In recent decades, there have been many studies reporting that antidepressants have a rapid onset of action, with improvement occurring in the first week. The current pilot study questions whether such findings reflect an artefact emerging from high rates of 'nonspecific' improvement and evaluates the phenomenon in a small sample of melancholic patients seemingly lacking nonspecific improvement propensities. Twenty-nine patients with a well-defined melancholic depression completed a 12-week treatment study comparing drug therapy versus cognitive behaviour therapy. The primary outcome measure was the Hamilton Rating Scale for Depression, and a self-report measure of depressed mood severity (the Daily Rating Scale) was completed daily. Analyses seeking time till onset of action were limited to those receiving drug therapies. The lack of improvement in the first 4 weeks for those receiving cognitive behaviour therapy argued for the melancholic patients lacking the capacity for a nonspecific response to therapy. Formal 12-week responder status in those receiving the antidepressant could not be predicted from improvement status until day 12 of the study, and not in the first week as reported in most previous studies of those with major depression. This pilot study argues for any study seeking to quantify the specific interval for onset of action of antidepressant drugs focusing on only those with well-defined melancholia.