Cancer patients diagnosed with solid tumors and undergoing a first course of chemotherapy were randomized to either conventional care or a 10 contact 20 week cognitive behavioral intervention designed to reduce emotional distress. Three hypotheses were tested. First, a test for a group effect found that patients in the intervention who entered the trial with higher symptom severity reported significantly lower depression at 10 but not 20 weeks. Patients in the experimental group who entered with higher depression were more depressed at 10 weeks than patients in the control group. The second hypothesis compared the intervention in two sub-indices of symptom severity; one comprised of symptoms with a depressive component, the other of symptoms not associated with depression. The intervention was more effective in lowering depression at 10 weeks through an interaction with the non-depressive symptom sub-index. At 20 weeks, a significant main effect for the intervention on depression was observed. No interaction with either sub-index was observed. The third hypothesis found no main or interaction effect between psychotropic drugs and the intervention. In conclusion, behavioral interventions may influence depression differently over time. Initially, the intervention lowered depression through certain symptoms and only later did it effect depression directly. Symptoms with an affective component are relatively unaffected by the intervention which was independent of the presence of psychotropic medications. Copyright (c) 2004 John Wiley & Sons, Ltd.