Forty-five symptomatic patients with aortic regurgitation underwent graded treadmill exercise testing before operation. Twenty-seven patients (group A) could not complete stage I of the National Institutes of Health exercise protocol because of limiting symptoms (exercise duration less than or equal to 22.5 minutes); 18 patients (group B) completed this stage without limiting symptoms (exercise duration > 22.5 minutes). Patients in group A had higher resting pulmonary capillary wedge pressures (mean 19 vs 13 mm Hg, p < 0.05) and left ventricular (LV) end-diastolic pressures (mean 24 vs 16 mm Hg, p < 0.05) than those in group B, but did not differ with respect to LV systolic dimension or fractional shortening by echocardiography or LV ejection fraction at rest or during exercise by radionuclide cineangiography. Among 32 patients with subnormal preoperative LV fractional shortening on echo, nine of 17 in group A and 0 of 15 in group B have died (p < 0.01); seven of the nine deaths were from late congestive heart failure. Group A patients also had less decrease postoperatively in LV diastolic size by echocardiography (mean decrease 8 vs 23 mm, p < 0.001) and less increase postoperatively in LV ejection fraction during exercise by radionuclide cineangiography (mean increase 11% vs 23%, p 0.05) than group B patients. No group A patient and 60% of group B patients had normal exercise ejection fractions postoperatively (p < 0.01). The differences in postoperative mortality and function were not predicted by the differences in preoperative hemodynamics between the two groups. Thus, exercise capacity is imprecise in assessing preoperative LV function in symptomatic patients with aortic regurgitation, but is useful in predicting long-term survival after operation and reversibility of LV dilatation and systolic dysfunction.