What variables will predict who should be treated with antidepressants and who should not? If someone should be treated, what variables will predict which treatment is most likely to work? If patients are treated, what will predict who will respond and who will not? Clinicians have asked these questions ever since researchers established that antidepressants were superior to placebo in the treatment of depression. Innumerable reports claim either biologic or psychosocial predictor variables of response, and every few years yet another review of predictors is published in the literature. This article reviews methodologic issues that have been neglected in the pursuit of predictors to antidepressant response and proposes improved requirements for predictor research that would impact clinical practice.