The role of contextual information in the generation of early hypotheses during the clinical interview was studied with expert and novice medical diagnosticians: 18 experienced family physicians versus 5 new physicians and 12 final-year medical students. The 18 experts and 17 novices were presented with 32 short case histories each presented on three slides: a portrait of a patient, a patient chart containing the previous disease history, and a slide with the present complaint. Because of the sequential nature by which patient data became available during a clinical interview, contextual information is expected to be important in the initial hypotheses. Differences in the number of correct diagnoses were predicted for the two groups, if experts used the contextual information, implicitly provided by picture and patient chart, in a more elaborate way. If this prediction was true, it would show in the amount of information explicitly recalled. The data confirmed these predictions. The experts produced 50% more correct hypotheses as compared to the novices and were able to reproduce a larger amount of relevant contextual information. In addition, a high correlation between problem-solving and recall measures was found only for the expert group. (Author/SW)