Research directed at an understanding of medical expertise is about 30 years old, and many developments in this literature parallel progress in cognitive psychology. Over the past 15 years or so, this research became much more closely identified with particular psychological theories. Initial forays into medicine were essentially direct applications of methods developed in the psychology lab to the more natural domain of medicine, with varying degrees of success. These attempts were followed by a second wave that took the psychological theories themselves more seriously in a more thoughtful application of psychological methods to the medical domain. I will argue in the present paper that the methods and theories used in the study of medical expertise have advanced to the point that there is some reverse flow and they are providing a unique and valuable perspective on the nature of thinking.

Medical diagnosis has considerable potential interest for psychological investigation. Specific diseases can often be described according to a formal rule-like structure, which is exemplified in the extreme by the . DSM-IV approach to psychiatric diagnosis, but in less extreme form in textbooks of clinical diagnosis (e.g., “myocardial infarction (heart attack) is associated with retmsternal chest pain with radiation down the left arm. Patients may be tachycardie, and diaphoretic. The blood pressure will be below normal. ECG changes include…”). Mastery of these diagnostic rules is a formidable learning task for all students. These list-like rules are complemented by causal models, rooted in physiology, biochemistry, etc., which comprise the first two years of most medical school curricula. After mastery of the formal knowledge in medical school, students enter an apprenticeship period (clerkship, internship, residency) lasting from 4 to 10 years, during which they acquire a wealth of examples to draw on. Thus, in contrast to the experimental psychology lab, where category rules are learned during a one hour session, these rules are acquired over years and shared by all. And in contrast to the causal rules that may emerge from the social psychology lab (”John struck out at the stranger because he is an aggressive person”), the causal mechanisms are a shared form of knowledge in medicine, expected of all practitioners. Finally, the fact that expertise takes many years gives the experimenter an additional variable, expertise, to be explored in situations where experts may use a qualitatively different strategy than novices.

Perhaps because of these characteristics, clinical reasoning has proven an intriguing area for testing cognitive theories, a pursuit that has preoccupied Lee Brooks and me ever since our collaboration began in the 1980s. I view it as an auspicious coincidence that led me to Lee, and my research has been incalculably enriched by Lee’s conceptual insights.

From Psychology to Medical Education

Early History

Early research conducted in the 1970s was essentially atheoretical, and was directed at understanding the “problem-solving process” of expert physicians. The goal was simply to improve teaching in medical schools, not to advance any theory of reasoning. Methods used in this first generation of studies were observational, and almost at the level of job task analysis. Expert physicians and students were placed in a setting very similar to their normal work environment, with some consistency imposed by using standardized patients (people trained to simulate a disease instead of real patients). Participants were told to approach the patient as they would in their normal practice, and were encouraged to think aloud or subsequently reviewed a videotape ol their performance and commented (stimulated recall; Neufeld, Norman, Barrows, & Feightner, 1981).

Implicit in the research strategy was the assumption that a general problem-solving process that discriminated experts from students would emerge from the observed performances and introspective comments. Such was not the case. Whatever characterized the problem-solving process - generation of early diagnostic hypotheses, focused search for data, etc. - did not look any different between experts and novices (Neufeld et al., 1981). Experts had better hypotheses but they went about the problem-solving process the same way. Further, being correct on one problem was a poor predictor of success on a second problem: a phenomenon labeled “content specificity” by Rlstein (Elstein, Shulman, & Sprafka, 1978), which casts doubt on the possibility of a general process acquired with expertise. The consequence was that a new generation of researchers began to more critically examine the nature of expertise from the perspective of knowledge organization, using methods derived from various domains of cognitive psychology. Some examples of these second-generation methods:

* Patel and Groen (1960, 1990) examined clinical reasoning using methods borrowed from text processing. They had doctors and students read a written case and “think aloud.” The transcription was then analyzed using the prepositional analysis methods of Kintsch (1974). Patel’s major finding was that medical experts, like physics experts (Larkin, McDermott, Simon, & Simon, 1980), tend to use “forward reasoning,” from data to hypothesis, whereas novices use “backward reasoning.” However, this finding has been challenged by more recent evidence (Eva, Brooks, & Norman, 2001a).