Fisher's stepwise discriminant analysis was carried out on 10 baseline variables searching for posterior prediction of success in a sample of sixty DSM-III obsessive-compulsive patients. The patients' median score on Hamilton rating scale of depression was 19. In a controlled study they were randomized into three groups: fluvoxamine with antiexposure, fluvoxamine with exposure, or placebo with exposure. In the whole sample, five variables accounted for 76% of the correctly classified patients: avoidance score of the behavioural avoidance test, behaviour therapy expectations, fluvoxamine expectations, Beck depression inventory and rituals repetition. High avoidance score predicted 68% of the correctly classified patients in the whole sample. The discriminant function classified correctly 70% of the patients in the antiexposure with fluvoxamine group, versus 75% in the exposure with fluvoxamine group, and 85% of the patients receiving exposure with placebo (χ2, d.f.2, NS).