The time variation of crustal velocities in tectonic regions is most reasonably attributed to stress-induced variations in crack porosity. The decrease in Vp/Vs before earthquakes is due primarily to a large decrease in Vp. This supports the Nur dilatancy hypothesis but not the effective stress hypothesis. New data from the San Fernando region verify the Vp drop, show that this drop cannot be entirely due to source depth effects, and give strong support to the explanation of material property, or path effect, rather than source effect variations. Calculations show that the crack-widening model works even for mid-crustal depths in saturated rock. Narrow cracks of low aspect ratio are required to satisfy the velocity and uplift constraints. The recovery of velocity prior to fracture can be due to fluid flow or crack closure. The t~L2 relation does not require diffusion. Diffusion of groundwater or crack closure leads to increased pore pressure and rock weakening. Observations of gravity, conductivity, and crustal distortions along with velocities should narrow the choice of models. The crust in regions of thrust tectonics is probably always dilatant to some degree. The aftershock region is smaller than the anomalous velocity region, which in turn must be smaller than the dilatant region. A simple relationship is derived for the relative sizes of the anomalous and aftershock regions.