Background: The definition of outcome in antidepressant treatment studies may be a crucial factor influencing the relationship between possible treatment variables and treatment response. We therefore wanted to investigate possible relationships between baseline severity of depression and anxiety, and different definitions of outcome among outpatients with major depressive disorder undergoing antidepressant treatment. Methods: Two hundred and forty-eight patients diagnosed with major depression with the Structured Clinical Interview for DSM-III-R Diagnosis – Patient Edition were treated with fluoxetine 20 mg/day for 8 weeks. Patients were evaluated both pretreatment and posttreatment with the 17-item Hamilton Rating Scale for Depression (Ham-D-17), the Clinical Global Impressions Scales for Severity (CGI-S) and Improvement (CGI-I). We chose four continuous definitions of response using a linear regression method to analyze the relationship to baseline anxiety and depression. We used a logistic regression analysis for the relationship between seven categorical definitions of response and baseline severity of depression and anxiety. Results: Greater endpoint severity was significantly positively related to greater baseline severity of depression and anxiety. Lesser baseline severity of depression or anxiety predicted a greater degree of response on either the endpoint CGI-I score (with lower scores indicating greater improvement) or the percent change in HAM-D-17 score (with higher percent change indicating greater improvement). For all seven categorical definitions of response, lower baseline scores were significantly related to the probability of being a responder. Conclusions: These findings support the impression that how outcome is defined affects the strength and direction of observed relationships with predictive variables. Methodological implications are discussed.