The evolution of clinical reasoning in medical students was studied. A cross-sectional sample consisted of randomly-selected medical students from three classes. Additionally, twenty-two students were observed at yearly intervals from the preclerkship period to the first post-graduate year. Subjects were observed in a clinical examination of a simulated patient, and their thought processes were abstracted from a 'stimulated recall' of the videotaped encounter. The data were transcribed and coded for computer analysis, yielding several variables characterizing the clinical reasoning process, and four measures of outcome of the encounter. Analysis of variance of differences between students at various educational levels and a doctor criterion group indicated that the majority of the process variables were unrelated to educational level. By contrast, diagnostic and management outcomes were positively related to education. The single process variable which was related to both educational level and outcome was an 'hypothesis aggregate score', a measure of the content of the student's diagnostic hypotheses. The results of the study indicate that the problem-solving or clinical reasoning process remains relatively constant from medical school entry to practice. This observation has important implications for clinical teaching and evaluation.