Quantitative characterization of melanin-based feather colour using reflectance spectrophotometry

Abstract Colour produced via melanin pigments is widespread in extant bird feathers and has been inferred in many fossil birds, feathered dinosaurs and pterosaurs. Current models of fossil feather colour are based on the geometry of feather melanosomes in extant birds, but to date, feather colour has been assigned on a purely visual basis. Here, we assess statistical support for visually assigned categories of melanin-based colour. We performed quantitative analysis of spectral reflectance data from black, brown, grey and rufous feathers, using both raw reflectance spectra and spectrally derived parameters. Only rufous feathers show a systematic division into distinct light and dark subcategories, whereas brown, grey and black categories are better treated as discrete groups. Analysis of spectral data demonstrates that colour categories are distinct. Predictive models for feather colour using linear discriminant analysis yield predicted accuracies of over 80% using parameter data and over 90% using raw spectral data. Results confirm the validity of these visually assigned colour categories. Future studies of feather colour should use spectral analysis to test whether visually assigned colour categories are statistically supported. This will provide a robust foundation for future efforts to investigate the evolution of feather colour in the fossil record.