Evaluating different metrics to study small color differences: the red bill and plumage of common waxbills as a case study

Animal coloration is a key biological trait in social communication and animal behavior research. Quantifying color differences among individuals using reflectance spectrophotometry has become standard practice but can be challenging due to measurement inaccuracies and color plasticity, especially when color differences are small. Choosing robust color metrics, among the many receiver-independent and visual-system-dependent metrics available, is thus an important consideration. Here, we measured reflectance spectra from carotenoid-based red bill and breast plumage of 18 females and 6 males common waxbills repeatedly throughout 1 year, computed alternative color metrics for each of three color properties (brightness, hue, and saturation), including 9 receiver-independent metrics and 4 metrics based on a tetrahedral model of avian vision, and also computed Principal Components (PCs) of spectral reflectance. Then, we estimated the adjusted repeatability of individual color metrics and correlations among metrics of the same color property for bill and plumage separately. Except for metrics based on PCs, all receiver-independent color metrics were strongly correlated to visual model metrics of the same color property. Individual differences in color were significantly repeatable across days, except when using PCs, but the repeatability estimates differed for alternative metrics of the same color property. Overall, visual model metrics were the most consistently repeatable in different tissues, while the repeatability of receiver-independent metrics differed depending on tissue and color property. Our results suggest that visual model metrics are appropriate to study subtle differences in animal coloration, perhaps even in the absence of detailed information on the performance of alternative color metrics. Significance statement Reflectance spectrophotometry is a standard method for objectively measuring animal coloration, and several alternative metrics are available to quantify color properties from reflectance spectra. However, we lack systematic testing of the reliability of different metrics to capture small color differences among individuals, despite measurement inaccuracies and a degree of color plasticity. We repeatedly measured the carotenoid-based red color in the bill and plumage of common waxbills throughout 1 year, to test which metrics captured individual differences in color more consistently. Metrics based on modeling the receiver visual system retrieved consistently repeatable individual differences in hue, saturation and brightness, in either tissue, while the repeatability of receiver-independent metrics often differed depending on the tissue and color property. Thus, compared to receiver-independent metrics, metrics based on visual models may generally be appropriate to detect and study subtle differences in carotenoid-based coloration.