Performance of temporary parallel left ventricular assistance was investigated and the theoretic conditions leading to optimal behavior of the mechanical system were explored. Computer models of nonpulsatile and pulsatile left ventricular assist devices (LVADs) were incorporated into a previously reported closed-loop simulation of the canine cardiovascular system. Assuming the assisted heart was capable of recovery, LVAD performance was assessed based on both myocardial oxygen balance and cardiac output. With a synchronous LVAD, and operating in a counterpulsation mode, these variables were sensitive to the phasing of pump ejection. Maximum reduction in cardiac oxygen consumption, maximum increase in oxygen availability, and maximum increase in cardiac output with the atrio-aortic device were obtained when pump ejection immediately followed aortic valve closure. These variables were directly proportional to the magnitude of bypass volume. The pulsatile asynchronous and nonpulsatile LVAD models affected oxygen balance in a similar manner, but neither performed so well as the synchronous model when equal bypass volumes were used. Ventricular uptake of blood provided a further 27% decrease in oxygen consumption and further 78% increase in oxygen availability than atrial uptake. In summary, the model predicted that the pulsatile synchronous LVAD, filling from the ventricle during heart systole and ejecting into either the ascending or descending aorta just after ventricular systole, would be most beneficial to both myocardial oxygen balance and cardiac output.