The Coronary Artery Surgery Study registry enrolled 24,959 patients from 1974 through 1979. Of these patients, 12,556 had anginal chest pain symptoms and significant angiographic coronary disease (greater than or equal to 70% lesion in a major vessel or greater than or equal to 50% in the left main coronary artery) and were considered coronary bypass surgery candidates. Stepwise linear discriminant-function analysis of clinical and angiographic determinants of therapeutic assignment revealed that myocardial jeopardy (a composite score of the relationship of proximal lesions to retained wall motion in anterior and inferior segments) was the most important determinant. Site-to-site variability, another important predictor of therapeutic assignment, was assessed after adjusting for differing patient populations and was found to be significant (p less than 0.001). Other significant predictors were angina severity, the number of operable vessels, change in activity level, unstable angina and the presence of a left main lesion. Temporal changes (between 1974 and 1979) in the use of beta-blocking drugs and in therapeutic assignment of patients with left main lesions and one-vessel disease suggested altered decision-making in response to the reporting of major coronary surgery trials. We conclude that the dominant factors in therapeutic assignment for patients with completed arteriographic evaluations reflect considerations of jeopardized myocardium and continuing diversity of opinion about the importance of therapeutic and prophylactic indications for coronary surgery.