The Hemopump is a useful left ventricular assist device. Because it is a rotary blood pump, the pump performance is not constant and is dependent on the cardiac cycle. We measured the static flow delivered by the pump at varying pressure heads (delta P) in a mock circulation. These data are compared to the pump performance in vivo. On the basis of these results, 5 sheep were instrumented for continuous Hemopump flow measurement as well as left ventricular and aortic pressure measurements. The Hemopump flow was relayed instantaneously to the pressure head. Low filling and ventricular failing (through intravenous administration of a beta-blocker) conditions were applied. The in vivo measured flows also are pressure head dependent, but the flow curve shows hysteresis resulting in a loop during each cardiac cycle. The in vivo peak flows (delta P = 0) are similar to the in vitro data. The in vivo means flows (delta = 50 mm Hg) are similar to the in vitro data for the lower pump speeds but are less than that at the higher pump speeds (3.74 +/- 0.55 L/min in vivo at Speed 7 versus 4.6 L/min in vitro). Low filling interrupts the delta P-flow loop and reduces flow. In the failing ventricle, delta P increases and flow is reduced. The cannula leaks and results in aortic insufficiency (0.36 +/- 0.05 L/min) when the pump is turned off. Several conclusions have been drawn from these tests: Cardiac activity is beneficial for the pump performance as well as when the aortic pressure curve is nonpulsatile; the longer the systolic phase, the higher the pump flow; the pump should never be turned off in clinical use, and filling is important for the pump's performance.