The onset of action of antidepressant drugs was investigated on the basis of two independent multicenter, double-blind efficacy studies comparing amitriptyline (n = 120), oxaprotiline (n = 120), imipramine (n = 506) and moclobemide (n = 580) with placebo (n = 189 + 191). The samples consisted of in- and outpatients diagnosed, according to Diagnostic and Statistical Manual (DSM)-III criteria, as suffering from major depressive disorder. Measures of efficacy criteria were the Hamilton Rating Scale for Depression (HAM-D), the Hamilton Rating Scale for Anxiety (HAM-A) and the Zung Self-Rating Depression scale. By using the Sustained Relative Improvement (SRI) criterion, onset of action was determined in each individual patient as that time point in the 30 day observation period at which a 20% baseline score reduction was achieved without subsequent deterioration. Analogously, a response to treatment was defined as a 50% baseline score reduction. As expected, highly significant differences between active drugs and placebo were found with respect to the total number of improvers and responders. Significant differences between treatment modalities surfaced in the percentage rate as well as the time distribution of premature withdrawals. Yet, unexpectedly, among improvers, the time spans to onset of improvement were found to be independent of treatment modality as indicated by virtually identical cumulative percentages of improvers throughout the whole observation period. The picture was essentially the same for the HAM-A and Zung assessments, except for a significant time lag between observer- and self-ratings. In particular, our analyses revealed no evidence for a delayed onset of action under various antidepressants with large biochemical and pharmacological differences in comparison to placebo. Moreover, the early onset of improvement was highly predictive of later outcome: on average, 70% of the patients showing improvement within the first 14 days became responders. Applying survival-analytical methods, we found that differences between active treatments and placebo emerged within the first 5 days and reached a point of maximum distinction around day 14. After this time point, differences between treatment modalities remained constant until the end of the observation period. According to our data, 20–25% of the patients were, on average, ‘true’ drug responders, thus suggesting that the therapeutic qualities of antidepressants do not lie in the suppression of symptoms, but rather are related to their ability to elicit and maintain certain conditions which allow recovery in a subgroup of patients who would otherwise remain non-responders.