BACKGROUND: The effectiveness of lithium is being questioned increasingly and requires clarification. AIMS: To assess the effectiveness of lithium treatment in depression and mania, syndromal types I and II, with predominantly mixed or psychotic episodes or rapid cycling, during treatment resumed following discontinuation, and across three decades. METHOD: The longitudinal course of 360 patients with bipolar disorder compliant with lithium treatment for at least 1 year and without comorbidity for substance use disorder was reviewed. RESULTS: Risk of single-episode recurrences, a common index of treatment failure, was similar to that in other reports. Both episode frequency and 'time ill' improved more in type II than type I cases. Reduced morbidity during treatment was similar in patients with mixed or psychotic episodes, or rapid cycling, and in less complex cases. Retreatment yielded minor decrements in response, and there was no tendency for lesser responses in more recent years. CONCLUSIONS: Based on overall affective morbidity, long-term lithium treatment in compliant patients without comorbid substance use disorder, though imperfect, remains effective, even in subgroups of supposedly poor prognosis.