We report an extensive study which compares cognitive therapy, antidepressant drugs and a combination of these two, in depressed patients seen either in general practice or an out-patient department. One-hundred and forty patients were screened for primary major depression and 64 patients completed the trial. All were rated on seven measures of mood, including independent observer-rated and self-rated depression and scales of anxiety and irritability. Patients were randomly assigned to cognitive therapy, antidepressants or a combination of the two. The antidepressant drug group did less well in both hospital and general practice and combination treatment was superior to drug treatment in both hospital and general practice. In general practice, cognitive therapy was superior to drug treatment. The presence of endogenous features did not affect response to treatment. The results are discussed in terms of Beck's cognitive theory of depression and factors of presumed causal importance of depression in general practice.