Segregation is important in many bulk handling industries due to its impact on product quality and mixing. Previous studies have focused on quantifying segregation as a coefficient for a particular process. Although this has provided insights into the effect of segregation during particular processes, it has not provided a rational approach to understanding the phenomena of segregation. Toward this end, a vertically oriented segregation shear cell was designed, fabricated and used to test the response of binary mixtures of spherical glass beads. Three size ratios of 10.9:1, 8.7:1, and 5.1:1 were used in this study. Variables such as strain, strain rate (i.e., cycle speed) and bed depth were used to quantify size-segregation characteristics under different input energies. Size ratio was the most dominant variable that affected the percolation rate of fines through a bed of coarse particles. Based on the differences in percolation between the smaller ratio of 5.1:1 and larger ratios of 8.7:1 and 10.9:1, there are indications of a critical size ratio that determines the mechanism of percolation. Two mechanisms were observed. First, the larger size ratios exhibited an initial free-fall discharge of fines at the beginning of the test. This initial discharge was followed by a diffusive behavior, i.e., monotonically increasing trend that approaches an asymptotic value. The smaller ratio did not exhibit the initial discharge and the responses were described adequately by the diffusive behavior. Strain also had an effect. Input energy, as related to strain, was found to be critical in the type of percolation exhibited, i.e., the presence or absence of an initial rapid discharge. For the size ratio of 10.9:1, the amount of input energy was critical because the size of the fine particles was near the size of the coarse bed pore space. When the input energy was below a critical limit, mechanical arching was occurring. Although, the input energy is a function of bed depth and strain rate, these parameters did not influence the results outside random behavior.