Hospitalized manic patients were withdrawn from psychoactive medications for 2 weeks after which they were randomized to double-blind treatment with carbamazepine plus lithium [CBZ-Li] or haloperidol plus lithium [HAL-Li] with benztropine. Unit dosages of Li 300 mg, CBZ 200 mg and HAL 2 mg were titrated to therapeutic plasma levels and maintained for 8 weeks. No rescue medications were permitted after 3 weeks. Standard ratings of psychopathology and side effects were accomplished weekly. Sixty patients entered the study but only 33 remained for randomization after drug washout. By 8 weeks both groups were improved from baseline without statistically reliable differences between them. However HAL-Li patients had more extrapyramidal side effects that were major reasons for dropout, whereas CBZ-Li patients were more often noncompliant and initially required more rescue medications. We conclude that either combination treatment can be beneficial but CBZ-Li has the advantage because of fewer neurologic side effects.