Biologists generally agree that most morphological variation between closely related species is adaptive. The most common method of comparative analysis to test for co–evolved character variation is based on a Brownian–motion model of character evolution. If we are to test for the evolution of character covariation, and we believe that characters have evolved adaptively to fill niches during an adaptive radiation, then it is appropriate to employ appropriate models for character evolution. We show here that under several models of adaptive character evolution and coevolution during an adaptive radiation, which result in closely related species being more similar to each other than to more distantly related species, cross–species analyses are statistically more appropriate than contrast analyses. If the evolution of some traits fits the Brownian–motion model, while others evolve to fill niches during an adaptive radiation, it might be necessary to identify the number of relevant niche dimensions and the modes of character evolution before deciding on appropriate statistical procedures. Alternatively, maximum–likelihood procedures might be used to determine appropriate transformations of phylogenetic branch lengths that accord with particular models of character evolution.