In this model for the evolution of large crustal faults, water that originally came from the country rock saturates the initially highly porous and permeable fault zone. During shearing, the fault zone compacts and water flows back into the country rock, but the flow is arrested by silicate deposition that forms very low permeability seals between the fault zone and the country rock. Because of variations in temperature and mineralogical composition and the complex structure of the fault zone, a three-dimensional network of seals is formed in the fault zone itself; thus, the high-pressure fluid is not evenly distributed. As in deep oil reservoirs, the fluid will be confined to seal-bounded fluid compartments of various sizes and porosity that are not hydraulically connected with each other or with the hydrostatic regime in the country rock. When the seal between two of these compartments is ruptured, an electrical streaming potential will be generated by the sudden movement of fluid from the high-pressure compartment to the low-pressure compartment. When the pore pressure in the two compartments reaches its final equilibrium state, the average effective normal stress across them may be lower than it was initially, and, if the two compartments are large enough, this condition may trigger an earthquake. During an earthquake, many of the remaining seals will be ruptured, and the width of the fault zone will increase by failure of the geometric irregularities on the fault. This newly created, highly porous and permeable, but now wider fault zone will fill with water, and the process described above will be repeated. Thus, the process is an episodic one, with the water moving in and out of the fault zone, and each large earthquake should be preceded by an electrical and/or magnetic signal.