OBJECTIVES We aimed to study the relation between onset and peak time of circumferential shortening and the direction of propagation of these parameters in both ischemic and nonischemic patients. BACKGROUND Peak time is often used to select patients for cardiac resynchronization therapy, whereas pacing influences only the onset times directly. Furthermore, it is unclear whether there is a consistent direction of propagation delay and whether this depends on the etiology. METHODS Magnetic resonance imaging myocardial tagging with high temporal resolution (14 ms) was applied to 29 patients (18 nonischemic, 11 ischemic) and 17 healthy control subjects. Time to onset (T(onset)), to first peak (T(peak,first)), and to maximum peak (T(peak,max)) of circumferential shortening were determined. Three-dimensional vectors were calculated to denote the main direction of asynchrony. RESULTS In both patient groups, T(onset) showed a significant positive relation with both T(peak,first) and T(peak,max); however, T(peak,first) correlated considerably better with T(onset) than did T(peak,max) (p < 0.0001 for nonischemic, and p < 0.01 for ischemic patients). Moreover, the relations between T(peak) and T(onset) were stronger in the nonischemic patients than in the ischemic patients (p < 0.001). In nonischemic patients, the propagation of T(onset) was consistently from septum to lateral wall. In the ischemic patients, however, no consistent direction of propagation was found. For both groups, the longitudinal propagation delays (between apex and base) were negligible compared with the short-axis delays. CONCLUSIONS The relation between peak time and onset time of shortening is strongest in nonischemic patients and is most consistent when time to first peak is used (instead of time to maximum peak).