Visual signal evolution along complementary color axes in four bird lineages

Animal color patterns function in varied behavioral contexts including recognition, camouflage and even thermoregulation. The diversity of visual signals may be constrained by various factors, for example, dietary factors, and the composition of ambient environmental light (sensory drive). How have high-contrast and diverse signals evolved within these constraints? In four bird lineages, we present evidence that plumage colors cluster along a line in tetrachromatic color space. Additionally, we present evidence that this line represents complementary colors, which are defined as opposite sides of a line passing through the achromatic point (putatively for higher chromatic contrast). Finally, we present evidence that interspecific color variation over at least some regions of the body is not constrained by phylogenetic relatedness. Thus, we hypothesize that species-specific plumage patterns within these bird lineages evolve by swapping the distributions of a complementary color pair (or dark and light patches in one group, putatively representing an achromatic complementary axis). The relative role of chromatic and achromatic contrasts in discrimination may depend on the environment that each species inhabits.

3 4 of the measured statistic, we compared this value to 1000 randomized values obtained 2 3 5 using the inbuilt functions of the phytools package. To further verify these results, we 2 3 6 additionally performed a second analysis. Using a phylogenetic distance matrix derived 2 3 7 from the ape package, we calculated Mantel correlations between this matrix and an 2 3 8 interspecific trait distance matrix derived for color and luminance for each body region  Pittas: PC1 (the major axis of variation) of the XYZ coordinates in color space explains 2 5 8 85% of chromatic variation ( Figure 2B). PC1 loads weakly negatively on X (-0.15), and 2 5 9 exhibits strong positive loadings (0.6 and 0.78) on Y and Z, respectively.

6 8
Across all four lineages, the Z coordinate loads most strongly on PC1, thus suggesting 2 6 9 that most variation in perceptual coordinate space occurs along the elevational rather 2 7 0 than azimuthal direction along the PC1 line. Therefore, in subsequent analyses, we used 2 7 1 the elevational coordinate Φ and the sign of this coordinate as an indicator of where 2 7 2 different colors lie along this line. Although this does not take variation in the azimuthal 2 7 3 plane into account, the results of our analysis suggest that this variation is negligible 2 7 4 compared to variation along the elevational axis in all four families. Thus, colors with 2 7 5 opposite signs of Φ in this dataset lie on opposite sides of the achromatic point (as is 2 7 6 evident from the spread of the data in Figure 2). We used PCA only to estimate the 2 7 7 proportion of variance along this line, and not in any subsequent analysis. between -1.54 and +1.57 across the family, i.e. on opposite sides of the achromatic point 2 9 0 and at roughly equal distances from it along the elevation axis, consistent with the 2 9 1 interpretation of a complementary color axis. For example, the deep-blue (to human 2 9 2 eyes) crown of the male Hydrornis baudii has, on average, a color score of -1.15, and 2 9 3 the deep-red crown of the sympatric(Erritzoe and Erritzoe 1998) Erythropitta granatina 2 9 4 scores +1.12. This is also consistent with a hypothesis of complementary colors, in that 2 9 5 these colors also represent opposite ends of the avian-visible light spectrum. Histograms relatively high CVs for color scores, but not body regions (except the wing, which does, 3 2 6 however, exhibit phylogenetic signal suggesting that this variation has a phylogenetic 3 2 7 component). Taken together, these results are also consistent with body colors being a 3 2 8 constrained feature within this lineage, but colors being swapped around on the head.

2 9
Parakeets: Color scores span between -1.47 and +1.57, again consistent with a  scores for each species lying on opposite sides of the achromatic point ( Figure 3C).

5
Neither color nor luminance scores exhibit significant phylogenetic signal across any 3 3 6 body regions (Table 1)  which is also consistent with signal diversification along a chromatic complementary 3 4 0 axis.

4 2
suggests that sandgrouse are clustered in chromatic space to one side of the achromatic 3 4 3 point, further supported by color histograms ( Figure 3D). However, aside from luminance 3 4 4 scores on the wing (Table 1) sandgrouse whose colors are found to only one side of the achromatic point.

7 3
Additionally, based on phylogenetic comparative analyses, we hypothesize that signals in luminance scores across body regions compared to color scores, leads us to 3 7 7 tentatively hypothesize that signal evolution in this family has occurred along an

9 2
We uncover evidence both that most species within a lineage possess complementary 3 9 3 colors in their plumage (Figure 3), and also that phylogenetic patterns of trait evolution 3 9 4 depart from Brownian motion over at least some body regions in all families (Table 1).

9 5
Taken together, this is consistent with complementary colors in their plumage, putatively 3 9 6 for high chromatic contrast (Endler 1992) under constraints, and with signal achromatic axis (or to changes in barring and speckling, which our study did not 4 3 0 investigate), albeit with the caveat that luminance variation is difficult to compare using 4 3 1 museum specimens. However, a comparison of plumage patterns in sandgrouse ( Figure   4 3 2 1) reveals that many species possess conspicuous black and white patches, whose predation these birds experience, and the light microhabitats they use.

6 5
Color vision is challenging to study comparatively in speciose bird lineages containing 4 6 6 rare or range-restricted species, and many of the species we examine are poorly known.

6 7
Thus, although our study design does not permit us to conclusively identify the 4 6 8 ecological driver of these patterns, we do find consistent evidence of overall patterns of  We are indebted to Helen James for providing access to the reflectance spectrometer