From a genetic point of view, the application of mathematical models to affective disorders has not yet been useful, since they do not indicate a specific mode of transmission. To use these models correctly, we need to identify homogeneous genetic subgroups among those sharing the common phenotypic feature of affective illness. Our useful criterion for this is outcome on long-term lithium therapy, since experimental data suggest the existence of a close relationship between the genetic mechanisms that underly the affective disorders and those that underly outcome on lithium. We have studied 145 subjects with primary affective disorders, 92 of whom did not relapse during lithium treatment and 53 of whom did, together with 864 of their first-degree relatives. The data for both groups fit both single major locus and multifactorial polygenic models for genetic analysis, including a sex effect and therefore neither mode of transmission can be excluded.