OBJECTIVE: The authors predicted that depressed elderly patients who responded to treatment would rate their baseline health more positively than nonresponders, that responders would again rate their health more positively once they were in remission, and that lower baseline self- ratings of health would predict lack of response to protocol treatment. METHOD: The Perception of Illness Scale was administered to 61 depressed elderly patients at baseline and again upon completion of the acute phase of a depression treatment protocol. A logistic regression was performed to ascertain whether Perception of Illness Scale scores predicted response to protocol treatment. RESULTS: Baseline Perception of Illness Scale scores were poorer among the nonresponders, accurately predicted response or lack of response in 75% of the subjects, and showed before-to after-treatment improvement among the responders. CONCLUSIONS: Patients who initially rated their health as fair to poor were less likely to recover from depression in a standardized treatment protocol. Self-ratings of health improved with resolution of depression.