BACKGROUND. From both a clinical and an aetiological perspective, major depression (MD) is probably a heterogeneous condition. We attempt to relate these two domains. METHOD. We examined which of an extensive series of clinical characteristics in 646 female twins from a population-based register with a lifetime diagnosis of MD predicts the risk for MD in co-twins. MD was defined by DSM-III-R criteria. RESULTS. Four variables uniquely predicted an increased risk for MD in the co- twin: number of episodes, degree of impairment and co-morbidity with panic disorder or bulimia. One variable uniquely predicted decreased risk: co-morbidity with phobia. Variables that did not uniquely predict risk of MD in the co-twin included age at onset, number and kind of depressive symptoms, treatment seeking, duration of the longest episode and co-morbidity with generalised anxiety disorder and alcohol dependence. CONCLUSIONS. Our results suggest that the clinical features of MD can be meaningfully related to the familial vulnerability to illness, particularly with respect to recurrence, impairment and patterns of co-morbidity.