Birds show striking interspecific variation in their use of carotenoid-based coloration. Theory predicts that the use of carotenoids for coloration is closely associated with the availability of carotenoids in the diet but, although this prediction has been supported in single-species studies and those using small numbers of closely related species, there have been no broad-scale quantitative tests of the link between carotenoid coloration and diet. Here we test for such a link using modern comparative methods, a database on 140 families of birds and two alternative avian phylogenies. We show that carotenoid pigmentation is more common in the bare parts (legs, bill and skin) than in plumage, and that yellow coloration is more common than red. We also show that there is no simple, general association between the availability of carotenoids in the diet and the overall use of carotenoid-based coloration. However, when we look at plumage coloration separately from bare part coloration, we find there is a robust and significant association between diet and plumage coloration, but not between diet and bare part coloration. Similarly, when we look at yellow and red plumage colours separately, we find that the association between diet and coloration is typically stronger for red coloration than it is for yellow coloration. Finally, when we build multivariate models to explain variation in each type of carotenoid-based coloration we find that a variety of life history and ecological factors are associated with different aspects of coloration, with dietary carotenoids only being a consistent significant factor in the case of variation in plumage. All of these results remain qualitatively unchanged irrespective of the phylogeny used in the analyses, although in some cases the precise life history and ecological variables included in the multivariate models do vary. Taken together, these results indicate that the predicted link between carotenoid coloration and diet is idiosyncratic rather than general, being strongest with respect to plumage colours and weakest for bare part coloration. We therefore suggest that, although the carotenoid-based bird plumage may a good model for diet-mediated signalling, the use of carotenoids in bare part pigmentation may have a very different functional basis and may be more strongly influenced by genetic and physiological mechanisms, which currently remain relatively understudied.