Since the initially discovery that secondary sexual traits are controlled by a circulating substance by Berthold (1848) many experiments confirmed that typical male traits are under the influence of testosterone in vertebrates. However, these studies also revealed that there is a large variability in the strength and extent of this relationship as testosterone profiles may be related to the type of mating system, degree of parental care, development of sexually selected traits, immuno-competence, and territorial defence, to cite a few. One particularly challenging issue is the fact that many male-specific morphological and behavioural traits that are usually associated to high testosterone levels are often expressed also when the production of gonadal androgens is very low. Several hypotheses have been advanced for explaining these special cases, such as local production of steroids or high sensitivity that make extremely small amounts of androgen sufficient to develop and/or activate the trait. Evidence supporting one of these hypotheses has been obtained in a number of case studies. Nevertheless, there is no general theory which allows us to predict what traits can be expected to be under control of testosterone. Such a theory would be particularly useful to study ecological and evolutionary aspects of behaviour, where the interest in the hormone-behaviour relationships has grown enormously in the last two decades. Here we propose that the testosterone-sensitivity of a trait depend on its co-option into the reproductive context during the evolutionary process and its present role within this context. If the trait has gained importance (and is still important) in male-male or male-female interactions related to reproduction, we predict that this trait will be influenced by circulating testosterone levels. If, on the contrary, the trait did not evolve or is not shown in its present form within a reproductive context, we predict that its expression will not be associated to fluctuations in the levels of the hormone.