Ninety-eight patients with acute myocardial infarction were examined by 3 clinicians who, independently of each other, gave an estimate of left ventricular (LV) and right ventricular (RV) ejection fraction (EF) in each patient. Their estimates were based on physical examination, chest x-ray, electrocardiogram, patient history and clinical course during admission. Ejection fractions were estimated as belonging to 1 of 4 categories: normal (LVEF greater than or equal to 0.53, RVEF greater than or equal to 0.57), mildly reduced (LVEF 0.40 to 0.52, RVEF 0.45 to 0.56), moderately reduced (LVEF 0.30 to 0.39, RVEF 0.35 to 0.44) or severely reduced (LVEF less than 0.30, RVEF less than 0.35). Radionuclide ventriculography was carried out immediately after the physical examination. LVEF was correctly estimated in 43% of all examinations, deviated from radionuclide LVEF by 1 LVEF category in 45% and by 2 LVEF categories in 12%. The 3 clinicians agreed on estimated LVEF in only 32% of the patients. RVEF was correctly estimated in 67% of the examinations, but none of the clinicians identified greater than 43% of the relatively few patients with reduced radionuclide RVEF and they greatly disagreed as to who among the patients had a reduced RVEF. Previous myocardial infarction, electrocardiographic infarct location, Killip class, physical signs of left- and right-sided heart failure, radiographic pulmonary congestion and cardiomegaly were analyzed to determine which were the most helpful in predicting LVEF and RVEF. The results disclosed that several variables, traditionally believed to be reliable indexes of reduced ventricular function, were surprisingly poor predictors of LVEF and RVEF.