Traits do not evolve independently. To understand how trait changes under selection might constrain adaptive changes, phenotypic and genetic correlations are typically considered within species, but these capture constraints across a few generations rather than evolutionary time. For longer-term constraints, comparisons are needed across species but associations may arise because of correlated selection pressures rather than genetic interactions. Implementing a unique approach, we use known patterns of selection to separate likely trait correlations arising due to correlated selection from those reflecting genetic constraints. We examined the evolution of stress resistance in >90 Drosophila species adapted to a range of environments, while controlling for phylogeny. Initially we examined the role of climate and phylogeny in shaping the evolution of starvation and body size, two traits previously not examined in this context. Following correction for phylogeny only a weak relationship between climate and starvation resistance was detected, while all of the variation in the relationship between body size and climate could be attributed to phylogeny. Species were divided into three environmental groups (hot and dry, hot and wet, cold) with the expectation that, if genetic correlations underpin trait correlations, these would persist irrespective of the environment, whereas selection-driven evolution should produce correlations dependent on the environment. We found positive associations between most traits in hot and dry environments coupled with high trait means. In contrast few trait correlations were observed in hot/wet and cold environments. These results suggest trait associations are primarily driven by correlated selection rather than genetic interactions, highlighting that such interactions are unlikely to limit evolution of stress resistance.