Trimipramine, a sedating tricyclic antidepressant, and imipramine were compared on polysomnographic parameters during a 4-week double-blind trial in depressed patients with insomnia and anxiety. Trimipramine eliminated objective evidence of sleep disturbance. This was not the case with imipramine, although depression improved similarly in both groups. Subjects' sleep appeared unchanged or more disturbed at the end of the treatment with imipramine. For trimipramine, the major changes in sleep parameters occurred during the first week of drug administration and did not parallel the gradual changes seen in the measures of depression. Additionally, trimipramine did not suppress REM sleep even in a subgroup of six trimipramine patients who had short rapid-eye-movement (REM) sleep latencies during the placebo baseline period, even though their depression was alleviated. The data demonstrate that (a) antidepressants may vary in their effects on sleep, even though they have similar effects on depression; (b) REM sleep suppression does not necessarily accompany improvement in depression; and (c) reports of improved sleep by patients undergoing antidepressant therapy may not reflect improvement on objective measures of sleep. The different sleep effects suggest the possibility of different antidepressant pathways.