The compressive failure of soda-lime glass under uniaxial shock loading has been the subject of much recent discussion. Evidence of failure occurring behind a traveling boundary that follows a shock front has been accumulated and verified in several laboratories. Such a boundary has been called a failure wave. The variations in material properties across this front include complete loss of tensile strength, reduction in shear strength, lowered acoustic impedance and sound speed, and opacity to light. It in many ways resembles a phase boundary. While these observations are generally held to be true, there is no universally agreed mechanism for the process or processes that gives rise to the failure. It is the object of this work to present the results of plate impact experiments that aim to identify the mechanism by which the failure wave propagates. The experiments indicate that the failure is nucleated by the shock wave at surfaces, and that deliberately introducing flaws by roughening the surface speeds the fracture of the material leading to a failure wave close to the shock front. ©1997 American Institute of Physics.