T he action potential duration (APD) of ventricular myocytes depends on the length of the preceding diastolic interval (DI) in a relationship termed "electrical restitution." The APD shortens with decreasing cycle length and thus with decreasing DI. 1 The teleological advantage of this phenomenon is that it provides increased time for excitation-contraction coupling at long cycle lengths while preserving diastolic time for coronary perfusion and ventricular filling at short cycle lengths.Restitution of the APD may also lead to rich behavior in beat-to-beat APD fluctuations. At a fixed cycle length, a perturbation in action potential that shortens the APD results in concomitant lengthening of the next DI, given that CLϭAPDϩDI. The restitution relationship dictates that a prolonged DI will lead to lengthening of the next APD. Thus, the initial perturbation results in alternation in APD between values that are shorter or longer than the steady-state APD. Whether the APD oscillations grow or decay depends on whether the slope of the restitution curve in the region near the steady state is greater than or less than unity. 2 In the case of growing APD alternans, regions of myocardium may develop sufficiently long APD on some beats as to be rendered refractory to activation on the subsequent beat. This results in wavebreak and the substrate for reentry and fibrillation. [3][4][5] But there is further complexity in electrical restitution in that the shape and duration of the action potential depend on more than just the DI of the preceding beat. 6 -9 The APD observed at any given DI is not unique and depends on the history of activation sequence leading up to that beat. In this issue of Circulation Research, Wu and Patwardhan 10 report on APD restitution in canine isolated ventricular tissue preparations using oscillatory sequences of DI. These investigators made use of a system they previously developed 11 to provide explicit control over DI by starting a timer when the action potential achieved 90% repolarization and stimulating the tissue when the elapsed time reached the prescribed DI for that beat. They utilized sequences in which DI varied linearly, sinusoidally, and randomly.Wu and Patwardhan 10 found that the relationship between APD and preceding DI followed a hysteresis loop rather than a fixed unimodal curve when DI varied sinusoidally. For a given DI, APD was consistently lower when DI was on the rising limb than when DI was falling. When DI varied randomly, the plot of APD versus preceding DI consisted of points that largely filled the interior of a hysteresis loop. When DI varied linearly (following a triangular wave), the APD-versus-DI curve followed a hysteresis loop similar to that observed during sinusoidally varying DI, although with a somewhat shallower set of slopes. These findings are consistent with those of Elharrar and Surawicz, 6 who showed two decades ago that the restitution curve is shifted higher for sequences with long cycle lengths than that found for sequences with shorter cycle lengths...