Parallel groups of depressed (DSM III-R) outpatients received moclobemide (n = 22) and fluoxetine (n = 19), double blind, for 6 weeks. Respective starting doses were 150 mg twice a day and 20 mg q.a.m. These could be doubled after 3 weeks for greater efficacy. Chronic users of benzodiazepine anxiolytics continued taking them as comedication. Therapeutic and side effects were assessed using conventional rating scales. Actual driving performance was assessed during the week before therapy and at 1, 3 and 6 weeks thereafter using a standardized test that measures standard deviation of lateral position (SDLP). Similar remissions in depressive symptoms and side effects occurred in both groups. Patients drove with normal and reliable (r = 0.87) SDLPs before treatments. Most continued to do so but a few drove with progressively rising SDLPs and the overall trends were significant in both groups (p < 0.03). A post-hoc multiple regression analysis was applied for identifying factors that correlated with SDLP in separate tests after the beginning of therapy. At 3 and 6 weeks there were significant (p < 0.03) relationships involving the same factor; patients who drove with progressively higher SDLPs appeared to be those using benzodiazepines that are metabolized by a P450 isozyme subject to inhibition by their particular antidepressant.