Thirty-two outpatient depressives were treated by experienced therapists during a 16- to 20-week, 20- to 24-session cognitive-behavioral therapy (CBT) protocol. Patients were classified as CBT responders (n = 22) or nonresponders (n = 10) on the basis of independent clinical ratings of Hamilton (1960) depression severity. Point and confidence interval estimation procedures yielded results consistent with hypotheses derived from the states-of-mind (SOM; Schwartz & Garamoni, 1986) model. At posttreatment, CBT responders shifted the balance of positivity and negativity to the optimal range, whereas nonresponders remained in a predominantly negative SOM. Response status was related more strongly to change in positivity than in negativity. Findings support the view that clinical response to CBT depends on reducing negativity and increasing positivity until an optimal balance is achieved.