Colour in a new light: a spectral perspective on the quantitative genetics of carotenoid colouration

Functional Ecology 215, 29, 96 13 doi: 1.1111/1365-2435.12297 Colour in a new light: a spectral perspective on the quantitative genetics of carotenoid colouration Simon R. Evans*,1,2 and Ben C. Sheldon 1 1 Edward Grey Institute, Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK; and 2 Department of Animal Ecology, Evolutionary Biology Centre, Uppsala University, Norbyv agen 18 D, 752 36, Uppsala, Sweden Summary 1. Carotenoid-based colours are model traits for research on animal signalling and sexual selection but, whereas the consequences of variable expression have been extensively studied, its causes are rarely quantified. This issue is complicated by the composite nature of carotenoid-based colour patches, which combine pigments and a reflective background. Ultimately, the evolution of such colours will be determined by the processes that govern variable expression of these mechanisms. 2. We present a novel approach to assessing the quantitative genetics of colour expression, in which reflectance spectra are analysed directly, thereby avoiding the data loss and inherent subjectivity of summary colour variables. Further, the influence of the component mechanisms can be distinguished in spectral analyses due to their contrasting wavelength-dependencies. 3. Using data from a 6-year study of carotenoid-based plumage reflectance in wild great tits (Parus major), we employ a multi-parallel animal modelling approach to estimate sources of variance for narrow (2 nm) wavebands across the visible spectrum. 4. Moderate heritability estimates were limited to the violet-blue region of the spectrum, diagnostic of the carotenoid content of plumage being heritable. The natal environment effect was limited entirely to the violet-blue, again indicating that it relates to variation in carotenoid content of feathers. Other wavelengths were sensitive to annual and permanent environmental variation but only marginally influenced by additive genetic variation. Hence, background reflectance is the component that is more sensitive to the environment. 5. Analysing reflectance spectra directly provided an objective perspective of the dynamics of colour expression that is not apparent when relying on summary colour scores. In this case, our results suggest that carotenoid deposition may be an effective target of selection and hence could explain the important role carotenoids frequently play in intraspecific signalling. Key-words: animal model, great tit, heritability, natal environment, plumage reflectance Introduction The phenotypic variability in colour traits has been appreciated by biologists for more than a century (Wallace 1881) and animal colouration has provided many model traits for the study of sexual selection and animal signalling. However, the study of within-population colour variation has historically been concerned with the consequences of variable expression; relatively few studies have examined how continuous colour variation is generated. In this regard, understanding lags behind that of other character types (e.g. reproductive and morphometric traits), which have been extensively considered within a quantitative *Correspondence author. E-mail: simon.evans@ebc.uu.se genetic framework (Kruuk 24; Kruuk, Slate & Wilson 28). Hence, there remains uncertainty about the relative importance of genetic and environmental factors in determining colour expression, despite the importance of this for determining responses to selection (Falconer & Mackay 1996; Mundy 26). For conspicuous colour patches, such knowledge is particularly valuable, given their obvious potential to function as signals and thereby communicate information about the bearer: elucidating the information content of a colour signal requires that we identify the factors determining phenotypic expression. Among animal colours, those based on carotenoid pigmentation have received particular attention in recent decades (McGraw 26). Initial interest was perhaps stimulated by the fact that they represent many of the 214 The Authors. Functional Ecology 214 British Ecological Society

Colour in a new light 97 gaudiest animal displays. However, the realisation that vertebrates are unable to autosynthesise carotenoids and must thus sequester them from the environment via their diet stimulated research interest: the dietary origin of pigments was expected to provide an inherent enforcement of signal honesty (Endler 198). However, environmental origins do not imply that pigment availability is limiting (Griffith, Parker & Olson 26) and, while dietary carotenoid restriction negatively impacts colour expression of captive animals (Hill 1992; McGraw & Hill 21), empirical demonstrations of dietary pigment limitation driving colour variation in wild populations are infrequent (Hadfield & Owens 26; Evans & Sheldon 213; though see Hill, Inouye & Montgomerie 22). Thus, while carotenoids are required in the diet in order for colouration to be expressed, we have relatively little evidence that they are a limiting factor in many natural environments. This potential mismatch between studies of captive and wild animals demonstrates the necessity of examining issues of colouration in wild populations for improving understanding of the evolutionary ecology of colour. Common garden experiments have found interpopulation differences in the allocation of carotenoids to sexual colouration of sockeye salmon (Oncorhynchus nerka), with crosses creating hybrid progeny of intermediate phenotype, indicating that genetic variation can generate differences between populations (Craig & Foote 21). More recently, the adoption of mixed effects modelling by evolutionary biologists (Kruuk 24) has allowed the heritability of carotenoid-based colouration to be estimated directly in study populations of wild birds, yielding estimates of significant additive genetic variation (Hadfield et al. 27; Evans & Sheldon 212). However, pigments are subtractive colourants (Andersson & Prager 26) such that carotenoid-based colours are actually composite traits, with reflectance determined by a combination of carotenoid content and background reflectance (Grether, Kolluru & Nersissian 24; Shawkey & Hill 25); focussing on the displayed or perceived colour (Montgomerie 26), while invaluable in the context of perception, overlooks the mechanistic level at which evolutionary change in colouration is mediated. For carotenoid-based plumage colours, both experimental removal of carotenoids using solvents (Shawkey & Hill 25; Shawkey et al. 26) and simulations (Andersson & Prager 26; Jacot et al. 21) support this bimechanistic model, in which plumage reflectance is determined by two attributes of the feathers carotenoid content and background reflectance both of which vary across and within individuals (Matrkova & Remes 212; Evans & Sheldon 213). Using reflectance spectra to generate estimates of both carotenoid content (Peters et al. 24; Andersson & Prager 26) and background reflectance (Jacot et al. 21) is a popular approach but, for carotenoid content at least, reliability is moderate (Isaksson et al. 28; Butler, Toomey & McGraw 211). An alternative approach is to directly analyse reflectance spectra, thereby making greater use of the data contained in reflectance spectra. Several studies have partitioned the visible spectrum into narrow wavebands, with each waveband analysed individually via a multi-parallel approach, allowing the wavelength-specificity of effects to be visualised (Brunton & Majerus 1995; Hunt et al. 1998; Griggio et al. 29; Evans, Summers & Sheldon 212; Evans & Sheldon 213). Wavelength-specific analysis provides an objective perspective on expression of a colour patch. Furthermore, for carotenoid-based colours, this approach allows the influence of the two component mechanisms to be discriminated based on their contrasting spectral influences: background reflectance has a relatively uniform, broad-spectrum effect (Shawkey & Hill 25; Shawkey et al. 26), while the influence of carotenoid pigmentation is limited almost exclusively to violet-blue wavelengths (4 5 nm) and has a characteristic spectral shape (Zsceile et al. 1942; Ruban, Horton & Young 1993). Here, we develop this approach further by examining the wavelength-specific quantitative genetics of plumage reflectance of a carotenoid-based colour trait. We use reflectance data from a 6-year study of plumage colouration in a wild great tit (Parus major) population to examine the quantitative genetics of reflectance for a carotenoid-based colour patch. By focussing on reflectance measures, we avoid the subjectivity of summary colour scores (Bennett, Cuthill & Norris 1994) and can identify the mechanistic route by which changes in phenotypic variation are achieved. The perspective facilitated by wavelength-specific analyses of reflectance spectra thus informs us of both the reflective changes that give rise to variation in perceived colour and of the mechanisms by which these are effected. Materials and methods STUDY SITE AND SPECIES Plumage reflectance was sampled from May 28 to June 213 (i.e. including six consecutive breeding seasons) from a nestboxbreeding great tit (P. major) population in Bagley Wood (51 42 N, 1 15 W), near Oxford, UK. Since January 27, 51 nestboxes have been in place at this site, arranged across 12 similarly sized plots. Sampling was conducted from September to June: in winter (September March), great tits were caught at baited mist-net sites; while in May and June, breeders were captured during chick provisioning. Unringed birds were fitted with a uniquely numbered aluminium leg-ring. Sex and age-class [first-year or adult ( 2 yearold)] were scored according to standard plumage characteristics (Svensson 1994). While we did not sample plumage reflectance of individuals until after the post-fledging moult, all nestlings were ringed prior to fledging, at 14 days post-hatching, so that native recruits could be later identified. In 29, a cross-fostering experiment was carried out (Evans & Sheldon 212); because variance components are expected to be altered in cross-fostered individuals they were excluded from the analyses. PLUMAGE REFLECTANCE AND SPECTRAL COMPARTMENTALISATION Ventral plumage reflectance across the bird-visible spectrum (32 7 nm: Andersson & Prager 26; Montgomerie 26) was

98 S. R. Evans & B. C. Sheldon measured directly (i.e. not on collected feathers) using a field-portable spectrophotometer (USB4; Ocean Optics) and xenon lamp (PX-2; Ocean Optics), as detailed elsewhere (Evans et al. 21). Plumage reflectance was sampled at the midpoint of a line between the sternum and the right shoulder, using a coincident normal geometry (Andersson & Prager 26), and was calculated relative to a dark box and a white Spectralon tile. Three reflectance measures were taken and the mean of these used as the plumage reflectance measure for that sample. To conduct wavelength-specific analyses, we divided the bird-visible spectrum into 2 nm-wavebands, for each of which we calculated the median reflectance value (Cuthill et al. 1999). Each 2 nm waveband was analysed as a distinct trait. This multi-parallel approach to analysis allowed us to quantify and visualise across-spectrum changes in variance components, which can be attributed to either of the component mechanisms determining plumage reflectance (carotenoid content versus background reflectance: Grether, Kolluru & Nersissian 24; Shawkey & Hill 25; Shawkey et al. 26; Jacot et al. 21) based on their contrasting wavelength-dependencies (Evans et al. 21; Evans, Summers & Sheldon 212; Evans & Sheldon 213). Given both the smooth reflectance spectra of objects in nature and the bimechanistic model of reflectance determination for this trait, adjacent wavebands will be non-independent, in the sense that any natural colour has a smooth spectral shape, allowing the value in one waveband to be predicted, to some extent, from that in the adjacent waveband. Hence, although each waveband is analysed independently of all others, spectrally auto-correlated responses of wavebands are to be expected. In this sense, our analyses are similar to any multi-trait analysis (e.g. morphology) where there may be underlying factors (e.g. size) influencing multiple traits. QUANTITATIVE GENETIC ANALYSES We used an animal model to partition the phenotypic variance (V P ) for each of the 19 wavebands into five components: V P ¼ V A þ V PE þ V BE þ V Y þ V R where V A is the additive genetic variance, V PE is the permanent environmental variance (other sources of between-individual variation), V BE represents variance attributable to the brood environment, V Y models the annual variance and V R represents error deviance. Random effects describing contemporary environment (nestbox plot in which the individual was captured), cohort (i.e. hatching year) and maternal (i.e. maternal identity) effects were included in initial models but yielded marginal estimates and so are not included in the model presented here. Individual-level variation could be partitioned into its additive genetic and permanent environmental components because we constructed a pedigree describing familial relationships, from which a relatedness matrix for all sampled individuals was calculated. Our pedigree assigned maternities and paternities on the basis of behavioural observations (provisioning of nestlings). While this will be subject to errors due to extra-pair paternity (EPP), rates of EPP in this species (13% for a nearby population: Patrick et al. 212) are not expected to have much impact on estimates of quantitative genetic parameters (Charmantier & Reale 25). The pruned pedigree (excluding uninformative links) featured 1691 individuals, with 551 maternities, 539 paternities and 446 full sibships and mean sibship size of 211 via both maternal and paternal linkages. Shared environmental experiences among relatives will potentially inflate heritability estimates because between-individual covariance of environmental origin may be erroneously attributed to shared heritable influences. Such an issue is perhaps most obvious between brood-mates, which share a natal environment, so we explicitly modelled the phenotypic variation attributable to the brood environment. This random effect is defined as the nestbox each individual fledged from and is year-specific, such that we have a brood-level effect. Many sampled individuals (178 of 1618; 666%) were first captured at maturity, such that their natal origin (and parentage) is unknown. These individuals are uninformative for the estimation of the brood environmental effect but are retained in the data set as they contribute to estimation of other model parameters. The standard errors of each estimate reflect the information available: an advantage of the animal model is that it is robust to unbalanced datasets and missing data (Kruuk 24). Year was defined relative to the annual moult cycle, rather than calendar date (Evans et al. 21). Sex and age-class (first-year or adult) were included as categorical fixed effects as they are known to influence plumage reflectance of great tits (Evans et al. 21; Evans & Sheldon 213). We also fitted Septemberday (day of sampling, relative to 31st August) as a continuous fixed effect to model seasonal change in plumage reflectance (Evans et al. 21; Evans, Summers & Sheldon 212). Statistical support for variance estimates was assessed via likelihood ratio tests, in which the log-likelihood value of the full model is compared to that of a model in which the focal component has been omitted. The test statistic, defined as twice the difference in log-likelihood between the two models is assumed to follow a mixture of two chi-squared distributions, with zero and one degree of freedom, respectively (Self & Liang 1987). Analyses were conducted in R v.2.14. using the ASReml-R package (VSN International, Hemel Hempstead, UK), with results based on restricted maximum likelihood. Phenotypic and pedigree datasets are archived in the Dryad Digital Repository (doi: 1.561/ dryad.1cn2; Evans & Sheldon 214), along with the R script used for the analyses. Results For each 2 nm waveband within the bird-visible spectrum we analysed 294 records representing 1618 mature individuals. Each waveband was analysed independently of all others but by arranging the wavebands in sequence, we are able to visualise how the variance composition of plumage reflectance changes across the spectrum (Figs 1 and 2). The resulting plots show that some variance components are highly wavelength-dependent, while others are more uniform across the spectrum. These differences reveal the influence of either mechanism (carotenoid content versus background reflectance) in mediating the effect on plumage reflectance: carotenoids selectively absorb violet-blue (4 5 nm) light, while the influence of background reflectance is relatively consistent across the spectrum (Shawkey & Hill 25; Shawkey et al. 26; Jacot et al. 21). Moderate and highly significant estimates of additive genetic variation were restricted to the violet-blue region of the spectrum (Fig. 1f). This implies heritable variation in reflectance is mediated largely via an influence on carotenoid content of the plumage, because carotenoids selectively absorb violet-blue light but have little influence on reflectance across the rest of the spectrum. Outside the violet-blue, where we expect reflectance to be determined by background reflectance alone, estimates of additive genetic variation were small [though weakly significant in

Colour in a new light 99 (a) Reflectance (%) (b) (c) (d) (e) (f) 6 4 2 15 1 5 6 5 4 3 2 1 6 4 2 35 3 25 2 15 1 5 2 15 1 5 Mean reflectance spectr um P > 5 P < 5 P < 1 Phenotypic variance, V P Annual variance, V Y Brood environmental variance, V BE Permanent environmental variance, V PE Additive genetic variance, V A 35 4 45 5 55 6 65 7 Wavelength (nm) Fig. 1. Estimates of the wavelength-specific phenotypic variance and its components for ventral plumage reflectance of great tits. The bird-visible spectrum (32 7 nm) is divided into 2 nm-wide wavebands, for each of which variance parameters are estimated using an animal model. (a) Mean (standard deviation) reflectance; (b) wavelength-specific estimates of phenotypic variance (standard error), conditioned upon fixed effects; (c f) wavelength-specific estimates (standard error) of variance components. Shading indicates the level of statistical support for each estimate (total number of wavebands = 19), based on likelihood ratio tests (see key in plot c). parts of the yellow-red (5 7 nm) and UV (3 4 nm) sectors], indicating that variation in background reflectance is little influenced by heritable factors. These Proportion of phenotypic variance, V P 1 8 6 4 2 across-spectrum changes in the additive genetic variance translate into spectral variation in heritability of reflectance (Fig. 2; heritability plotted in red): heritability exceeds 2% within the violet-blue sector but is more limited (approximately 3 5%) across the rest of the spectrum. With respect to environmental sources of variation, the effect of the natal environment on plumage reflectance at maturity is observed solely in the violet-blue (Fig. 1d), implying that it is mediated solely via carotenoid content. Conversely, estimates of annual variance were largest in the yellow-red sector of the spectrum (Fig. 1c), suggesting that annual variation in plumage reflectance is mediated via background reflectance. Indeed, this effect, which was larger in magnitude than the other factors considered here, appears to be responsible for driving the greater phenotypic variance in the yellow-red sector of the spectrum, compared to shorter wavelengths (Fig. 1b). The permanent environmental effect (Fig. 1e) similarly exhibits a relatively consistent influence on reflectance across the spectrum. Discussion 35 4 45 5 55 6 65 Wavelength (nm) V RESIDUAL V YEAR V BROOD ENVIRONMENT V PERMANENT ENVIRONMENT V ADDITIVE GENETIC Fig. 2. Standardised variance decomposition plot, depicting the estimated division of total phenotypic variance (conditional upon fixed effects; see Methods) in ventral plumage reflectance of great tits. The bird-visible spectrum (32 7 nm) is divided into 2 nmwide wavebands, for each of which variance parameters are estimated using an animal model. Our previous work (Evans & Sheldon 212) typifies the approach of much research on animal colouration, which has generally focussed on quantifying and understanding colour expression either in anthropocentric terms (e.g. tristimulus colour variables) or, by incorporating models of animal vision (Kelber & Osorio 21), from the perspective of conspecifics (Hadfield et al. 26; Stoddard & Prum 28). Such colour measures (e.g. hue, saturation and brightness, or the SWS ratio of Evans & Sheldon 212) do not determine the influence of specific mechanisms on

1 S. R. Evans & B. C. Sheldon reflectance, but instead describe observable variation from the perspective of human or animal perception. While highly informative, particularly with respect to understanding the performance of colour signals, this focus neglects an additional level at which colour variation can be considered that provides insight into the evolution of colour traits: one that focuses on the mechanisms underpinning colour variation. Wavelength-specific analyses more fully exploit the data-rich nature of reflectance spectra which describe a physical property of the plumage and are free of the subjective assumptions inherent in summary variables (e.g. tristimulus colour scores). The resulting spectral plots are highly informative and provide a novel perspective on variable colour expression. Given an appreciation of the composite nature of carotenoidbased colours and the contrasting spectral influences of the two component mechanisms (carotenoid content and background reflectance: Grether, Kolluru & Nersissian 24; Shawkey & Hill 25; Shawkey et al. 26; Jacot et al. 21), they also provide a mechanistic understanding of the means by which variation in reflectance occurs. When applied to a 6-year data set describing ventral plumage reflectance in a wild great tit (P. major) population, this approach indicated that the carotenoid content of plumage is moderately heritable while background reflectance is more sensitive to environmental variation. Even without detailed consideration of the spectral variance decomposition plot (Fig. 2), it is clear that the birdvisible spectrum can be broadly divided into three regions, within which plumage reflectance tends to respond consistently to each variance effect: UV (32 4 nm), violetblue (4 5 nm) and yellow-red (5 7 nm) sectors. Broadly, these same sectors were demarcated by MacDougall & Montgomerie (23) based on an understanding of the wavelength-specific effect on reflectance of carotenoids deposited into feathers: carotenoids exhibit highly specific absorptance of violet-blue light (Zsceile et al. 1942). Beyond the violet-blue sector, plumage reflectance is determined by background reflectance, this having a more achromatic influence on reflectance. This three-sector division, consistent with our bimechanistic model of carotenoid colouration, is instructive when interpreting the wavelengthspecific effects we report. Given the large number of mechanistic measures in the literature (Andersson & Prager 26; Jacot et al. 21; Butler, Toomey & McGraw 211) and reports of limited accuracy when estimating carotenoid content from reflectance spectra (Isaksson et al. 28; Butler, Toomey & McGraw 211), it is difficult to know how well such indices quantify variation at the mechanistic level without conducting pigment extraction procedures on sampled feathers. Using an alternative, objective method of analysing reflectance measures multi-parallel wavelength-specific analyses enabled us to circumvent these issues. The wavelength-specific analyses provide visual representations of the heritability of plumage chromaticity we reported previously for this trait (Evans & Sheldon 212). The additive genetic effect was highly wavelength-dependent, with the peak in the violet-blue indicating that a large proportion of the heritable variation in reflectance is realised via deposition of carotenoids; indeed, the characteristic, multi-peaked absorptance spectrum of carotenoids (Zsceile et al. 1942; Ruban, Horton & Young 1993) is increasingly apparent as spectral resolution is improved (i.e. narrower wavebands; cf. Evans, Summers & Sheldon 212). Marginal estimates outside the violet-blue sector suggest that background reflectance may also be weakly heritable. Carotenoid colouration is widespread across multiple animal taxa and, within these, often exhibits remarkable variety (Mahler, Araujo & Tubaro 23; Hofmann, Cronin & Omland 26), indicating that processes of elaboration, diversification and convergence have repeatedly taken place within lineages (Badyaev 27). The evolvability this implies is problematic, given the historical view of withinpopulation variation in carotenoid colouration being the result of limited pigment availability (Endler 198; Hill 1992): if phenotypic variation lacks a heritable component then selection would not result in evolutionary change (Falconer & Mackay 1996). Our results suggest that the importance of heritable factors in contributing to variable carotenoid content of colour patches may have been overlooked and, if representative of carotenoid-based colours more generally, suggest there is sufficient additive genetic variation to facilitate an evolutionary response to selection. Evans & Sheldon (212) reported a significant and persistent natal environment effect on plumage chromaticity at reproductive maturity in this population, an influence that we show here to be mediated exclusively via the carotenoid content of plumage. While the impact of the contemporary environment on carotenoid colouration has received extensive attention (McGraw 26), the potential for long-term environmental effects is less frequently considered. For a trait that is replaced each year, such as plumage colouration, the persistence of a natal environmental effect is particularly interesting, especially given the general expectation for contemporary environmental variation to dominate determination of carotenoid-based colour expression. The first moult occurs within weeks of fledging (Jenni & Winkler 1994); all our analyses relate to measurements of plumage grown since leaving the nest. Thus, the natal environment effect we report is not being expressed in integumentary tissue that was grown at that time (Jacot et al. 21; Matrkova & Remes 212) but is instead a persistent influence that is observed in subsequent, temporally distinct tissue growth. Given that colouration as a nestling may not be related to plumage colouration at maturity in this species (Fitze, Kolliker & Richner 23), it may even be that this natal environment effect on carotenoid deposition is only expressed after the post-fledging moult. Clearly, the dynamics of the developmental pathway of carotenoid colouration during feather growth must be more complex than a simple source-sink translation, even for species such as great tits where ingested carotenoids

Colour in a new light 11 are not chemically modified (McGraw 26), and there are indeed multiple points along the pathway between carotenoid procurement and deposition into feathers that may be subject to heritable sources of variation (Parker 1996; Walsh et al. 212). Our results raise the possibility that components of this pathway exhibit irreversible plasticity, being sensitive to environmental cues early in life but fixed thereafter (Utz et al. 214). In contrast to the violet-blue, plumage reflectance in the UV and yellow-red sectors which is largely determined by background reflectance was only weakly influenced by additive genetic variance. Although little work has addressed the heritability of purely structurally based plumage colours, low but statistically significant heritability estimates have been reported (Johnsen et al. 23; Hadfield et al. 27). Nonetheless, our results show that annual variation was a major contribution to phenotypic variation in the yellow-red and, to a lesser extent, UV sectors. Indeed, the greatest phenotypic variance in reflectance was observed in the yellow-red sector, where carotenoids have little influence on reflectance. A study of burrowing parrots (Cyanoliseus patagonius) reported marked annual variation in broad-spectrum reflectivity of a structurally based plumage colour (Masello, Lubjuhn & Quillfeldt 28). The authors attribute this year-level variation to climatic factors, which may also be driving the annual variance effect we report here (although we have insufficient data to explore the causes of variation between years). If so, our results suggest background reflectance is a more obvious candidate as a signal of current condition than the carotenoid component. The contrasting sensitivities of the two mechanisms to environmental and heritable variation indicated by our analyses are consistent with a multiple messages perspective on pigment-based colours, in which separate pieces of information may be communicated within a single colour patch (Senar et al. 28): variation in each mechanism corresponds neatly with the variation we observe in this trait when plumage reflectance is described using an avian visual model (Evans & Sheldon 212), with carotenoid content of plumage translating to plumage chromaticity and background reflectance driving changes in plumage brightness. These two visual dimensions chromaticity and brightness correspond with our understanding of avian visual processing, in which independent neural pathways handle chromatic and luminance cues (Kelber & Osorio 21). It is possible, therefore, that expression of the two mechanisms can be perceived separately by conspecifics, and this could be exploited to simultaneously convey multiple messages if the colour patch has a signalling role (Shawkey & Hill 25; Bitton & Dawson 28). In fact, as already outlined for carotenoid-based plumage colours by Shawkey & Hill (25), background reflectance is essential if carotenoid content of plumage is to be observable (pigments are subtractive colourants: Andersson & Prager 26). Thus, it may be that on an evolutionary time-scale carotenoids are incorporated into achromatic plumage patches (such as are seen in many bird species, e.g. Ferns & Hinsley 24; Griggio et al. 211) to restore information content (i.e. individual genetic quality) that had been eroded from the original signal through the action of selection. Such a transition would provide a second dimension to variation in signal expression, as expected by the multiple messages hypothesis, in which the information content of signals is enhanced by incorporation of novel, non-redundant components (Bro-Jørgensen 21). In summary, we present a wavelength-specific approach to quantifying colour variation, in which we avoid condensing spectral variation into a small number of summary variables and instead focus more directly on reflectance spectra. We find that moderate genetic variation in reflectance is limited to violet-blue wavelengths, indicative of heritability for the carotenoid content of feathers. Similar wavelength-specificity of the natal environmental effect likewise indicates that early environmental experiences influence the carotenoid content of plumage at maturity (i.e. following replacement of nestling feathers). In contrast, for regions of the spectrum beyond the major influence of carotenoids, where background reflectance alone is expected to determine reflectance, heritability was marginal and reflectance was more sensitive to contemporary environmental variation. If representative of other carotenoidbased colours, these results have implications for our understanding of the information content of colour signals and suggest carotenoid pigmentation may retain evolutionary potential comparable to many other traits. Acknowledgements We are grateful to: St. John s College (Oxford) for allowing us to conduct research at the study site and to their staff for accommodating us; Hannah Edwards, Camilla Hinde, Julian Howe, Caroline Isaksson, Sam Jones, Adele Mennerat, Nicole Milligan, Alexia Mouchet, Sarah Roberts, Teddy Wilkin, Helen Wilkinson and Stuart Will for assistance during breeding seasons; Staffan Andersson and Kevin McGraw for insightful discussion; Bj orn Rogell and Ian White for assistance with data analysis; Simon Pickett for providing breeding records from preceding years; and the National Environment Research Council (NE/D11744/1; to BCS), the European Research Council (AdG 25164; to BCS), the Swedish Research Council (grant to Lars Gustafsson) and Zoologisk Stiftelsen (to SRE) for funding. Data accessibility Data deposited in the Dryad repository: http://doi.org/1.561/dryad.1cn2 (Evans & Sheldon 215) References Andersson, S. & Prager, M. (26) Quantifying colors. Bird Coloration. Volume I: Mechanisms and Measurements (eds G.E. Hill & K.J. McGraw), pp. 41 89. Harvard University Press, Cambridge, MA. Badyaev, A.V. (27) Evolvability and robustness in color displays: bridging the gap between theory and data. Evolutionary Biology, 34, 61 71. Bennett, A.T.D., Cuthill, I.C. & Norris, K.J. (1994) Sexual selection and the mismeasure of colour. American Naturalist, 144, 848 86. Bitton, P.-P. & Dawson, R.D. (28) Age-related diferences in plumage characteristics of male tree swallows Tachycineta bicolor: hue and brightness signal different aspects of individual quality. Journal of Avian Biology, 39, 446 452.

12 S. R. Evans & B. C. Sheldon Bro-Jørgensen, J. (21) Dynamics of multiple signalling systems: animal communication in a world in flux. Trends in Ecology and Evolution, 25, 292 3. Brunton, C.F.A. & Majerus, M.E.N. (1995) Ultraviolet colours in butterflies: intra- or inter-specific communication? Proceedings of the Royal Society of London, Series B: Biological Sciences, 26, 199 24. Butler, M.W., Toomey, M.B. & McGraw, K.J. (211) How many color metrics do we need? Evaluating how different color-scoring procedures explain carotenoid pigment content in avian bare-part and plumage ornaments. Behavioral Ecology and Sociobiology, 65, 41 413. Charmantier, A. & Reale, D. (25) How do misassigned paternities affect the estimation of heritability in the wild? Molecular Ecology, 14, 2839 285. Craig, J.K. & Foote, C.J. (21) Countergradient variation and secondary sexual color: phenotypic convergence promotes genetic divergence in carotenoid use between sympatric anadromous and nonanadromous morphs of sockeye salmon (Oncorhynchus nerka). Evolution, 55, 38 391. Cuthill, I.C., Bennett, A.T.D., Partridge, J.C. & Maier, E.J. (1999) Plumage reflectance and the objective assessment of avian sexual dichromatism. American Naturalist, 153, 183 2. Endler, J.A. (198) Natural selection on color patterns in Poecilia reticulata. Evolution, 34, 76 91. Evans, S.R. & Sheldon, B.C. (212) Quantitative genetics of a carotenoidbased colour: heritability and persistent natal environmental effects in the great tit. American Naturalist, 179, 79 94. Evans, S.R. & Sheldon, B.C. (213) Pigments versus structure: examining the mechanism of age-dependent change in carotenoid-based plumage coloration. Journal of Animal Ecology, 82, 418 428. Evans, S.R. & Sheldon, B.C. (215) Data from: Colour in a new light: a spectral perspective on the quantitative genetics of carotenoid coloration. Dryad Digital Repository. http://doi.org/1.561/dryad.1cn2 Evans, S.R., Summers, A.G.R. & Sheldon, B.C. (212) Seasonality of carotenoid-based plumage coloration: modelling wavelength-specific change through spectral reconstruction. Journal of Avian Biology, 43, 234 243. Evans, S.R., Hinks, A.E., Wilkin, T.A. & Sheldon, B.C. (21) Age, sex and beauty: methodological dependence of age- and sex-dichromatism in great tits. Biological Journal of the Linnean Society, 11, 777 796. Falconer, D.S. & Mackay, T.F.C. (1996) Introduction to Quantitative Genetics. Pearson Education, Harlow, UK. Ferns, P.N. & Hinsley, S.A. (24) Immaculate tits: head plumage pattern as an indicator of quality in birds. Animal Behaviour, 67, 261 272. Fitze, P.S., Kolliker, M. & Richner, H. (23) Effects of common origin and common environment on nestling plumage coloration in the great tit (Parus major). Evolution, 57, 144 15. Grether, G.F., Kolluru, G.R. & Nersissian, K. (24) Individual colour patches as multicomponent signals. Biological Reviews, 79, 583 61. Griffith, S.C., Parker, T.H. & Olson, V.A. (26) Melanin- versus carotenoid-based sexual signals: is the difference really so black and red? Animal Behaviour, 71, 749 763. Griggio, M., Serra, L., Licheri, D., Campomori, C. & Pilastro, A. (29) Moult speed affects structural feather ornament in the blue tit. Journal of Evolutionary Biology, 22, 782 792. Griggio, M., Valera, F., Casas-Criville, A., Hoi, H. & Barbosa, A. (211) White tail markings are an indicator of quality and affect mate preference in rock sparrows. Behavioral Ecology and Sociobiology, 65, 655 664. Hadfield, J.D. & Owens, I.P.F. (26) Strong environmental determination of a carotenoid-based plumage trait is not mediated by carotenoid availability. Journal of Evolutionary Biology, 19, 114 1114. Hadfield, J.D., Burgess, M.D., Lord, A., Phillimore, A.B., Clegg, S.M. & Owens, I.P.F. (26) Direct versus indirect sexual selection: genetic basis of colour, size and recruitment in a wild bird. Proceedings of the Royal Society of London, Series B, 273, 1347 1353. Hadfield, J.D., Nutall, A., Osorio, D. & Owens, I.P.F. (27) Testing the phenotypic gambit: phenotypic, genetic and environmental correlations of colour. Journal of Evolutionary Biology, 2, 549 557. Hill, G.E. (1992) Proximate basis of variation in carotenoid pigmentation in male house finches. Auk, 19, 1 12. Hill, G.E., Inouye, C.Y. & Montgomerie, R. (22) Dietary carotenoids predict plumage coloration in wild house finches. Proceedings of the Royal Society, Series B: Biological Sciences, 269, 1119 1124. Hofmann, C.M., Cronin, T.W. & Omland, K.E. (26) Using spectral data to reconstruct evolutionary changes in coloration: carotenoid color evolution in New World orioles. Evolution, 6, 168 1691. Hunt, S., Bennett, A.T.D., Cuthill, I.C. & Griffiths, R. (1998) Blue tits are ultraviolet tits. Proceedings of the Royal Society of London, Series B, 265, 451 455. Isaksson, C., Ornborg, J., Prager, M. & Andersson, S. (28) Sex and age differences in reflectance and biochemistry of carotenoid-based colour variation in the great tit Parus major. Biological Journal of the Linnean Society, 95, 758 765. Jacot, A., Romero-Diaz, C., Tschirren, B., Richner, H. & Fitze, P.S. (21) Dissecting carotenoid from structural components of carotenoid-based coloration: a field experiment with great tits (Parus major). American Naturalist, 176, 55 62. Jenni, L. & Winkler, R. (1994) Moult and Ageing of European Passerines. Academic Press, London. Johnsen, A., Delhey, K., Andersson, S. & Kempenaers, B. (23) Plumage colour in nestling blue tits: sexual dichromatism, condition dependence and genetic effects. Proceedings of the Royal Society of London, Series B, 27, 1263 127. Kelber, A. & Osorio, D. (21) From spectral information to animal colour vision: experiments and concepts. Proceedings of the Royal Society of London, Series B, 277, 1617 1625. Kruuk, L.E.B. (24) Estimating genetic parameters in natural populations using the animal model. Philosophical Transactions of the Royal Society B: Biological Sciences, 359, 873 89. Kruuk, L.E.B., Slate, J. & Wilson, A.J. (28) New answers for old questions: the evolutionary quantitative genetics of wild animal populations. Annual Review of Ecology, Evolution, and Systematics, 39, 525 548. MacDougall, A.K. & Montgomerie, R. (23) Assortative mating by carotenoid-based plumage colour: a quality indicator in American goldfinches, Carduelis tristis. Naturwissenschaften, 9, 464 467. Mahler, B., Araujo, L.S. & Tubaro, P.L. (23) Dietary and sexual correlates of carotenoid pigment expression in dove plumage. Condor, 15, 258 267. Masello, J.F., Lubjuhn, T. & Quillfeldt, P. (28) Is the structural and psittacofulvin-based coloration of wild burrowing parrots Cyanoliseus patagonus condition dependent? Journal of Avian Biology, 39, 653 662. Matrkova, J. & Remes, V. (212) Environmental and genetic effects on pigment-based vs. structural component of yellow feather colouration. PLoS One, 7, e3664. McGraw, K.J. (26) Mechanics of carotenoid-based coloration. Bird Coloration. Volume I: Mechanisms and Measurements (eds G.E. Hill & K.J. McGraw), pp. 177 242. Harvard University Press, Cambridge, MA. McGraw, K.J. & Hill, G.E. (21) Carotenoid access and intraspecific variation in plumage pigmentation in male American Goldfinches (Carduelis tristis) and Northern Cardinals (Cardinalis cardinalis). Functional Ecology, 15, 732 739. Montgomerie, R. (26) Analyzing colors. Bird Coloration. Volume I: Mechanisms and Measurements (eds G.E. Hill & K.J. McGraw), pp. 9 147. Harvard University Press, Cambridge, MA. Mundy, N.I. (26) Genetic basis of color variation in wild birds. Bird Coloration. Volume I. Mechanisms and Measurements (eds G.E. Hill & K.J. McGraw), pp. 469 56. Harvard University Press, Cambridge, MA. Parker, R.S. (1996) Absorption, metabolism, and transport of carotenoids. FASEB Journal, 1, 542 551. Patrick, S.C., Chapman, J.R., Dugdale, H.L., Quinn, J.L. & Sheldon, B.C. (212) Promiscuity, paternity and personality in the great tit. Proceedings of the Royal Society, Series B: Biological Sciences, 279, 1724 173. Peters, A., Denk, A.G., Delhey, K. & Kempenaers, B. (24) Carotenoidbased bill colour as an indicator of immunocompetence and sperm performance in male mallards. Journal of Evolutionary Biology, 17, 1111 112. Ruban, A.V., Horton, P. & Young, A.J. (1993) Aggregation of higher plant xanthophylls: differences in absorption spectra and in the dependency on solvent polarity. Journal of Photochemistry and Photobiology: Biology, 21, 229 234. Self, S.G. & Liang, K.-Y. (1987) Asymptotic properties of maximum likelihood estimators and likelihood ratio tests under nonstandard conditions. Journal of the American Statistical Association, 82, 65 61. Senar, J.C., Negro, J.J., Quesada, J., Ruiz, I. & Garrido, J. (28) Two pieces of information in a single trait? The yellow breast of the great tit (Parus major) reflects both pigment acquisition and body condition. Behaviour, 145, 1195 121. Shawkey, M.D. & Hill, G.E. (25) Carotenoids need structural colours to shine. Biology Letters, 1, 121 124. Shawkey, M.D., Hill, G.E., McGraw, K.J., Hood, W.R. & Huggins, K. (26) An experimental test of the contributions and condition

Colour in a new light 13 dependence of microstructure in yellow plumage colouration. Proceedings of the Royal Society of London, Series B, 273, 2985 2991. Stoddard, M.C. & Prum, R.O. (28) Evolution of avian plumage color in a tetrahedral color space: a phylogenetic analysis of New World buntings. American Naturalist, 171, 755 776. Svensson, L. (1994) Identification Guide to European Passerines, 4th edn. Svensson, Stockhom, Sweden. Utz, M., Jeschke, J.M., Loeschcke, V. & Gabriel, W. (214) Phenotypic plasticity with instantaneous but delayed switches. Journal of Theoretical Biology, 34, 6 72. Wallace, A.R. (1881) Island Life: Or, the Phenomena and Cause of Insular Faunas and Floras. Harper & Brothers, New York. Walsh, N., Dale, J., McGraw, K.J., Pointer, M.A. & Mundy, N.I. (212) Candidate genes for carotenoid coloration in vertebrates and their expression profiles in the carotenoid-containing plumage and bill of a wild bird. Proceedings of the Royal Society of London, Series B: Biological Sciences, 279, 58 66. Zsceile, F.P., White, J.W., Beadle, B.W. & Roach, J.R. (1942) The preparation and absorption spectra of five pure carotenoid pigments. Plant Physiology, 17, 331 346. Received 23 December 213; accepted 2 May 214 Handling Editor: Kevin McGraw