Plumage reflectance and the objective assessment of avian sexual dichromatism

Assessment of color using human vision (or standards based thereon) is cent ral to tests of many evolutionary hypotheses. Yet fundamental differences i n color Vision between humans and other animals call this approach into que stion. Here we use techniques for objectively assessing color patterns that avoid reliance on species-specific (e.g., human) perception. Reflectance s pectra are the invariant features that we expect the animal's color cogniti on to have evolved to extract. We performed multivariate analyses on princi pal components derived from >2,600 reflectance spectra (300-720 nm) sampled in a stratified random design from different body regions of male and fema le starlings in breeding plumage. Starlings possess spatially complex pluma ge patterns and extensive areas of iridescence. Our study revealed previous ly unnoticed sex differences in plumage coloration and the nature of irides cent and noniridescent sex differences. Sex differences occurred in some bo dy regions bur not others, were more pronounced at some wavelengths (both u ltraviolet and human visible), and involved differences in mean reflectance and spectral shape. Discriminant analysis based on principal components we re sufficient to sex correctly 100% of our sample. If hidden sexual dichrom atism is widespread, then it has important implications for classifications of animals as mono- or dimorphic and for taxonomic and conservation purpos es.