SUMMARY Evolvability is a function of the way genetic variation interacts with the mechanisms that produce the phenotype. We explore an explicitly mechanistic way of studying the evolvability of phenotypes that are produced by a relatively simple genetic mechanism, the mitogen-activated protein kinase (MAPK) cascade. We developed a quantitative model of MAPK activation that can be used to study the effects of genetic variation on the various components of this signaling cascade. We show how some standard tools of applied mathematics, such as steady-state formulations and nondimensionalization, can be used to elucidate the relative importance of variation in each gene of this mechanism. We also give insights into non-intuitive patterns of dependence and trade-off among the genes. The mechanism produces several different phenotypes (ultrasensitivity to stimulation, switch-like behavior, amount of MAPK-PP delivered, persistence of MAPK-PP activity), each of which is sensitive to different (but partially overlapping) combinations of genes. We show that the mechanism imposes clear limitations on the evolvability of each of the different phenotypes of the pathway, even in the presence of genetic variation in the components of the mechanism. This approach to the study of evolvability is generally applicable and complements the traditional approach through statistical genetics by providing a mechanistic understanding of the genetic interactions that produce the phenotype.