As part of a study in which reduced rapid eye movement latency was used to predict treatment response, fluoxetine and placebo were compared in 89 outpatients with major depression with (n = 52) and without (n = 37) DSM-III-R melancholia, to determine whether the presence or absence of melancholia predicted antidepressant and/or placebo response. Following a 2-week, single blind placebo lead-in, men and women were assigned by random allocation to double-blind fluoxetine, 20 mg/day, or placebo for 8 weeks. Fluoxetine was statistically significantly superior to placebo in patients with melancholia (endpoint change in the Montgomery-åsberg Depression Rating Scale [MADRS] score, response rates and remission rates). A weekly analysis demonstrated statistical superiority of fluoxetine compared with placebo at week 3 and continuing for the remainder of the study. Fluoxetine was statistically significantly more likely to reduce suicidal ideation compared with placebo, using the MADRS item 10 (suicidal ideation question).