The association of a significant increase in seismic activity with the filling of some large water reservoirs is well documented. One possible such ase is the earthquake sequence at Oroville, California, in 1975. It has been suggested that this activity may be triggered on faults close to failure by the weight of the water, the increase in pore pressure, or both. Two-dimensional half-space models with surface loading are used to study the change in strength produced by these mechanisms for thrust, normal, and strike slip faulting. For the case of the water weight alone the strength increased across fault planes dipping as in the Oroville sequence. It siconcluded that the water load alone is an unlikely mechanism. The inclusion of pore pressure effects in these models, however, is shown to be very encouraging. Using the Biot linearized quasi-static elasticity theory for fluid-infiltrated porous materials, two types of loaded half-space models were studied. The first was homogeneous with respect to permeability and found to produce broad zones of weakening. The second type included a fault zone of different permeability. This model intensifies the magnitude and broadens the zone of strength drop if the fault zone is highly permeable in relation to the rest of the half space. For fault dips roughly corresponding to the Oroville situation it was found that although weakening occurs for ambient stresses favoring either thrust or normal faulting, the magnitude and breadth of strength drop were significantly greater for normal faulting. Similar effects were observed for strike slip faulting. It was also found that all types of models developed significant zones of weakening very rapidly, permitting reasonable values of permeability to be estimated. The case of the impervious reservoir bottom was examined, with significantly differing results. Just as in the permeable case, zones of weakening were initially developed. However, in the impermeable case these dissipated with time, eventually reaching the same strength distribution as that of the water weight alone. Thus with sufficiently slow impoundment the impervious reservoir may not weaken at all, whereas the permeable reservoir will continue to weaken in time. The results also show that rapid unloading may casue instantaneous crustal weakening owing to excess pore pressure. Although the models are very simplified, it is interesting to apply them to the case of Lake Oroville. Using reasonable parameters, strength drops for normal faulting of 8--14 bars were calculated, which compared favorably to the fluid injection experiments at Rangely and Matsushiro. The observed lag time between initial impoundment and seismicity yielded a permeability of 0.2 mdarcy for the homogeneous model and 20 mdarcy for the greater strength drops produced by the heterogeneous model, both of which are reasonable rock values.If the period of rapid filling is assumed to have triggered the activity, then the permeability was found to be 3.6 mdarcy for the homogeneous mode and 360 mdarcy for the heterogeneous model, the latter of which is appropriate for fractured rock. It is concluded that the pore pressure trigger mechanism is very promising and that the methods of this paper may be extended to three dimensions to study reservoirs in more detail. The mechanism is limited, however, by the lack of knowledge of, first, the ambient pore pressure in situ and, second, the in situ permeability, coupling parameters, reservoir bottom permeability, and local tectonic stress. Perhaps the most important implications of the result is the presence of water to depths of 10 km or more in the crust. Thus at least in some places the crust may be weak, and tectonic stresses low. This also suggests an important potential for geothermal energy at great depth in the crust.