Many of the secondary fault structures developed during triaxial friction experiments have been generally correlated with the structures of natural fault zones. Therefore, any physical differences that can be found between laboratory samples that slide stably and those that show stick-slip motion may help to identify the cause of earthquakes. We have examined petrographically the run products of many triaxial friction experiments using clayey and quartzofeldspathic gouges, which comprise the principal types of natural fault gouge material. The examined samples were tested under a wide range of temperature, confining and fluid pressure, and velocity conditions. The clayey and quartzofeldspathic gouges show some textural differences, owing to their different mineral contents and grain sizes and shapes. In the clayey gouges, for example, a clay mineral fabric and kink band sets are commonly developed, whereas in the quartzofeldspathic gouges fracturing and crushing of the predominately quartz and feldspar grains are important processes. For both types of gouge, however, and whatever the pressure-temperature-velocity conditions of the experiments, the transition from stable sliding to stick-slip motion is correlated with: (i) a change from pervasive deformation of the gouge layer to localized slip in subsidiary shears; and (ii) an increase in the angle betweem the shears that crosscut the gouge layer (Riedel shears) and ones that form along the gouge-rock cylinder boundaries (boundary shears). This suggests that the localization of shear within a fault zone combined with relatively high Riedel-shear angles are somehow connected with earthquakes. Secondary fracture sets similar to Riedel shears have been identified at various scales in major strike-slip faults such as the San Andreas of the western United States (Wallace, 1973) and the Luhuo and Fuyun earthquake faults of China (Deng and Zhang, 1984; Deng et al., 1986). The San Andreas also contains locked and creeping sections that correspond to the stick-slip and stably sliding experimental samples, respectively. We plan to study the physical structure of the San Andreas fault, to see if the experimentally observed differences related to sliding behavior can also be distinguished in the field. ?? 1990.