Theoretical and empirical studies of local adaptation in species interactions have increased greatly over the past decade, yielding new insights into the conditions that favor local adaptation or maladaptation. Generalizing the results of these studies is difficult, however, because of the different experimental designs that have been used to infer local adaptation. Particularly challenging is comparing results across empirical studies conducted in a common laboratory or garden environment with results of those conducted using transplants in natural environments. Here we develop simple and easily interpretable mathematical expressions for the quantities measured by these two different types of studies. Our results reveal that common-garden designs measure only a single component of local adaptation-the spatial covariance between the genotype frequencies of the interacting species-and thus provide only a partial description of local adaptation. In contrast, reciprocal-transplant designs incorporate additional terms that measure the contribution of spatial variability in the ecological environment. Consequently, the two types of studies should yield identical results only when local adaptation is caused by spatial variability in the genotype frequencies of the interacting species alone. In order to unify these disparate approaches, we develop a new methodology that can be used to estimate the individual components of local adaptation. When implemented in an appropriate experimental system, this partitioning allows the examination of fundamental questions such as the relative proportion of local adaptation attributable to interactions between species or to the abiotic environment.