Abstract Most of the models that simulate fractured rock masses assume fully persistent discontinuities, simplifying the fact that, in nature, fractured rock masses are made of non-continuous sets of joints. A rock bridge gives an effective cohesion to the fracture and a block of rock cannot fall or slide until all the rock bridges fail. This failure involves the failure of the intact rock, which can be orders of magnitude stronger than the shear strength of the rock joint. In this study we focus on how the distribution of rock bridges influences the overall rock mass behaviour, to contribute to the understanding of how the presence of rock bridges influences the ‘scale’ effect that is observed between strength values measured on intact rock in the laboratory and those observed at the rock mass scale. To estimate the influence of spatial fracture, parameters of the rock mass strength and deformation were determined, using the orthogonal arrays method, UDEC and variance analysis. The numerical model was first calibrated on shear tests of samples made of continuous joints, and then used to investigate the shear behaviour of a fractured rock mass with non-persistent joints. The 2D approach was successfully extended to 3D models using 3DEC with the aim of providing a better approach for simulating the stability of an underground cavern in a fractured rockmass.