Sealing faults may have a major impact on reservoir compartmentalization and play a key role in advanced oil recovery strategies. Because of the different approaches and scales used to characterize faults, a considerable gap remains between engineers’ and geologists’ conceptions of faulted reservoirs. The fundamental difficulty in integrating these views lies in constructing realistic reservoir-scale models incorporating relevant fault properties on several scales, so that fault connectivity and its impact on fluid flow can be simulated and quantified properly. This paper is a step towards examining quantitatively the impact of fault connectivity on fluid flow at a reservoir scale. A stochastic reservoir model incorporating sealing faults was constructed. The number of faults in the model were gradually increased and several fluid flow simulations performed in order to verify the impact of fault density and fault connectivity on the spatial pressure variation. Despite being synthetic, the model incorporates some properties of an actual carbonate reservoir, the Xaréu oil field (XOF) in the Ceará Basin, NE Brazil. In the XOF, thin (<15 m) oil-bearing carbonate layers are interlayered in a thick (>50 m) shale bed. Sealing effects may be caused by the faults because they usually have a normal slip component, so that the thick shale layer becomes juxtaposed against the thin carbonate layers, behaving as barriers to fluid flow in the carbonate layers. It appears that there is no reservoir compartmentalization in the carbonate layers because there is overall pressure depletion in XOF – all the new wells drilled in the carbonate layers, before starting production, showed formation pressure values smaller than the formation pressure measured in the first wells drilled in the field. Nevertheless, it was not understood how such communication throughout the entire reservoir (approx. 19 km2) could occur, given the presence of laterally extensive faults, with up to 5 km in strike and maximum vertical displacement greater than the reservoir thickness. In the synthetic model, fault population follows a displacement-length property derived from seismic data. Sub-seismic faults were also incorporated in the model using a downscaling law. The location, direction and length of each fault were also generated stochastically. In order to quantify the ‘degree of faulting’, two criteria were used: fault density (the fraction's ‘total fault length per occurrence area’; f1) and fault connectivity (the fraction's ‘number of fault intercepts per occurrence area’; f2). By comparing f1 and f2 for the real and simulated versions of the XOF carbonate reservoir it is concluded that the carbonate layers remain interconnected so that compartmentalization does not exist. From a methodological viewpoint, the most important aspect of the modelling is the change of focus regarding sealing. At least for the purposes of investigating interconnection, the adequate focus is on the ‘percolation core’ of the reservoir (its hydraulically connected and undamaged parts) and not on the faults themselves, regardless of their sealing nature.