No evidence for general condition-dependence of structural plumage colour in blue tits: an experiment

Abstract

Condition-dependence is a central but contentious tenet of evolutionary theories on the maintenance of ornamental traits, and this is particularly true for structural plumage colour. By providing diets of different nutritional quality to moulting male and female blue tits, we experimentally manipulated general condition within the natural range, avoiding deprivation or stressful treatments. We measured reflectance of the structural-coloured UV/blue crown, a sexually selected trait in males, and the white cheek, a nonpigmented structural colour, directly after moult and again during the following spring mating season. We employed a variety of colour indices, based on spectral shape and avian visual models but, despite significant variation in condition and coloration, found no evidence for condition-dependence of UV/blue or white plumage colour during either season. These and previously published results suggest that structural colour might be sensitive to stress, rather than reduced body condition, during moult.

Introduction

Condition-dependence is often presumed to be a general feature of sexual traits, males in superior condition finding the costs of elaboration of sexual signals easier to bear. This is supported by a relatively large number of studies demonstrating positive relationships between ornaments or courtship and some aspects of condition (reviewed in Johnstone, 1995). However, the vast majority of these studies are correlational, and a direct link between individual condition and trait expression has not been unambiguously demonstrated. Importantly, as was noted by Cotton et al. (2004), because many nonsexually selected traits also show condition-dependence, what needs to be shown is heightened condition-dependence of sexual traits compared to suitable control traits. Such control traits can be corresponding nonsexually selected traits, or the homologous trait in females, that presumably reflects the unexaggerated state. However, despite the common claim that ornaments are condition-dependent, good support from well-designed experiments is rare (Cotton et al., 2004). Moreover, the notion of heightened condition-dependence of sexually selected traits was recently challenged by Johnstone et al. (2009). They modelled whether ornamental traits as they undergo exaggeration inevitably become more sensitive to condition. Their results showed that, depending on the precise form of the relationship between cost of trait exaggeration and trait expression, as ornaments become more exaggerated, condition-dependence may increase, decrease or remain unchanged. Currently, therefore, the question of the expected condition-dependence of ornamental traits remains unanswered. This is particularly true for ornamental coloration in birds that is not predominantly or exclusively pigment-based, the structural colours (Hill, 2006).

Brightly coloured plumages feature prominently in sexual selection research, both at ultimate and at proximate levels (review Hill & McGraw, 2006a,b). Two basic mechanisms produce coloured feathers: deposition of pigments, the commonest of which are carotenoids and melanins, and regularly arranged nanostructures within the keratin matrix that coherently scatter incident light and reflect specific wavelengths (Prum, 2006). Condition-dependence of pigment-based ornamental colours is fairly well supported theoretically, as well as correlationally and experimentally (Hill, 2006; McGraw, 2008; Peters et al., 2008; Catoni et al., 2009). Structural coloration, in contrast to pigment-based colours, is not based on the use of specific valuable and/or limited nutrients. This, in combination with the self-assembly inherent in colour-producing keratin nanostructures, provides no strong theoretical expectations of condition-dependence of structural coloration (Prum et al., 2009). Nonetheless, several authors have argued that depressed nutritional condition may compromise the development of feather nanostructure (Keyser & Hill, 1999; McGraw et al., 2002; Hill, 2006), thereby producing less regular, less precisely aligned scattering structures and plumage colour with lower chroma or saturation (Shawkey et al., 2003). Indeed, a few experimental studies resulted in the production of deteriorated structural colours by manipulations involving food withdrawal (McGraw et al., 2002; Siefferman & Hill, 2005), parasite infection (Hill et al., 2005) and reduced moult speed (Griggio et al., 2009). However, these experiments involved rather severe and/or stressful protocols and did not manipulate body condition per se (available resources/reserves, typically expressed as body mass corrected for structural size, for reviews see for instance Green, 2001; Peig & Green, 2010). Other studies were correlational and did not control for potential confounding factors (review by Prum, 2006). Thus, convincing demonstrations of condition-dependence of structural colours remain largely lacking, and controlled experimental investigations of moulting birds are needed (Hill, 2006).

We experimentally examined condition-dependence of a structural plumage ornament, the UV/blue crown colour of the blue tit, Cyanistes caeruleus. The crown is sexually dichromatic and a large body of evidence indicates that male crown coloration is a sexually selected trait (review in Delhey & Peters, 2008). Blue tit crown coloration is produced through coherent scattering by quasi-ordered keratin microtubules (Prum, 2006). Nanostructural variation in such an arrangement has been shown to affect colour chroma and saturation in eastern bluebirds (Sialia sialis, Shawkey et al., 2003). Additionally, we assessed the effect of our experimental manipulation on the reflectance of the white cheek patch, which is also sexually dichromatic, but less strongly so (Delhey & Peters, 2008; Griggio et al., 2009). White is also a structural colour but it is produced through incoherent scattering of light by unpigmented feather keratin (Prum, 2006). As the mechanism of colour production here does not depend on the regularity of the feather microstructure, one might argue that the white cheek colour should be less sensitive to variation in condition (see also Griggio et al., 2009).

We compared condition-dependence of UV/blue and white plumage colour in yearling males undergoing the first moult into adult plumage, using the homologous female traits as a control, expecting that females show reduced condition-dependence (following Cotton et al., 2004). We specifically aimed to avoid severe deprivation or stressful conditions and to remain largely within the natural range of relevant parameters (condition, plumage colour). Therefore, we kept birds in large naturalistic individual aviaries and manipulated condition through diet quality, providing a minimalistic but adequate standard diet and an optimized, enhanced diet. We determined the effect of this manipulation on plumage reflectance after the moult was completed. We measured plumage reflectance of these birds again the following breeding season, because plumage, particularly structural plumage, can show large and significant seasonal variation in colour (Delhey et al., 2010a). As different aspects of structural colour are not equally likely to be subject to condition-dependence (Prum, 2006), we included a variety of traditional colour indices including brightness, chroma and hue as well as indices derived from our current understanding of avian colour perception ability. Despite significant variation in condition and coloration, we found no evidence for condition-dependence of UV/blue or white plumage colour.

Methods

Model species

The blue tit is a well-known model organism for studies of sexual selection, particularly with reference to structural plumage coloration. Although both sexes in blue tits develop the UV/blue head plumage, in males reflectance of the crown peaks much further into the UV (for reflectance spectra, see Delhey et al., 2010a). The coloration most likely derives from coherent scattering of short wavelengths by the quasi-ordered array of keratin in the spongy layer of the feather barbs as has been shown for several other species with similar reflectance spectra (Prum, 2006). From experimental and correlational studies, it is known that male crown coloration is important for female reproductive investments, including mate choice (Hunt et al., 1999; Kurvers et al., 2010), extra-pair paternity (Delhey et al., 2003, 2007a), brood sex ratio (e.g. Sheldon et al., 1999; Korsten, 2006; Delhey et al., 2007b; but see Dreiss et al., 2006) and nestling provisioning (Limbourg et al., 2004; Johnsen et al., 2005), as well as in male agonistic and social interactions (Alonso-Alvarez et al., 2004; Rémy et al., 2010; Vedder et al., 2010). Crown colour is related to male heterozygosity (Foerster et al., 2003), and it is mediated by testosterone (Peters et al., 2006; Roberts et al., 2009) and moult speed (Griggio et al., 2009). Neither genetic effects nor conditions during early development appear to affect crown UV/blue reflectance (Roberts et al., 2009), and crown colour is only weakly heritable in this species (Hadfield et al., 2006), indicating that there is a large potential for individual condition to affect its expression.

Study design

The experimental blue tits were raised and housed in large naturalistic individual aviaries (300 × 300 × 190 cm, with some shrubs, perches and one nest box). We collected blue tit families (complete broods with their parents and nest box) during May 2006 from a population in south-west Germany (47°45′N, 8°59′E). Each brood was raised in one aviary by their parents until the young could forage independently. Then, 48 experimental birds, two males and two females from 12 families, were placed separately in individual aviaries for the remainder of the experiment. One male and one female of each family were randomly assigned to a standard or an enhanced diet, respectively, that started 15 July, before the start of the post-juvenile moult. We regularly inspected birds to monitor the progress of moult; moult speed and timing of moult completion did not vary between the two diet treatments (Kurvers et al., 2008). Body mass was measured before the start of the experiment (19–21 June; tarsus length was also measured at this time), during early moult (31 July–1 August) and during mid-moult (23–25 August). Immediately after completion of the moult, when all feathers had completely emerged (26–30 October 2006), and during the following spring (26–27 March 2007), we measured plumage reflectance. For comparison with free-living blue tits captured in the local area, we measured body mass of moulting blue tits in August and September 2005–07 (n = 53) and plumage reflectance of freshly moulted juvenile birds during October 2005–07 (n = 26). For the purpose of a different experiment, all birds received control implants during moult (containing only inert carrier material) and were immunized with sheep red blood cells and phytohaemagglutinin during August and September 2006 (Roberts & Peters, 2009), and in February 2007, males received a temporary colour treatment of the crown that faded within hours (Roberts et al., 2009; Kurvers et al., 2010). In the course of the experiment, some birds died, escaped or became injured, and final sample sizes are presented in Figs 1 and 2.

Figure 1.  No evidence for an effect of diet quality on expression of the UV/blue crown colour of blue tits measured (a) immediately after moult (October 2006) and (b) during the following spring breeding season (March 2007). Shown are mean ± SE UV chroma for juvenile males and females that had been provided with a standard and enhanced diet during moult. Numbers refer to sample sizes.

Figure 2.  No evidence for an effect of diet quality on expression of the white cheek colour of blue tits measured (a) immediately after moult (October 2006) and (b) during the following spring (March 2007). Shown are mean ± SE brightness (R_avg) for juvenile males and females that had been provided with a standard and enhanced diet during moult. Numbers refer to sample sizes.

Dietary treatment

We manipulated bird condition through the provision of two specially formulated experimental semi-synthetic diets. The standard diet was formulated to be adequate, fulfilling minimal requirements only, but not deficient, whereas the enhanced diet was formulated to be more easily digestible (lower fibre content, more water) and more nutritious, with higher protein, vitamin and antioxidant (carotenoid) content. Protein is important because feathers consist largely of keratin, a highly stable, insoluble protein that is synthesized from amino acids during feather growth (Brush, 1983), and during moult, whole-body protein turnover is highly accelerated to ensure a continuous supply of amino acids as feather production progresses (Murphy & Taruscio, 1995). Antioxidants and vitamins could alleviate the oxidative stress (Catoni et al., 2008) associated with the large increase in metabolic rate during moult, one of the most energy demanding processes in small birds (Dietz et al., 1992; Lindström et al., 1993; Klaassen, 1995). The standard diet consisted of 20% protein, 0.4% vitamin, 42% carbohydrate and 20% fat and the enhanced diet contained 41% protein, 0.05% lutein (FloraGlo, Pfannenschmidt), 2% vitamin, 15% carbohydrate and 15% fat (for further details on diets, see Kurvers et al., 2008). Prior to the experimental diets, birds received a varied diet of several types of adult invertebrates, invertebrate larvae and egg food, a mix of mashed boiled hens’ eggs, crushed rusk and soured milk with added vitamins. We did not provide the experimental diets over winter, and from November 2006 onwards, all birds received a similar winter diet consisting of live mealworms, fat balls and sunflower seeds.

Measurement and analysis of reflectance spectra

From five different but standardized spots (each around 11 mm²) of the crown and the cheek, reflectance spectra between 300 and 700 nm were obtained (for typical reflectance spectra, see Delhey & Peters, 2008; Delhey et al., 2010a,b). We used an Avaspec 2048 spectrometer, a DH-S light source and a bifurcated fibre optic probe (all Avantes, Eerbeek, The Netherlands) placed perpendicular to the surface, with a black plastic cylinder at the tip to standardize measuring distance and exclude ambient light. Reflectance (R) was computed relative to a WS-2 white standard using the program Avasoft 6.2.1. We computed brightness or average reflectance R_avg, hue (λ_max; crown only) and UV chroma (R_300–400/R_300–700), descriptors used in numerous previous studies of structural colours, including blue tits (Delhey et al., 2007a,b). Although these measurements are useful to uncover patterns of colour variation, it is unclear how they relate to colour perception by birds. Therefore, we additionally applied current models of avian colour vision physiology. We computed cone quantum catches, using formula 1 of Vorobyev et al. (1998) as described in detail by Delhey et al. (2010a) using generalized cone sensitivity for U-type birds from Endler & Mielke (2005) and standard daylight (D65) as illuminant. Cone quantum catches can be transformed into three independent variables, x, y, z (following Kelber et al., 2003), that define for each reflectance spectrum a position in the three-dimensional visual colour space of birds. The distance between two points in this visual space corresponds to the chromatic difference between them. In our analyses, higher values of x represent greater stimulation of the long-wavelength-sensitive cone and lower stimulation of the medium-wavelength-sensitive cone, higher y values represent greater stimulation of the short-wavelength-sensitive cone, and higher values of z represent greater stimulation of the very short-wavelength-sensitive cone. Xyz coordinates are usually correlated and measurements of a particular colour type (e.g. UV/blue crown) form discrete cigar-shaped clouds in the visual space of birds that usually have one clear axis of variation (Endler et al., 2005; Delhey et al., 2010b). We expressed the variation in reflectance for each colour (crown/check in October and March) in units that provide information about the magnitude of perceptual differences (i.e. whether a particular chromatic difference between treatments is discriminable by the birds). Following the methods by Delhey et al. (2010a) for each set of x, y and z, we calculated the first principal component (PC1) that explained 94.8% (October) and 94.0% (March) of chromatic variation for the crown and 89.2% (October) and 98.1% (March) of chromatic variation for the cheek. We then used PC1 scores to select, separately for each plumage patch and period, the measurement with the lowest value of PC1 and used formulas 2, 3 and 8 by Vorobyev et al. (1998) to compute chromatic contrast (ΔS, also called discriminability) between this point and all other points in the sample using a Weber fraction of 0.05 and relative abundance of single cones as measured in the blue tit (Hart et al., 2000). ΔS is measured in just noticeable differences (jnd), and values of contrast below 1 jnd are usually considered not discriminable by birds (Vorobyev et al., 1998). ΔS is thus a measure of discriminable variation within a sample, and in our data set, it correlates strongly with UV chroma (October, crown: R² = 0.91, P < 0.0001, ΔS = −22.6 + 70.1 × UVC; cheek: R² = 0.50, P < 0.0001, ΔS = −7.0 + 35.3 × UVC; March, crown: R² = 0.80, P < 0.0001; ΔS = −24.0 + 81.8 × UVC; cheek: R² = 0.79, P < 0.0001, ΔS = −13.5 + 72.9 × UVC) and thus provides partly similar information but has the advantage to be expressed in units that are relevant to the visual perception of birds.

Statistical analyses

Body condition is generally expressed as a function of body mass and structural size, the most widely applied nondestructive index of energy reserves (Green, 2001; Peig & Green, 2010), and we used the well-established method in avian ecology of correcting for structural size by including tarsus length as a covariate in our analyses (Darlington & Smulders, 2001; Green, 2001). Tarsus length did not vary between diet groups (F_1,31.5 = 0.04, P = 0.84) but males were larger than females (F_1,32.5 = 25.7, P < 0.001; controlling for ‘family’ as a random term). Not including tarsus length as a covariate did not qualitatively alter the results (data not shown). We used the restricted maximum likelihood procedure to test for differences between treatment groups in body condition and plumage colour. The initial models consisted of diet, sex and their interaction as the fixed effects, with bird ‘family’ (nest box of origin) as the random term. To investigate relationships between individual condition and coloration, we constructed similar mixed models, with family as random effect and body mass during early moult as fixed effect, controlling for structural size (tarsus length) and sex (the interaction between sex and body mass was nonsignificant for all colour variables; using body mass during mid-moult as fixed effect yielded similar results, data not shown). Fixed models were stepwise reduced until only significant terms remained in the final model. Statistical details for excluded terms were based on their re-introduction in the final models. The random term ‘family’ was conservatively always retained, although no significant family effects on coloration were evident (cf. Roberts et al., 2009): for all colour parameters in spring and autumn (cf. Tables 1 and 2), only 1 in 26 P-values was below 0.05 and all others ranged from 0.21 to 0.93 (mean = 0.59, SD = 0.24). We corrected for multiple testing by implementing the false discovery rate procedure (Benjamini & Hochberg, 1995; Peters et al., 2007) defining the statistical tests for each plumage patch and season as one family of tests. We generated an estimate of the sample sizes that would have been required to render the observed mean differences significant at α = 0.05 by applying a power analysis (separately for each sex) on the effect of diet on all colour parameters (as listed in Tables 1 and 2). All data were analysed using JMP^®7.0 (SAS Institute Inc., Cary, NC, USA).

Table 1.   No effect of dietary manipulation during moult on reflectance of the crown and cheek in male and female blue tits measured just after moult completion (October 2006). Shown are statistical details of diet (standard, enhanced), sex and their interaction, controlling for family of origin in mixed models. Values indicated in bold are statistically significantly different after FDR correction for multiple testing.

	Diet			Mean ± SE*		Sex			Mean ± SE*		Sex × Diet
	F	d.f.den	P	Standard	Enhanced	F	d.f.den	P	Female	Male	F	d.f.den	P
x, y and z are the relative cone quantum catches. ΔS is the chromatic distance between all individuals in the sample and the individual with the least shortwave-rich reflectance spectrum as indicated by the lowest PC1 score. UVC, UV chroma; Ravg, average reflectance (‘brightness’); λmax, wavelength of maximum reflectance (‘hue’); FDR, false discovery rate. *Model predictions.
Crown
UVC	0.17	32.1	0.68	0.369 ± 0.003	0.371 ± 0.003	65.1	33.2	< 0.0001	0.353 ± 0.003	0.387 ± 0.003	0.11	31.2	0.75
Ravg	0.09	32.4	0.77	18.8 ± 0.5	19.0 ± 0.5	15.0	33.4	< 0.0001	17.6 ± 0.5	20.0 ± 0.5	2.02	31.4	0.16
λmax	0.25	32.7	0.62	339.6 ± 1.6	338.4 ± 1.6	9.94	33.6	0.003	342.6 ± 1.6	335.4 ± 1.6	0.09	31.7	0.77
x	0.00	33.4	0.95	−0.037 ± 0.001	−0.037 ± 0.001	0.33	33.4	0.57	−0.037 ± 0.001	−0.038 ± 0.001	0.58	31.4	0.45
y	0.04	32.3	0.85	0.096 ± 0.003	0.095 ± 0.003	44.2	33.3	< 0.0001	0.082 ± 0.003	0.108 ± 0.003	0.08	31.3	0.78
z	0.59	32.0	0.45	−0.146 ± 0.002	−0.147 ± 0.002	66.2	33.0	< 0.0001	−0.155 ± 0.002	−0.138 ± 0.002	0.06	31.0	0.80
ΔS	0.25	32.1	0.62	3.26 ± 0.27	3.42 ± 0.27	57.5	33.1	< 0.0001	2.16 ± 0.27	4.51 ± 0.27	0.05	31.2	0.83
Cheek
UVC	0.18	32.7	0.67	0.234 ± 0.002	0.234 ± 0.002	4.19	32.3	0.05	0.231 ± 0.002	0.236 ± 0.002	1.35	30.2	0.25
Ravg	1.15	33.8	0.29	57.3 ± 1.3	55.4 ± 1.3	0.54	33.9	0.47	55.7 ± 1.3	57.0 ± 1.3	0.32	32.0	0.57
x	3.25	32.1	0.08	0.0042 ± 0.0003	0.0049 ± 0.0003	10.1	32.4	0.003	0.0053 ± 0.0003	0.0038 ± 0.0003	0.00	31.1	0.98
y	5.26	30.6	0.03	−0.029 ± 0.001	−0.033 ± 0.001	0.53	30.6	0.47	−0.032 ± 0.001	−0.031 ± 0.001	4.58	28.4	0.04
z	0.18	31.4	0.67	−0.200 ± 0.001	−0.201 ± 0.001	3.94	31.3	0.05	−0.201 ± 0.001	−0.199 ± 0.001	2.90	29.1	0.10
ΔS	2.63	29.5	0.12	1.29 ± 0.09	1.07 ± 0.09	0.97	29.2	0.33	1.11 ± 0.09	1.25 ± 0.09	32.3	27.9	0.07

Table 2.   No effect of dietary manipulation during moult on crown and cheek reflectance in male and female blue tits, measured the following spring (March 2007). Shown are statistical details of diet (standard, enhanced), sex and their interaction, controlling for family of origin in mixed models. Values indicated in bold are statistically significantly different after FDR correction for multiple testing.

	Diet			Mean ± SE*		Sex			Mean ± SE*		Sex × Diet
	F	d.f.den	P	Standard	Enhanced	F	d.f.den	P	Female	Male	F	d.f.den	P
x, y and z are the relative cone quantum catches. ΔS is the chromatic distance between all individuals in the sample and the individual with the least shortwave-rich reflectance spectrum as indicated by the lowest PC1 score. UVC, UV chroma; Ravg, average reflectance (‘brightness’); λmax, wavelength of maximum reflectance (‘hue’); FDR, false discovery rate. *Model predictions.
Crown
UVC	0.75	27.1	0.39	0.349 ± 0.003	0.344 ± 0.003	43.5	29.7	< 0.0001	0.331 ± 0.003	0.363 ± 0.003	0.12	27.8	0.73
Ravg	0.62	25.4	0.44	20.9 ± 0.8	19.8 ± 0.9	6.07	28.3	0.02	18.8 ± 0.8	22.0 ± 0.9	0.16	26.3	0.69
λmax	0.0	25.3	0.92	354.1 ± 22	353.9 ± 2.3	9.44	27.3	0.005	357.6 ± 2.2	350.4 ± 2.3	0.13	25.2	0.72
x	4.31	26.5	0.05	−0.038 ± 0.001	−0.034 ± 0.002	0.41	28.3	0.53	−0.035 ± 0.001	−0.037 ± 0.002	0.89	26.6	0.35
y	2.01	26.7	0.17	0.094 ± 0.004	0.087 ± 0.004	46.7	29.0	< 0.0001	0.072 ± 0.004	0.109 ± 0.004	0.36	27.0	0.36
z	1.07	28.0	0.31	−0.152 ± 0.002	−0.155 ± 0.002	47.0	30.3	< 0.0001	−0.162 ± 0.002	−0.144 ± 0.002	0.14	28.4	0.71
ΔS	2.00	27.3	0.17	4.71 ± 0.30	4.10 ± 0.32	42.5	29.6	< 0.0001	3.00 ± 0.31	5.85 ± 0.33	0.27	27.7	0.61
Cheek
UVC	0.84	28.4	0.37	0.209 ± 0.002	0.212 ± 0.002	15.0	30.8	0.0005	0.205 ± 0.002	0.216 ± 0.002	0.09	28.7	0.77
Ravg	0.63	26.8	0.43	57.4 ± 1.6	55.8 ± 1.7	0.33	28.2	0.57	56.1 ± 1.7	57.3 ± 1.7	1.47	25.6	0.24
x	0.01	27.9	0.91	0.0034 ± 0.0006	0.0035 ± 0.0007	18.5	29.4	0.0002	0.0054 ± 0.0007	0.0014 ± 0.0007	0.01	28.3	0.92
y	1.38	26.8	0.25	−0.028 ± 0.002	−0.031 ± 0.002	18.6	28.3	0.0002	−0.035 ± 0.002	−0.024 ± 0.002	0.03	27.2	0.86
z	0.03	27.5	0.87	−0.205 ± 0.0.001	−0.205 ± 0.001	14.7	29.0	0.0006	−0.208 ± 0.001	−0.203 ± 0.001	0.04	27.9	0.84
ΔS	0.29	27.2	0.59	1.85 ± 0.17	1.73 ± 0.18	18.2	29.3	0.0002	1.29 ± 0.17	2.39 ± 0.18	0.01	27.7	0.91

Results

Condition

There was no difference in body mass (F_1,32.1 = 1.41, P = 0.24, correcting for sex) or body condition (mass correcting for tarsus length, F_1,32.0 = 0.55, P = 0.48) between experimental groups before the start of the experiment. During early moult, birds fed the enhanced diet were significantly heavier for their size (F_1,32.0 = 4.97, P = 0.03, controlling for tarsus length; there was no sex difference in condition, F_1,34.6 = 0.02, P = 0.88). Birds provided with the enhanced diet then weighed on average 11.7 ± 0.15 g (95% interval: 11.4–12.0 g) compared to 11.4 ± 0.16 g (95% interval: 11.1–11.8 g) for birds provided with the standard diet. During mid-moult, the effect of diet was similar (F_1,31.4 = 6.53, P = 0.016, controlling for tarsus length; there was no sex difference in condition, F_1,31.1 = 3.05, P = 0.09). Birds provided with the enhanced diet then weighed on average 12.0 ± 0.11 g (95% interval: 11.8–12.3 g) compared to 11.7 ± 0.16 g (95% interval: 11.4–12.1 g) for birds provided with the standard diet. Accordingly, birds provided with the enhanced diet gained more mass during moult (F_1,33.4 = 4.50, P = 0.04).

Free-living blue tits

During moult, free-living blue tits captured in the local area showed a similar but slightly narrower range in body mass (11.6 ± 0.11 g; 95% interval 10.1–13.0) as the captives. Birds on the enhanced diet at that time were significantly heavier (F_1,72 = 10.2, P = 0.002; controlling for sex) whereas birds on the standard diet did not differ significantly from the wild birds (F_1,73 = 1.45, P = 0.23; controlling for sex). All colour indices measured in October fell within the natural range for the corresponding variables in blue tits captured in the local area (data not shown). We compared mean UV chroma of the crown and brightness of the cheek, the most integrative of the colour parameters, between freshly moulted wild and captive juveniles and found no significant difference (crown UV chroma: F_1,68 = 3.00, P = 0.09; cheek brightness: F_1,65 = 0.02, P = 0.88; controlling for sex).

Crown colour

As expected, there were strong and highly significant differences in crown coloration between the sexes immediately after moult (Table 1, Fig. 1) as well as during the following spring (Table 2), with males presenting a more UV crown (higher UV chroma, smaller hue, greater y, z). Diet quality did not affect coloration of the crown for any of the colour descriptors during October (Table 1) as well as March (Table 2). Differences between treatment groups in predicted mean ΔS were well below 1 jnd in October as well as March. Power analysis indicated that a median of 318 (October, range 205–1430) or 361 (March, range 73–1999) males and 3186 (October, range 64–68307) or 65 (March, range 26–50509) females would have been required to render the observed differences between diet treatment groups significant at the 0.05 level. In accordance, individual condition during moult (body mass controlling for tarsus and controlling for sex) did not have an effect on the crown colour developed during that moult (Table 3, Fig. 3). Likewise, condition during moult did not affect crown colour during the following spring (data not shown).

Table 3.   No effect of individual condition during moult on post-moult crown and cheek coloration in male and female blue tits. Shown are statistical details of body mass during early moult controlling for structural size (tarsus length), sex and family of origin in mixed models.

	Crown				Cheek
	F	d.f.den	P	ß ± SE	F	d.f.den	P	ß ± SE
x, y and z are the relative cone quantum catches; ΔS is the chromatic distance between all individuals in the sample and the individual with the least shortwave-rich reflectance spectrum as indicated by the lowest PC1 score. β is the effect size. UVC, UV chroma; Ravg, average reflectance between 300 and 700 nm (‘brightness’); λmax, wavelength of maximum reflectance (‘hue’). *Multiplied by 1000 for presentation.
UVC	0.96	22.6	0.34	−4.2 ± 4.3*	0.12	24.1	0.73	0.9 ± 2.6*
Ravg	1.92	24.3	0.18	0.91 ± 0.66	1.52	22.5	0.23	−2.13 ± 1.72
λmax	0.02	17.8	0.90	0.26 ± 2.06
x	0.07	28.5	0.79	−0.39 ± 1.45*	1.16	25.5	0.29	0.48 ± 0.45*
y	0.70	29.8	0.41	−3.7 ± 4.4*	0.28	3.9	0.63	−0.79 ± 1.5*
z	1.12	24.3	0.30	−2.4 ± 2.2*	0.00	21.1	0.95	0.00 ± 0.00*
ΔS	0.95	28.7	0.34	−0.34 ± 0.35	0.30	11.5	0.59	−0.07 ± 0.13

Figure 3.  No evidence for an effect of condition during early moult on expression of the UV/blue crown colour of blue tits measured immediately after moult (October 2006). Shown are observed values of UV chroma and body mass for (a) males and (b) females. Lines are predicted by mixed models controlling for tarsus length, sex and family (random effect, see Table 3).

Cheek colour

There were sex differences in cheek colour immediately after moult, but these were not quite significant (Table 1). In spring, however, cheek reflectance was significantly sexually dichromatic (Table 2). Moult diet did not affect reflectance of the cheek for any of the colour descriptors during October (Table 1) as well as March (Table 2). Differences between treatment groups in predicted mean ΔS were well below 1 jnd in October as well as in March. Power analysis indicated that a median of 87 (October, range 205–1430) or 2129 (March, range 73–1999) males and 220 (October, range 50–42156) or 367 (March, range 44–296104) females would have been required to render the observed differences between diet treatment groups significant at the 0.05 level. Individual condition during moult did not have any effect on reflectance of the cheek plumage during that moult, measured directly after moult (Table 3), or during the following spring (data not shown).

Discussion

Experimental manipulation of condition did not affect the reflectance of the UV/blue crown and the white cheek of wild-caught yearling male and female blue tits. Our diet treatments generated significant variation in condition and plumage coloration, and we examined plumage reflectance not only directly after moult but also during the following mating season, when ornament variation might be most closely related to fitness. We expressly aimed to manipulate condition within or just beyond the natural range, avoiding stressful conditions or harsh deprivations, because extreme treatments provide little information on the evolutionary relevance of the natural connection between individual condition and ornament elaboration. By employing a comprehensive selection of the latest methods of analysis of reflectance spectra, we could demonstrate no significant effect of variation in condition on structural white or UV/blue plumage colour. These results were similar at the diet-treatment level as well as the individual level. The lack of significance does not seem to simply relate to insufficient sample size: during autumn as well as spring, for crown as well as cheek plumage, the mean differences in reflectance between treatment groups fell far short of 1 jnd (just noticeable difference), the threshold below which discrimination between two visual stimuli is no longer theoretically possible. Moreover, the observed effect was positive for post-moult colour only and negative for breeding season colour (compare Fig. 1a, b). Finally, similar sample sizes could demonstrate a significant experimental effect of moult testosterone levels (Roberts et al., 2009) and moult duration (Griggio et al., 2009) on blue tit crown reflectance, suggesting that effects of general body condition on colour – if they exist – are much smaller than those of testosterone or moult duration.

Our diet manipulation generated a relevant, though not extreme, range of body condition during moult. The diets resulted in consistent differences in body mass, including (and slightly surpassing) the entire natural range of moulting juvenile blue tits. Birds fed the enhanced diet were relatively heavier for their size, with an average difference between diet groups of 39% of the standard deviation in body mass. Furthermore, previous research using the same diets demonstrated that they resulted in varied differences in health status and important aspects of overall condition: blue tits fed the enhanced diet were heavier and fatter and produced a greater humoral immune response (Kurvers et al., 2008; Roberts & Peters, 2009), whereas closely related great tits (Parus major) fed the enhanced diet were heavier, had higher haematocrit and were better able to maintain condition during an immune challenge (Peters et al., 2011). We avoided any concomitant confounding effects of stress or other physiological effects associated with treatments such as restricted access to food or parasite infestation. Moreover, the experimental birds developed a natural-looking crown and cheek plumage, with reflectance parameters spanning the natural range, and our experiment confirmed patterns of sexual dichromatism and seasonal changes in plumage reflectance previously identified in free-living blue tits (compare figures 1 and 2 with Delhey et al., 2010a). We therefore conclude that we successfully created treatment groups that differed in body condition in a physiologically realistic and relevant manner – consequently, we are confident that our experiment was a robust test of the hypothesis that white and/or UV/blue structural colour in the blue tit is condition-dependent. Our results provide no evidence that this is the case.

Evidence for a causal link between general condition during moult and resultant structural coloration remains elusive, as most studies are correlational and/or inferred condition during moult indirectly/retrospectively. Correlations between colour during breeding and an index of body condition have been demonstrated for example in blue-tailed bee-eaters (Merops philippinus, Siefferman et al., 2007) and western bluebirds (Sialia mexicana, Budden & Dickinson, 2009), whereas in satin bowerbirds (Ptilonorhynchus violaceous), iridescent structural plumage coloration is negatively related to blood parasites (Doucet & Montgomerie, 2003a,b). Although such results indicate that plumage colour can signal current condition or parasite burden, it does not constitute evidence that plumage colour is determined by condition during moult. Some studies have therefore related plumage colour to feather growth bars, whereby larger bars indicate faster feather growth rate, and thereby presumably better condition, during moult (Grubb, 1989). Larger feather growth bars have been shown to correlate with structural blue plumage of male blue grosbeaks (Guiraca caerulea, Keyser & Hill, 1999) and blue-black grassquits (Volatinia jacarina, Doucet, 2002). This contrasts with an experimental study in blue tits, where a (photoperiodically induced) faster moult resulted in a severe decline in UV reflectance of the crown (Griggio et al., 2009). Body condition was not affected by the photoperiod treatment, and although feather growth bars were not reported, this experiment challenges the interpretation of the correlational studies. Alternatively, the experimental abrupt change in photoperiod in the blue tits (Griggio et al., 2009) could have caused stress and this affected plumage colour. The same is true for the most extensive experimental studies to date of condition-dependence of structural plumage. Here, male brown-headed cowbirds (Molothrus ater, McGraw et al., 2002) and female eastern bluebirds (Siefferman & Hill, 2005) in moult were experimentally exposed to nutritional stress, which consisted of random, temporary (6 h) complete food withdrawal. This treatment resulted in less saturated iridescent (cowbirds) and structural blue (bluebirds) plumage. However, the experimental regime did not result in reduced body mass. Thus, the question remains to what extent the observed effects on plumage colour were related to condition rather than, for example, stress responses associated with unpredictable food withdrawal for half a day at a time. The same applies to experimental evidence that severe infection with coccidial parasites can reduce UV reflectance of iridescent feathers in wild turkeys (Meleagris gallopavo, Hill et al., 2005). Nonetheless, these experimental studies show that some aspects of iridescent structural plumage can signal food stress or acute parasite infection during moult.

Whereas the anatomical details underlying chromatic structural colours of birds vary, all are produced by the same physical mechanism – coherent scattering (Prum, 2006). Structural white colour, on the other hand, is produced through incoherent scattering of all UV/visible wavelengths by the unordered matrix of keratin and air vacuoles (Prum, 2006). Because white feathers are unpigmented, and because incoherent scattering does not depend on the regularity of the feather microstructure, we hypothesized that the white cheek colour might be less sensitive to variation in condition. We found no evidence for condition-dependence of the white cheek colour, and likewise, Griggio et al. (2009) found that the dramatic experimentally increased moult speed that strongly affected the blue crown had no effect on reflectance of the white cheek. Similarly, Shawkey et al. (2006) in an experimental manipulation of carotenoid and food availability during moult found that none of the experimental treatments affected the expression of the structural white colour of depigmented feathers in American goldfinches (Carduelis tristis). Brood size manipulation affected structural white of depigmented yellow feathers in great tit nestlings, but the authors did not relate this to condition (Jacot et al., 2010). Although dark-eyed juncos (Junco hyemalis) fed a protein-enriched diet regrew plucked tail feathers that had brighter white patches controlling for their original colour, faster-growing feathers were duller (McGlothlin et al., 2007). On balance, although not conclusive, the available experimental evidence suggests that condition during moult does not influence reflectance of white plumage.

Structural plumage colour and stress?

Our results and the results of previous studies suggest that there is currently no convincing experimental evidence that general condition during moult directly affects structural plumage colour. Although lack of condition-dependence of ornaments is theoretically possible (Johnstone et al., 2009; Prum, 2010), it remains to be explained that structural colours show significant variation that is biologically relevant (see for example the review of functional significance of the UV/blue crown of the blue tit in the Methods section and Hill, 2006). Possibly, structural colour is not sensitive to general condition or protein content of the diet (as our manipulation), but to availability of specific amino acids (cf. Poston et al., 2005 for such an approach). Additionally or alternatively, some of these variations could be related to plumage maintenance, as structural colours may be particularly sensitive to soiling or abrasion (Fitzpatrick, 1998). Indeed blue tit crown colour shows a strong decline during late winter and spring that is individually highly variable (Delhey et al., 2006) and that can be counteracted by increased preening behaviour in spring, possibly mediated by testosterone levels (Roberts et al., 2009). However, such gradual mechanisms cannot explain differences in structural coloration immediately after moult.

Alternatively or additionally, structural colour expression may be sensitive to stress. Our experiment was designed to manipulate condition but impose the least stress possible. However, previous experiments, which impinged on structural colour but did not affect body mass, involved stressful protocols, such as food-deprivation stress (McGraw et al., 2002; Siefferman & Hill, 2005), time stress (Griggio et al., 2009), parasite-induced stress (Hill et al., 2005; Mougeot et al., 2010) or developmental stress (Jacot & Kempenaers, 2007; Siefferman & Hill, 2007). Feather structure appears stress-sensitive: keratin abnormalities are considered symptoms of stress (Jovani & Blas, 2004), and, for example, handling stress during moult can result in the production of fault bars, areas of keratin deficit (Murphy et al., 1988). In songbirds, both basal levels of stress hormone (corticosterone) and stress responsiveness are at their annual lowest during the period of post-breeding moult (Romero, 2002). This is deemed important to ensure that the protein catabolic effects of corticosterone do not interfere with protein metabolism required for feather production (Romero et al., 2005). Whereas only very severe protein malnutrition slows the rate of moult (Murphy et al., 1988; Murphy & King, 1991), short treatment with corticosterone can affect feather growth rate during induced and natural moult in adults (Romero et al., 2005) and in nestlings (Butler et al., 2010). We hypothesize that stress, possibly by interfering with the balance between protein synthesis and breakdown that is important for keratin production and deposition (Chilgren & DeGraw, 1977; Murphy & Taruscio, 1995; Taruscio & Murphy, 1995), can affect the structures responsible for structural colour. This could be tested by relating structural plumage elaboration to individual basal and stress-induced levels of corticosterone during moult (for example by examining feather corticosterone content Bortolotti et al., 2008) and by observing the effect on plumage coloration of treating moulting birds with corticosterone or repeated stressors while controlling for general condition. In line with recent demonstrations that stress hormones can affect melanin-based (Roulin et al., 2008) and carotenoid-based (Loiseau et al., 2008; Mougeot et al., 2010) coloration as well as mate choice (Husak & Moore, 2008), we propose that also structural plumage colour could serve as an indicator not only of stress experienced during moult but also of individual ability to effectively modulate the stress response (see also Almasi et al., 2010).

Acknowledgments

Removal of birds from the wild and animal experimental protocols was approved by the Regierungspräsidium Freiburg (Aktenzeichen 55-8852.15/05 and Registriernr. G-06/05, Aktenzeichen 35-9185.82/3/339, respectively). We are grateful to Scott McWilliams for help in designing the semi-synthetic diets. We thank the staff of the Vogelwarte Radolfzell for their help with bird capture, measurement and maintenance. Special thanks to Monika Krome and Evi Fricke for help with spectrometry and to Evi Fricke for processing of reflectance spectra and performing the molecular sexing. Michaela Hau commented on an earlier version of the manuscript. We thank Allen Moore and two anonymous reviewers for constructive comments that improved the manuscript. This study was funded by the ‘Minerva Sonderprogramm zur Förderung hervorragender Wissenschaftlerinnen’ of the Max Planck Society to AP.