Designed to examine the structural differences in the representation of medical problems in subjects with varying degrees of medical expertise, this study uses an online, thinking-aloud technique to investigate the validity of Feltovich and Barrows' model of expert medical knowledge and illness scripts. Study methodology involved asking one medical expert, two intermediates, and a novice to think aloud while being presented with 48 cards arranged in sequential order that characterized the patient or presented history and physical or lab findings for a clinical case of alcohol-induced pancreatitis. The subjects were also asked to provide a differential diagnosis after viewing all 48 cards. Study results indicate that the model is congruent with the elaborate, interrelated structure of the expert subject's case representation, which contained no basic science propositions, but not with the novice and intermediate protocols, which did utilize basic scientific knowledge. It is concluded that these findings affirm the proposition of a direct relationship between characteristics of the knowledge structure and the accuracy of the outcome of the problem-solving process, and it is suggested that an addition to the model is needed to describe how knowledge is activated and applied at different levels of medical expertise. The text is supplemented with 2 figures, and 15 references are provided. (KM)