The close association of overpressured shales and the coincident flattening of listric normal faults has been noted and is usually explained in terms of gravity-sliding models. However, incorporation of overpressure into Mohr-Coulomb frictional-faulting theory does not predict normal faults dipping at angles less than 45°, even for very low coefficients of friction. We suggest that low-angle normal faults may form because of a rotation of the maximum principal stress away from vertical with depth. Such a rotation occurs if a significant contrast in viscosity exists between two adjacent strata, such as a normally pressured sandstone and an underlying overpressured shale or evaporite. The stress-refraction model renders the presence of a preexisting fault plane or an extremely low coefficient of friction unnecessary to explain listric fault geometry and offers a failure mechanism in agreement with frictional-faulting theory based on rheological arguments.