Fifteen general physicians were given profiles of symptom combinations representing 80 depressed patients. They were asked to judge the severity of the disorder and to prescribe. There were two phases to the experiment, the formal structure of both being the same. In the first phase, all physicians based their judgments only on the cues of the Hamilton Depression scale; in the second, each physician defined his own cues. Multivariate regression analysis was applied to the observations. Agreement between the judges was low. Most had a more complex policy, in regard to both judgment of the severity of the depression and prescribing, when selecting their own cues than when restricted to textbook variables, but used it more consistently. The observations have important implications for training and research in psychiatry and in psychopharmacology.