A bivariate model of response to antidepressant medication (tricyclic) and psychosocial interventions is proposed and a specific test of the drug response aspect of this model was designed, using objectively defined nonpathognomonic behaviors. Repeated measures analyses of variance and interrupted time-series analyses were used to test whether tricyclics affected different constellations of behaviors in depressed inpatients. Results showed that drug interventions increased the frequencies of activity-related behaviors only. However, positive clinical response was related to elevated levels of both activity-related and interpersonal behaviors. The implications of these results for the treatment of depression and the elucidation of behavioral mechanisms of drug action are discussed.