The spatial and temporal organization of excitation during atrial fibrillation is still a question of debate. To further clarify the mechanisms maintaining chronic atrial fibrillation (CAF), a method for system identification of activation patterns of the atria during sinus rhythm (SR), during artificial atrial pacing, and during atrial fibrillation was developed. Data were obtained as repeated recordings of 8 s in duration, each of 56 atrial bipolar electrograms, simultaneously acquired during open-heart surgery at multiple sites in the right atrial-free wall in 12 patients with stable SR and during artificial pacing and 16 patients with CAF, Realization theory (Ho-Kalman and Juang-Pappa algorithms) was applied to fit the multivariate electrograms corresponding to individual beats during SR and type-I activations during AF, respectively. It was possible to reproduce the multivariate electrograms corresponding to individual beats, including their waveform patterns during SR as well as during type-I atrial fibrillation activation-in all cases and for all subjects tested with good accuracy. It was concluded that the local electrogram is a predictable signal during SR, artificial pacing, as well as during type-I atrial activation, and the proposed method may be used for further investigations concerning the deterministic and predictive nature of atrial activation during atrial fibrillation.