At the macroscopic scale, the process of brittle fracture, leading to the formation of a macroscopic fault, involves several stages of fracture development. The usual sequence of events consists of the following stages: load-parallel tensile fracture, load-normal shear fracture, inclined shear fracture, strength failure, second order fracture and faulting. The first two stages can be defined quantitatively in terms of an elastic flaw concept which is based on the elastic stress distribution existing around the flaw prior to the appearance of the first fracture. The appropriate theories are developed through a stress model which, in departure from other brittle fracture theories, involves the stress gradient along the fracture path. Experimental results are introduced to demonstrate the general validity of the stress gradient model. The initiation of the first inclined shear fracture seems to be contingent on the existence of a quasi-cataclastic state produced by the earlier sets of load-parallel tensile and load-normal shear fractures. Because the original elastic stress distribution, including the stress anomaly around the flaw, is destroyed before the initiation of the inclined shear fracture, later events of fracture cannot be interpreted in terms of the elastic flaw model• Probably a Coulomb-type mechanism, similar to the one involved in the failure of cohesionless, granular materials, controls the initiation and propagation of inclined shear fractures as well as the development of maximum resistance.