In pacing-induced models of atrial fibrillation (AF) that mimic atrial high-frequency foci, the increase in AF susceptibility over time is not paralleled by any increase in dispersion of repolarization (DOR). Measurements of effective refractory periods (ERP), however, were performed by using extrastimuli, which bears significant limitation because it cannot evaluate any dynamic increase in DOR that may arise at rapid rates. Repolarization alternans (Re-ALT), a beat-to-beat alternation in action potential duration, enhances DOR above a critical heart rate. It is unknown, however, whether Re-ALT plays a role in promoting AF. Method and results: two DDD pacemakers, each with right atrial (RA) and ventricular leads, were implanted in 4 male sheep (50-80 kg). The 1st pacemaker was used to deliver 1) electrophysiology protocols for measurements of RA ERP (S1S2) and of RA Re-ALT threshold (S1S1), and 2) intermittent RA burst pacing (for 5 sec followed by 2 sec of sinus rhythm) until sustained AF developed. The 2nd pacemaker was used to record a single broadband unipolar RA electrogram (EGM). The AV junction of the 3rd and 4th sheep has been ablated by RF in order to dissociate far-field ventricular activity from RA EGM. RA repolarization wave was detected following depolarization. Sustained AF was successfully induced after 1-14 weeks of intermittent RA burst pacing. RA ERP decreased progressively from 185±39 ms (pre-activation) to 125±15 ms and to 110±20 ms after 2 and 4 weeks of burst pacing respectively. Importantly, Re-Alt threshold (255±15 ms) did not change during the time course RA burst pacing, but the range of pacing CL during which RA Re-ALT was observed increased until sustained AF developed (45±5 ms before vs 105±25 after 2 weeks of RA pacing). Interestingly, despite aggressive atrial burst pacing protocol, significant changes in atrial histology were not observed. Conclusion: we report here for the first time in vivo measurements of atrial Re-Alt using standard pacemaker technology in a chronic sheep model of pacing-induced AF. Interestingly, Re-Alt threshold did not change during the time course of pacing-induced AF but the pacing CL range during which Re-Alt was observed increased proportionally to the decrease in RA ERP. These parameters were unrelated to significant atrial tissue alterations, but involved cellular alterations including at least important proteins involved in intracellular Ca cycling. Our findings suggest that atrial Re-Alt might be a mechanism by which DOR transiently increases, promoting wavebreaks and AF at rapid rates.