The aim of this study was to determine the presence of organization of atrial activation processes during atrial fibrillation (AF) by assessing whether the activation sequences are wholly random or are governed by deterministic mechanisms. We performed both linear and nonlinear analyses based on the cross correlation function (CCF) and recurrence plot quantification (RPQ), respectively. Recurrence plots were quantified by three variables: percent recurrence (PR), percent determinism (PD), and entropy of recurrences (ER). We recorded bipolar intra-atrial electrograms in two atrial sites during chronic AF in 19 informed subjects, following two protocols. In one, both recording sites were in the right atrium; in the other protocol, one site was in the right atrium, the other one in the left atrium. We extracted 19 episodes of type I AF (Wells' classification). RPQ detected transient recurrent patterns in all the episodes, while CCF was significant only in ten episodes. Surrogate data analysis, based on a cross-phase randomization procedure, decreased PR, PD, and ER values. The detection of spatiotemporal recurrent patterns together with the surrogate data results indicate that during AF a certain degree of local organization exists, likely caused by deterministic mechanisms of activation.