This paper extends the spring-slider model of Segall and Rice <1995> for dilating fluid-infiltrated faults to include geometry-induced normal stress variations. Because of the coupling between normal and shear stresses on a dipping plane the shear stress drop accompanying slip is associated with a decrease in normal stress for reverse faulting and with an increase for normal faulting. This coupling stabilizes frictional sliding on normal faults and destabilizes it on reverse faults. When a fluid phase is present, these normal stress variations during slip events also trigger, through poroelastic coupling, compensating pore pressure changes: stabilizing pressure decreases for reverse geometry and destabilizing increases for normal geometry. Fault stability is quantitatively studied using two parameters: the critical spring stiffness kcr separating the stable and unstable slip domains and, when numerical simulations exhibit stick-slip cycles, the stress drop during the slip phases Δ&tgr;. As expected, geometric coupling is found to reduce kcr and Δ&tgr; for normal faults (in comparison with Segall and Rice's <1995> strike-slip geometry) and to increase these parameters for reverse faults. Poroelastic coupling shows opposite consequences. The combined effect of shear-induced dilatancy and poroelastic coupling is stabilizing for reverse geometry but can be either stabilizing or destabilizing for normal geometry, depending on the dilatancy magnitude. Nevertheless, the stability differences between the types of faults are primarily controlled by the geometric coupling between shear and normal stresses: For all the parameters we investigated, normal faults exhibit lower stress drops (and critical stiffnesses) than reverse faults. Finally, extrema appearing in the graphs featuring Δ&tgr; or kcr versus the diffusion rate demonstrate that the influence of the fluid effects is not maximum in the undrained case. Surprisingly, their influence is maximum at a point intermediate between the drained and undrained limits, when the diffusion time is roughly equal to the frictional relaxation time associated with stick phases. This phenomenon is presumably caused by subtle second-order mechanisms involving the instantaneous pore pressure changes induced by the poroelastic coupling in response to instantaneous friction changes and the interseismic pressure recovery process. ¿ 2001 American Geophysical Union