The authors studied 53 patients hospitalized for depression to evaluate possible associations between pretreatment responses to the Minnesota Multiphasic Personality Inventory and behaviorally rated responses to lithium. Patients were randomly assigned to two groups (A and B) in which responders and nonresponders were balanced for sex. For the 18 women in group A, a lithium response scale discriminated responders with 89% accuracy. For the 9 males, a separate scale discriminated responders with 100% accuracy. In a cross-validation, the two scales discriminated responders and nonresponders with 100% accuracy in group B. The empirical methodology of this study suggests an alternative to the theoretical-rational approach of predicting response to antidepressant drugs based on pretreatment depressive symptoms.