BACKGROUND: Endoepicardial delays in activation observed in patients with atrial fibrillation have been hypothesized to be associated with structural remodeling and layer dissociation. We aim to explore in a computer model how synchrony in endo- and epicardial activation during fibrillation is affected by local layer dissociation. METHODS: A bilayer interconnected cable model of the left atrium was used to simulate 919 episodes of atrial fibrillation (up to 5 sec simulation duration) on different arrhythmogenic substrates with epicardium-only diffuse fibrosis (up to 30% uncoupling). Endoepicardial connections were heterogeneously distributed following random spatial patterns (characteristic length scale from 1.6 to 11.4 mm), resulting in 75% reduction in interlayer coupling relative to control in all cases. This heterogeneous interlayer dissociation divided the atrial bilayer into connected and disconnected regions. Activation time series were extracted at 12,669 sites in both layers to compute endoepicardial delays. Synchrony between endo- and epicardial time series was quantified using a variant of Agmon's jitter-based synchrony index (ranging from -1 to 1) which detected coincidences within a 6-ms time window and compensated for random coincidences. RESULTS: Because of epicardial fibrosis, fibrillatory waves were driven by the endocardium, which generated an endoepicardial delay. Layer dissociation prolonged that delay (4.0±2.2 ms within connected regions vs 11.0±5.0 ms in disconnected regions) within the clinical range of endoepicardial mapping. The delay in the disconnected regions increased with the length scale of the pattern (Pearson correlation of 0.6, p<0.001). The synchrony index was 0.48±0.31 within connected regions and -0.01±0.16 outside. The error rate for classifying connected vs disconnected regions based on the mean delay was 17% (overlap between the two distributions) and 12% when based on the synchrony index. CONCLUSION: Endoepicardial dissociation aggravates activation delays. The synchrony index provides a complementary experimentally applicable tool for quantitative interlayer comparison of fibrillatory activation patterns.