In studies to ascertain the basis of dynamic changes in the P wave, bipolar epicardial potentials were recorded from multiple atrial electrodes in dogs. One hundred to 120 activation times were displayed by a digital computer and used to construct atrial isotemporal activation sequence maps. Changes in heart rate or beat-to-beat cycle length were induced by vagal stimulation or infusion of autonomic mediating drugs. Changes in cycle length were associated with dynamic changes in the atrial activation sequence and surface P-wave. A conspicuous finding was that epicardial atrial depolarization began at three widely separated locations. These three points were consistently present in all animals and were generally located at the 12, 3, and 6 o'clock positions of the superior vena cava-right atrial junction. The dynamic changes in P waves and atrial activation sequence which accompanied the changes in cycle length were due to sudden shifts in the point of earliest activity between the three early sites. Asymmetric atrial depolarization with more rapid conduction along the crista terminalis, superior interatrial band, and pectinate muscles was present in all dogs. Although the anisotropic atrial geometry played an important role in the asymmetric conduction, the widely distributed onset of activity contributed significantly to the uneven spread. The multiple points of origin of the atrial wavefront might be explained by either a trifocal, distributed pacemaker or the epicardial exits of three specialized pathways conducting an impulse emanating from a single focus. These data explain the dynamic variation in P-wave morphology in normal hearts and also imply a relationship between the altered origin of atrial depolarization, atypical P waves, brady- or tachyarrhythmias, and heart rate control.