A sample of 137 child and adolescent outpatients with major depressive disorder were examined to identify baseline clinical characteristics that predicted symptom severity at the end of a 3-week evaluation period and to determine whether change in symptom severity between week 1 and week 2 predicted symptom severity at week three. Subjects underwent three consecutive weekly evaluations prior to being considered for entry into a double-blind, placebo-controlled treatment trial of fluoxetine. Results indicated that the combination of age, social functioning, family history, Children's Depressive Rating Scale-Revised (CDRS-R) (Poznanski et al. (1985) Psychopharmacol. Bull. 21, 979-989) total score at visit one, and percent change in symptom severity between visit one and visit two were predictors of symptom severity at visit three. These findings suggest that (1) subjects should not be excluded from randomized controlled clinical treatment trials based solely on improvement of symptom severity between visits and (2) an extended evaluation period is warranted, especially for adolescents whose symptom severity tends to fluctuate from week to week.