A recently published paper by Prioleau and colleagues has questioned whether the benefits of psychotherapy exceed those of placebo in real patients. A critique of that paper is presented drawing attention to key methodological weaknesses, to the subjective interpretation of the meta-analysis data, and to the selective and tendentious presentation of the authors' conclusions. Arguing that psychotherapy is best evaluated in relation to a limited group of conditions, and to disorders in which high efficacy is claimed, controlled studies of psychotherapy for neurotic and reactive depressive disorders are reviewed. The view that psychotherapy may have a key effect in influencing the natural history of these depressive disorders is explored.