Here we critically review the scale and extent of adaptive genetic variation in Atlantic salmon (Salmo salar L.), an important model system in evolutionary and conservation biology that provides fundamental insights into population persistence, adaptive response and the effects of anthropogenic change. We consider the process of adaptation as the end product of natural selection, one that can best be viewed as the degree of matching between phenotype and environment. We recognise three potential sources of adaptive variation: heritable variation in phenotypic traits related to fitness, variation at the molecular level in genes influenced by selection, and variation in the way genes interact with the environment to produce phenotypes of varying plasticity. Of all phenotypic traits examined, variation in body size (or in correlated characters such as growth rates, age of seaward migration or age at sexual maturity) generally shows the highest heritability, as well as a strong effect on fitness. Thus, body size in Atlantic salmon tends to be positively correlated with freshwater and marine survival, as well as with fecundity, egg size, reproductive success, and offspring survival. By contrast, the fitness implications of variation in behavioural traits such as aggression, sheltering behaviour, or timing of migration are largely unkown. The adaptive significance of molecular variation in salmonids is also scant and largely circumstantial, despite extensive molecular screening on these species. Adaptive variation can result in local adaptations (LA) when, among other necessary conditions, populations live in patchy environments, exchange few or no migrants, and are subjected to differential selective pressures. Evidence for LA in Atlantic salmon is indirect and comes mostly from ecological correlates in fitness-related traits, the failure of many translocations, the poor performance of domesticated stocks, results of a few common-garden experiments (where different populations were raised in a common environment in an attempt to dissociate heritable from environmentally induced phenotypic variation), and the pattern of inherited resistance to some parasites and diseases. Genotype × environment interactions occurr for many fitness traits, suggesting that LA might be important. However, the scale and extent of adaptive variation remains poorly understood and probably varies, depending on habitat heterogeneity, environmental stability and the relative roles of selection and drift. As maladaptation often results from phenotype-environment mismatch, we argue that acting as if populations are not locally adapted carries a much greater risk of mismanagement than acting under the assumption for local adaptations when there are none. As such, an evolutionary approach to salmon conservation is required, aimed at maintaining the conditions necessary for natural selection to operate most efficiently and unhindered. This may require minimising alterations to native genotypes and habitats to which populations have likely become adapted, but also allowing for population size to reach or extend beyond carrying capacity to encourage competition and other sources of natural mortality.