Considerable amounts of data indicate that the thickness of the gouge zone for brittle faults increases linearly with total slip. This implies that gouge is produced continuously and at a steady rate as faulting progresses. A simple model of wear which accounts for gouge formation predicts a constant thickness/slip ratio for a given rock type and normal stress. The model is compared with measurements of gouge production in friction-sliding experiments, and there is satisfactory agreement. Natural faults have higher wear rates than found in the experiments, probably because of their greater roughness. Estimates are also made of the fraction of the energy of faulting that goes into gouge production, and it is found to be 10−3–10−4 of the frictional work in both the experimental and natural cases. The model predicts that fault zones should thicken with depth in simple cases.