A group of 30 patients suffering from endogenous depression was treated with 150 mg amitriptyline (AT) for 21 days. Depression ratings and determinations of total and free plasma AT and nortriptyline (NT) were performed weekly. No correlation between clinical improvement and any of the biochemical parameters was found. Thus, this study does not support the existence of a therapeutic window for AT. A highly significant correlation was calculated between free and total AT and free and total NT, and also between the free fractions of AT and NT; moreover, age correlated significantly and positively with total plasma AT, but not with NT, and negatively with the free fractions of both AT and NT. The absence of correlation between clinical improvement and pharmacokinetic parameters is discussed for its possible significance. The finding that responders are also found in patients with "low" levels of antidepressants (corroborating the pharmacokinetic and pharmacodynamic data obtained in animals submitted to a long-term treatment with antidepressants) suggests that the concept of the need for steady-state levels with low fluctuations should be re-examined. In the light of these results the clinical effectiveness of treatment with higher drug doses, administered at larger intervals, in order to produce high amplitude fluctuations of the antidepressant should be studied.