It has been suggested recently that new quantitative methods for analy
zing comparative data permit the identification of evolutionary proces
ses. Specifically, it has been proposed that new comparative methods c
an distinguish the direct effects of natural selection on the distribu
tion of a trait within a clade from the effects of drift, indirect sel
ection, genotype-by-environment interaction, and uncontrolled environm
ental variation. Such methods can supposedly unravel the relative impo
rtance of these factors by the phylogenetic analysis of traits, perfor
mance attributes, and habitats. We argue that they cannot. We show tha
t many different evolutionary mechanisms can, in principle, account fo
r any one interspecific pattern, and we illustrate our case using exam
ples from the comparative literature. We argue that these confounded m
echanisms can only be unraveled if patterns of selection or genetic va
riation and covariation are directly measured in many species within a
clade. Even though comparative methods are valuable for examining the
evolutionary history of traits, they will often mislead in the study
of adaptive processes.