Isolated canine hearts with a critical stenosis on one coronary artery were perfused for 2 h with blood from supporting dogs using a new roller pump system that can deliver pulsatile or non-pulsatile flow perfusion. Non-pulsatile perfusion caused a decrease in coronary venous oxygen tension of 22% (P less than 0.05) accompanied by increasing carbon dioxide tension of 50% (P less than 0.02). With pulsatile flow coronary venous oxygen and carbon dioxide tensions remained stable. Non-pulsatile perfusion decreased the coronary arteriovenous oxygen difference by 35% (P less than 0.02), coronary blood flow by 40% (P less than 0.02), and myocardial oxygen consumption by 54% (P less than 0.01) whereas pulsatile flow did not change any of these variables. Subendocardial blood flow distal to the stenosis fell by 0.15 +/- 0.04 ml/min per gram myocardium (mean +/- S.E.M.) (P less than 0.01) during linear perfusion. The endocardial/epicardial-flow ratio was less than one and decreased further during fibrillation period indicating underperfusion of the endocardial muscle region. With pulsatile flow subendocardial flow remained unaltered during the two hours of fibrillation. Edema formation was 24% in hearts subjected to non-pulsatile flow but only 14% in hearts perfused by pulsatile perfusion (P less than 0.05). Accordingly, the ischemic area involved 40% of the left ventricle during non-pulsatile flow but 25% of the left ventricle in hearts perfused by pulsatile perfusion (P less than 0.05). The results indicate that pulsatile flow perfusion may prevent severe hemodynamic, hematologic, and metabolic alterations in fibrillating isolated canine hearts. It is suggested that pulsatile perfusion may be useful for fibrillating hearts during open heart surgery.