Interdisciplinary collaboration between physicians and engineers is now widespread in fundamental and clinical studies, due to the increasing complexity of the physiological problems tackled, and the development of novel measurement devices and data analysis techniques. The present work is a clear illustration of this fact. In the study that triggered it, elaborate experiments on an animal model yielded large amounts of data perturbed by unavoidable interferences, and the phenomenon of interest had an elusive nature. This thesis describes the physiological motivation for this study, introduces the signal processing techniques developed to remove interferences and extract the parameters of interest, and proposes an interpretation of the results obtained. The present thesis is aimed at providing new insights into the mechanisms promoting atrial fibrillation during rapid pacing. Atrial fibrillation is the most common arrhythmia in the developed world, affecting millions of individuals. Atrial fibrillation initiates when triggers such as pulmonary vein tachycardias interact with substrates. However, the exact nature of the electrophysiological substrates that favor transition from runs of pulmonary veins tachycardia to persistent atrial fibrillation remains unclear. Interestingly, repolarization alternans, a beat-to-beat alternation in action potential duration or amplitude, has been mechanistically implicated in transitions from rapid pacing to atrial fibrillation by facilitating dispersion of atrial refractory periods. During rapid pacing, however, transitions from 1:1 to 2:1 atrial capture may supposedly antagonized the dispersion of repolarization driven by repolarization alternans. The ability of atrial unipolar depolarization and repolarization parameters (i.e. activation time and repolarization alternans) to predict the imminence of atrial fibrillation initiation by rapid pacing was evaluated at baseline (i.e. before any burst-pacing induced remodeling occurred). A chronic free-behaving ovine pacing model was developed to study the interplay between atrial repolarization alternans that may facilitate atrial fibrillation initiation during rapid pacing, and the onset of reduced excitability that may cause atrial capture failure and prevent atrial fibrillation initiation. We observed in human atria using monophasic action potential recordings transient decrease in excitability (as exemplified by activation time prolongation) and capture failure that appeared to quench repolarization alternans driven by rapid pacing. Using our ovine model of rapid pacing, human observations were successfully reproduced, which allowed us to analyze repolarization alternans and activation time kinetics until the first beat of capture failure. Prolongation of activation time suggestive of decreased excitability and repolarization alternans preceded 90% of the transitions from 1:1 to 2:1 atrial capture during rapid pacing. Notably, 2:1 atrial capture was imminent at pacing cycle lengths <40 ms above effective refractory period. In the rare cases where capture failure did not occur, repolarization alternans preceded most episodes of non sustained atrial fibrillation, suggesting that intermittent capture plays a protective role against repolarization alternans-induced reentrant arrhythmias. Moreover, the dynamics of atrial repolarization alternans as a function of pacing rate were studied in details. Repolarization alternans amplitude increased as a function of pacing rates, but appeared intermittently without periodicity. This latter finding was suggestive of the presence of nodes (i.e. sites separating islands in opposite phase) spanning the atrial surface. In summary, since rapid atrial tachycardias may also slow propagation velocity, known to be proarrhythmic, transitions to 2:1 atrial capture may protect against dispersion of repolarization and atrial fibrillation. A better understanding of the mechanisms causing intermittent atrial capture may allow treatments to be tailored for preventing atrial fibrillation induction at lower rates.