Forty unipolar patients satisfying DSM-III criteria for major depression who discontinued lithium therapy were retrospectively compared with 105 similar patients who continued the drug and served as a control group. The time to readmission from starting lithium was compared while both groups were still on lithium, and after discontinuation in one group and further continuation in the control group. The progressive increase in the probability of recurrence over two years was the greatest after discontinuation of lithium. For the patients who eventually discontinued lithium, the cumulative probability of recurrence in two years was 0.08 on lithium and 0.58 after stopping it. The probability of recurrence was unchanged over the duration of the study for patients who continued to take lithium. There was no evidence of a lithium withdrawal syndrome within three months of stopping the drug. The results support the view that the everyday clinical use of lithium as prophylaxis is beneficial in unipolar depression.