The prophylactic effect of lithium was studied prospectively in 63 patients with recurrent affective disorders. Lithium treatment was started at recovery from a current episode of illness when the patients were able to participate in the decision to start long-term treatment. The median treatment time was 23.7 months and the mean serum lithium concentration approx. 0.63 mmol/l. Comparison with equally long control periods before lithium showed that the treatment resulted in statistically significant reductions in number of episodes, number of months ill, and number of months hospitalized. There were 40 (63%) responders (frequency of episodes during lithium lower than before lithium) and 23 (37%) non-responders (frequency of episodes during lithium higher than or the same as before lithium). Females responded slightly better than males. Responders scored significantly higher than non-responders in the psychasthenia and muscle tension subscales of the Karolinska Hospital Personality Inventory. Both responders and non-responders showed falls in CPRS during lithium treatment, the fall was statistically significant in the responders. During the first 4 months of lithium treatment the responders showed a significant rise in serum calcium, while serum calcium remained unaltered in the non-responders. These variables may be predictive of response to long-term lithium treatment. The low numbers of patients who discontinued treatment and the even serum levels of lithium suggest that good compliance to treatment was achieved with the present approach.