A total of 130 people attending psychiatric hospitals within 6 months of onset or relapse of an episode of depressive disorder were interviewed about their symptoms and treatment at the time of their initial contact. After a mean 4-month interval, 119 were reassessed to test the hypothesis that patients treated with antidepressants would be significantly more likely to be clinically improved compared with those untreated. Severity and duration of the episode emerged as the only significant clinical predictors of clinical improvement. Patients on treatment with antidepressants at the start of the study showed a nonsignificant trend for a lesser degree of clinical improvement, even when clinical severity and compliance were taken into account. Those who were not commenced on treatment until later in the study also fared no better than those who were never prescribed antidepressants. The effect of low doses of antidepressants (almost always a tricyclic) appeared to be less beneficial than either higher doses or clinical management without antidepressant drugs. The need for further experimental and naturalistic studies conducted over various periods of time and the implications for clinical practice, medical audit and the appropriate use of health outcome indicators are discussed.