Frictional discontinuities in rock masses can be viewed as fractures and their behavior can be approximated using Linear Elastic Fracture Mechanics (LEFM) theory. Slip along a frictional discontinuity can be approached as initiation and propagation of a mode II fracture along its own plane. Fracture mechanics theories predict that under pure mode II loading initiation will occur when the energy release rate of the fracture attains a critical value (G_IIC), which is generally taken as a material property. The research conducted shows that this may not be always the case. Identification and quantification of the mechanisms for the onset and propagation of slip along frictional surfaces have been carried out by testing in biaxial compression two different sets of brittle specimens: acrylic and gypsum. The specimens consist of two or three blocks with perfectly mated contact surfaces. The contact surface between blocks is equally divided in two areas, one with a lower frictional strength (weak area) and the other one with a higher frictional strength (strong area). Experiments show that slip starts first in the weak area and progresses towards the strong area with increasing load. Once slip has reached the strong area, a sharp contact is created between the area that has slipped (weak) and the area that has not (strong). Results from the tests show that the critical energy release rate, G_IIC, depends on the frictional characteristics of the surface and on the critical displacement required to decrease the frictional strength from peak to residual. Furthermore, experiments conducted on surfaces with and without cohesion indicate that cohesive debonding and frictional mobilization may not occur simultaneously. The experimental results together with an analytical formulation within the framework of fracture mechanics explain some of the inconsistencies found in the literature and provide a clear picture of how slip initiates. This paper presents the experiments, analyses and formulations carried out in support of the conclusions.