It has sometimes been suggested that the term adaptation should be reserved for differences with a known genetic basis. We argue that adaptation should be defined by its effects rather than by its causes as any difference between two phenotypic traits (or trait complexes) which increases the inclusive fitness of its carrier. This definition implies that some adaptations may arise by means other than natural selection. It is particularly important to bear this in mind when behavioural traits are considered. Critics of the 'adaptationist programme' have suggested that an important objection to many adaptive explanations is that they rely on ad-hoc arguments concerning the function of previously observed differences. We suggest that this is a less important problem (because evolutionary explanations generally claim some sort of generality and are therefore testable) than the difficulties arising from confounding variables. These are more widespread and more subtle than is generally appreciated. Not all differences between organisms are directly adapted to ecological variation. The form of particular traits usually constrains the form or value that other traits can take, presenting several obstacles to attempts to relate variation in morphological or behavioural characteristics directly to environmental differences. We describe some of the repercussions of differences in body size among vertebrates and ways in which these can be allowed for. In addition, a variety of evolutionary processes can produce non-adaptive differences between organisms. One way of distinguishing between these and adaptations is to investigate adaptive trends in phylogenetically different groups of species.