Stress pulses are generated whenever a brittle fracture occurs, and this paper first describes the phenomena observed when a Hertzian fracture is produced by the impact of a spherical indentor on a block of glass. The block is sufficiently large to be regarded as a half space, and the wave propagated over the surface of the glass is diverging from the region of impact and consists of a combination of dilatation, distortional and Rayleigh surface waves. At large distances of travel, the last of these predominates since it diverges cylindrically, while the others are diverging spherically. Satisfactory agreement was observed between the wave shapes predicted theoretically and those observed experimentally. The next problem was concerned with the fracture of glass rods in simple tension, here the fracture was initiated at a scratch on the surface of the rod and diverged cylindrically across the rod at high velocity from the point of initiation. The growth of the fracture results in the propagation of two pulses one longitudinal in nature and the other flexural. The shape of these pulses was measured by means of strain gages fixed to opposite sides of the surface of the rod parallel to the diameter which contained the point of initiation. In order to account for the shapes of these pulses theoretically, some assumption had to be made about the stress distribution over the cross section during the fracture process. The stress on the fractured portion clearly vanished, and it was assumed that since the fracture growth was so rapid it would be reasonable to assume that the stress in the unfractured portion retained the value it had before fracture commenced. On this assumption the shapes of the longitudinal pulse and the flexural pulse could be predicted, and from these the records of the strain pulses to be expected from the two strain gages could be determined. It was assumed that, the longitudinal pulse was propagated without change in form, while the propagation of the flexural pulse was calculated by the use of the Timoshenko theory. Excellent agreement between theory and experiment was obtained. Experiments were next carried out on the fracture of thin rods in flexure, and it was found that the observed pulses had very much longer durations. The type of treatment used for extensional fracture could here clearly not account for the whole phenomenon, since it would appear to predict that fracture started at the surface, ran to the neutral axis, and then stayed there permanently. In fact it was found that with short specimens the fracture rapidly approached the neutral axis and stayed in the neighborhood of this axis until an unloading wave arrived from the end of the rod. As longer and longer rods were employed the lengths of the pulses increased, and where very long rods were used, it could be seen that the stress was continuing to decrease even in the absence of reflected pulses from the ends. The program of experimental and theoretical work on the problems is continuing.