Induced seismicity is a major geological hazard associated with fluids injection (water and gases) into deep geological strata. The induced pore pressure due to injection may fracture intact rock, or more likely, trigger slip across preexisting geological faults. The critical pore pressure required to trigger slip has typically been predicted based on Terzaghi’s effective stress concept and Coulomb’s law of friction. However, faults are known to be anything but smooth, with roughness variations controlling the topography of the fault surface over many decades of length scales. Fault roughness has been shown to control the peak shear resistance of faults, the shear and normal stiffnesses of the fault, and the characteristics of stick slip oscillations. One of the major parameters that is influenced by fault roughness during slip is the dilation across the fault plane during shear. It would be intuitively expected that once the fault begins to dilate the pore pressure at the fault aperture will decrease, thus strengthening the fault, yes experimental evidence suggest that this is not always the case. In this study we first explore the micromechanics of stick slip oscillations using direct shear tests of interfaces 10 cm in length with particular emphasis on dilation through shear. We find that the amplitude of stick slip oscillations strongly depends on roughness under a wide range of normal stresses, as well as the dilative tendency of the fault. We continue with triaxial tests of inclined faults under controlled pore pressure conditions and find that the flow pattern across the fault plane also strongly depends on the initial roughness, as well as the stress drop and the slip velocity at the point of incipient failure. We conclude that initial roughness characteristics must be incorporated in predictions of the critical pore pressure required to induce slip across geological faults.