In 27 inpatients with primary affective disorder the urinary excretion of 3-methoxy-4-hydroxyphenylglycol (MHPG) was measured prior to a 4-week treatment with 150 mg amitriptyline (AT)/day. Ratings according to the Hamilton depression scale were performed before therapy and repeated after 2 and 4 weeks. Plasma levels of AT, nortriptyline (NT), and E-10-hydroxynortriptyline (OHNT) were assayed weekly, and binding of AT to plasma proteins was determined in one sample. Better therapeutic results were obtained at intermediate, as compared to low and high concentrations of AT or AT plus NT. Independent evaluation of AT and metabolite levels revealed that patients with AT of 50--125 ng/ml responded particularly well when NT did not exceed 95 ng/ml or when NT plus OHNT was below 150 ng/ml. Outside this "therapeutic window' the outcome was markedly poorer. Interindividual variation of AT binding was much smaller than variation of total concentrations. Evaluation of free, instead of total levels did not help to clarify the relationship between clinical and pharmacokinetic variables. Plasma levels within the optimal ranges were found in more patients with high than with low MHPG excretion. The free fraction of OHNT in plasma of healthy subjects was about 35%.