The authors have carried out an investigation of psychiatric morbidity in families of patients who responded and failed to respond to long-term lithium treatment. The study included 121 probands with RDC primary affective disorders and 903 first-degree relatives and spouses. Seventy-one probands were responders and 50 were nonresponders to long-term lithium treatment. Extended to 20 years, the follow-up of patients and their families provided substantial information relevant for the diagnosis and reliable assessment of lithium response. The diagnoses were based on all available information, SADS-L interviews and RDC criteria. The principal statistical methods were survival analysis and Cox regression analysis. The results revealed a significantly higher frequency of bipolar disorder in the relatives of lithium responders (3.8% vs. 0%). Schizophrenia was more common in the families of nonresponders (2.4% vs. 0.3%). There were no significant differences in the rates of other psychiatric disorders. Both family history and the proband's diagnosis contribute independently to predicting response to long-term lithium.