Of an original cohort of 39 epileptic patients, 28 were observed for a median of 26.5 months after an initial psychiatric interview. As had been noted in the original cohort, treatment with phenobarbital, as compared with treatment with carbamazepine or no anticonvulsant, was associated with higher rates of depression (38% vs 0%, P = .04). Depression associated with phenobarbital treatment at intake did not remit spontaneously on follow-up as long as phenobarbital use continued. Those patients treated with phenobarbital who were previously depressed and whose medication was changed to either carbamazepine or no medication showed nonsignificant trends toward declines in both the frequency and severity of depressive symptoms. These results provide further evidence that treatment with phenobarbital increases the risk for depression in epileptic patients and should be avoided when clinically feasible, particularly in patients with a personal or family history of affective disorder. Epileptic patients who receive phenobarbital because of clinical considerations should be monitored closely for symptoms of an affective disorder, and if depression is detected, a change to an alternative anticonvulsant may result in amelioration of depressive symptomatology.