Results from placebo-controlled trials of antidepressants can be used to identify patients most likely to benefit from medication. Using data from a randomized clinical trial of fluoxetine versus placebo for 671 elderly outpatients with major depression, we evaluated characteristics of those who improved with and without active medication. We found that the choice of outcome measure made a difference when evaluating the effectiveness of fluoxetine relative to placebo and determining the accuracy of predictive variables in both treatment groups. Generally, less severe depression predicted favorable response (greater than 50% improvement on the 21-item Hamilton Rating Scale for Depression [HAM-D-21], less than 3 on the Clinical Global Impressions [CGI] and Patient Global Impressions [PGI] improvement scales) and remission (less than 9 on 6-week HAM-D-21) with both fluoxetine and placebo. Less anxiety/somatization was associated with favorable fluoxetine response, and lower levels of cognitive and sleep disturbance were associated with remission in the placebo group. By contrast, higher levels of psychomotor retardation in the placebo group were associated with clinician and patient ratings of much or very much improved. The similarities among responders in both groups may indicate that some in the fluoxetine group would have improved with placebo.