A multihospital collaborative study evaluated prophylactic lithium carbonate therapy in 205 patients with bipolar manic-depressive illness. Factors associated with poor lithium carbonate response are the following: (1) a recent history of frequent affective episodes requiring hospitalization, and (2) previous failure of lithium carbonate treatment. Results also suggest that lithium carbonate response may be related to the presence of schizo-affective illness and a family history of bipolar affective illness, but the small numbers of patients in these groups limit interpretation. Most failures on lithium carbonate therapy occur during the first year. Ability to remain on the maintenance schedule with no episodes for a year may be the most potent predictor of ultimate prophylactic success. There are important implications in these findings for the clinician selecting patients for prophylactic treatment.