Common mental disorders such as schizophrenia, bipolar disorder, and severe major depression are highly heritable, but differ from single-gene (Mendelian) diseases in that they are the end products of multiple causes. Although this fact may help explain their prevalence from an evolutionary perspective, the complexity of the causes of these disorders makes identification of disease-promoting genes much more difficult. The "endophenotype" approach is an alternative method for measuring phenotypic variation that may facilitate the identification of susceptibility genes for complexly inherited traits. Here we examine the endophenotype construct in context of psychiatric genetics. We first develop an evolutionary theoretical framework for common mental disorders and differentiate them from simpler, single-gene disorders. We then provide a definition and description of endophenotypes, elucidating several features that will make a proposed endophenotype useful in psychiatric genetic research and evaluating the methods for detecting and validating such endophenotypes. We conclude with a review of recent results in the schizophrenia literature that illustrate the usefulness of endophenotypes in genetic analyses of mental disorders, and discuss implications of these findings for models of disease causation and nosology. Given that in mental disorders as in behavior generally, the pathways from genotypes to phenotypes are circuitous at best, discernment of endophenotypes more proximal to the effects of genetic variation will aid attempts to link genes to disorders.