An increasingly popular research method for identifying etiologic heterogeneity in psychiatric illness has been to compare the frequency of a risk factor in affected individuals with no affected relatives (sporadic cases) and in affected individuals with one or more affected relatives (familial cases). This paper presents a power analysis of this familial vs sporadic method, assuming a multifactorial model with a normally distributed liability to illness resulting from the additive effect of polygenes and multiple environmental factors. Various parameter estimates for a risk factor that identifies a component of either the genetic or the environmental contribution to disease liability are examined. Almost without exception, large sample sizes of probands and relatives need to be studied to have a substantial probability of detecting etiologic heterogeneity. The required sample size decreases dramatically when monozygotic twins are studied. If the multifactorial model accurately depicts the etiology for psychiatric disorders, these results suggest that the familial versus sporadic design is useful only when pursued in the context of large sample nuclear family studies or studies of monozygotic twins.