The most unequivocal demonstration of efficacy with putative or reference antidepressants is normally obtained from placebo-controlled studies; heavy reliance is therefore placed on such investigations, particularly in the early stages of development of any new therapeutic agent. However, many clinicians are understandably reluctant to allow some patients to be treated with placebo, which may lead to a sampling bias. It is equally understandable that doctors taking part in such studies are anxious about clinical deterioration or lack of response, and this in turn may result in a higher drop-out rate than would be seen in studies using reference antidepressants without a placebo treatment group. However, in spite of these and other drawbacks, it is still generally accepted that the efficacy of potential antidepressants is established in a more scientific and unequivocal way using placebo-controlled methodology. Fluoxetine has been the subject of a number of such investigations (Table I); their findings are reviewed in this paper.