is supported by the relatively big size of some islands in the Malayan peninsula, or that directional selection for conspicuity is taking place in these species. Since, plumage traits are long believed to be very labile, with high levels of homoplasy and rapid evolution (e.g. Hackett and Rosenberg, 1990; Burns 1998; Omland and Lanyon 2000; Blomberg et al 2003), it seems intriguing that we found strong correlation between phylogeny and plumage traits. A Brownian motion model suggests that either genetic drift or randomly fluctuating selection has played an important role in pittas plumage evolution (Felsenstein 1988; Harmon et al 2010). However, although genetic drift is a likely mechanism in small, insular populations, it leads to lost in genetic diversity rather than gains, and it is very likely that pittas went through a process of rapid radiation during islands colonization, which is reflected in their huge color diversity. Also, many subspecies are being raised to species level recently, and many of those differ substantially in their plumage pattern (Rheindt and Eaton 2010; Irestedt et al 2013). It seems then, that Brownian motion is not the only evolutionary mechanism behind pittas radiation. The seemingly puzzling association between plumage pattern and phylogeny is easier to grasp when we realize that even though pittas are distributed among many islands and continental areas, their main habitat is very conservative. All pittas inhabit tropical and equatorial forests, and since one of the main forces behind adaptive radiation is ecological opportunity (reviewed in Yoder et al 2010), it is not entirely surprising that other mechanisms, rather than convergent evolution, can explain plumage patterns in this group. Our results are also consistent with Irestedt et al (2006) observation that the Pittidae molecular phylogeny resembled well old phenetics classifications based only on plumage coloration. For an instance, although the genus Pitta is widely distributed, with some species inhabiting small and isolated islands while others big landmasses, they have a somewhat homogeneous pattern of plumage colors. Still, we should have expected that variation in nest placement and altitudinal distribution among pitta species would have affected plumage coloration somehow. So why it did not? Nests that are exposed, by being built open or at ground level, should be more vulnerable to predation (Wallace 1868; Soler and Moreno 2012) and inconspicuous plumage is expected to be advantageous, at least to the sex attending it. In pittas however, it has been observed wide individual variation in nest placement, which suggests

25 that some species can adapt to the presence of local predators (Erritzoe 2003; Hutchinson and Mears 2006). This behavioral plasticity, together with bi-parental care, which is believed to reduce predation pressures on the nest, can explain the lack of correlation between nesting behavior and sexual dichromatism in this family. The Pittidae mating system is also a likely explanation for the lack of plumage variation along elevational gradients. Sexual dichromatism has been associated with increase in male care in higher altitudes, due to colder micro-climate and spatial separation of feeding resources (Badyaev, 1997a; Badyaev and Ghalambor, 2001). However, male parental care occurs in all species of pittas studied so far, regardless of the elevational gradient in which they are more abundant. Also, although there a representative number of pittas which are lowland specialists, there are just a couple that are truly montane. The majority of pittas are distributed along a high elevational range, which can vary greatly between localities. For an instance, H. cyanea occurs from as low as 60m to up to 2000m. Once more, it seems that pittas lack strong niche separation that would allow directional selection to shape differences in plumage colors. Although we could not find any relationship between ecology and plumage colors at the family level, we did find support for the migratory regime in the Pitta genus. This result is consistent with Irestedt et al (2006) observation that the shift from migratory to sedentary habits in the genus Pitta had promoted changes in plumage coloration. They hypothesized that plumage diversification could be either constrained by the migratory habit or promoted by genetic drift in small and sedentary populations. In the first case, migratory species could be gaining from a duller plumage by increasing camouflage during flight, or rather, the energy-cost of migration could be limiting plumage diversification. We found support for the migratory behavior model for male color span and color volume and average chroma for both males and females. The fact that we found different results for males and females is very interesting, because differential migration has been reported for two species of pitta (Erritzoe 2003). It seems reasonable them, that differential migration affected the evolution of plumage coloration in both genders, but is not consistently associated with the intensity of sexual dichromatism. This result is similar to Dunn et al (2015) who found an association between migration and plumage brightness in both male and females, but not in sexual dichromatism. In general, sedentary species have lower values for color metrics than migratory ones, mainly because they have more black patches in their plumage. The exception is P. steerii, which has high color span and volume, due to a bright blue wing and breast patches. Since migratory species are more conspicuous

26 than are their sedentary relatives, we do not find support for the camouflage hypothesis proposed by Irestedt et al (2006). We believe that genetic drift has played a more important role in these insular species, since the majority of them have lost, rather than gained color diversity, which is expected in small, isolated populations. In fact, the only truly monochromatic species belong to the Pitta genus, and are precisely those with the smallest reported population and distributional range. Although we assessed the influence of natural selection on plumage coloration, the lack of data on pittas population density and reproductive behavior prevented us to investigate sexual selection mechanisms directly. For an instance, data on courtship behavior are virtually non-existing and studies on testis size and extra-pair paternity have never been done for pittas. Some of our results, however, provide interesting insights that can be used as a starting point for future research on the topic. For an instance, we discovered that males from highly sexually dichromatic species in the genus Hydrornis have more ultraviolet-blue patches than females of the same species. This is interesting because UV reflectance has been widely associated with mate choice (e.g. Johnsen et al 1998; Hunt et al 1999; Hausmann et al 2003), which suggests that, even though the mating system is the same among pittas, intensity of sexual selection may vary. In some species of the genus Erythropitta, however, both males and females have many violet patches that, in combination with its highly saturated red patches, allow these species to reach some of the highest plumage color span among pittas. E. granatina is specially remarkable, in which it has astonishingly high levels of UV reflectance if compared to other pittas. There are currently no distinguishable ecological or behavioral data on this species that would justify this peculiarity, and it will be interesting to investigate why E. granatina seems to rely so much in UV reflectance. Although overall plumage contrast is generally smaller in females, chiefly when sexual dichromatism is high, some of female pittas are quite conspicuous. In fact, color span is correlated between males and females, and in the genus Hydrornis there was an increase in sexual dichromatism with both males and females, which led to an increase in plumage conspicuousness. Basal species of the genus have mainly buff patches whereas derived species evolved a number of structural and carotenoid-based colors. It is possible then, that both male and female plumage coloration are being sexually selected or that female conspicuousness is driven by genetic correlation with males (e.g. Amundsen 2000; Kraaijeveld et al 2007; Cardoso and Mota 2010). Also, although pittas are socially monogamous with bi parental care as a rule, extra-pair paternity has been shown to be quite high in many bird species, regardless of their mating system (Møller and Birkhead 1994; Birkhead and Møller 2008). The fact that many pittas are strongly territorial, seems to us as an indirect evidence that intra-sexual conflict may be common among