Data for 401 depressed outpatients with mood reactivity who participated in a randomized trial comparing placebo, imipramine, and phenelzine were analyzed for predictors of differential response by stepwise multiple regression techniques. Features of the Columbia criteria for atypical depression including oversleeping, overeating, severe anergy, and pathologic rejection sensitivity were each predictive of a poorer response to imipramine than to phenelzine only when compared to those patients with none of the features. These features were not additive in their contribution to differential outcome. Lack of endogenous features was not predictive of a differential drug treatment response. Compared with patients who have no symptoms of atypical depression, patients with any of the four features had an inferior imipramine response rather than a superior phenelzine response. These analyses indicate that the clear differential responsivity to medication treatment in atypical depression is not simply related to any one defining symptom and that further correlates of this apparent biological heterogeneity need to be explored.