Evaluation of the efficacy of antidepressant agents needs to take into account factors which increase the "effect size", such as dosage, treatment duration, the use of two-phase trials, and pattern analysis of responders. Although many patients are thought to receive inadequate doses of antidepressants, relatively few dose-response studies have been performed. However, trials of both tricyclic antidepressants and phenelzine have shown that statistically significant improvements in outcome result from the use of higher dosages; the length of treatment may also affect results. In some studies, the proportion of patients showing a clear-cut response increased significantly among patients treated with active drug instead of placebo when the treatment period was extended from 4 to 6 weeks, independent of the dose used. There may thus be a distinct advantage in extending trials of antidepressants for a minimum of 6 weeks. Two-phase drug trials can be used to extend the trial period still further in responding patients, to emphasize the contrast between treatment and placebo effects and to eliminate type-2 errors. Twelve-week trials might increase the statistical power of the evaluation by 10-20%, in studies where the drug effect size is small. Pattern analysis of the timing and duration of patients' responses has been shown to aid the distinction of true responses from non-specific or placebo effects, and may be useful for evaluating data from studies of antidepressant agents which yield ambiguous results.