Previously we verified experimentally that a generalized form of the Greenwood-Williamson theory describing the elastic contact of two random surfaces adequately predicts the closure of joints in fused silica glass samples. Here we present a similar analysis for experiments on rock. We have tested the theory quantitatively by comparing joint closure from experiment with joint closure predicted by the theory. These comparisons were made using topography data from the surfaces of the joints used in the closure experiments. Three types of tests were done: (1) Granite and quartzite samples were studied where the joints consisted of surfaces ground with ♯80 and ♯60 grits. Here the surfaces were random and uncorrelated with each other. These experiments permitted a direct test of the theory. As with the experiments on glass, the theory agrees with experiment. However, uncertainty in the values of the elastic constants for rock result in larger possible errors in predicting joint closure. (2) Tension fractures in granite and quartzite samples were studied. Here the surfaces were highly correlated with each other. The amount of closure expected for correlated surfaces depends on the degree of correlation, since this determines the effective roughness of the joint. Here the surface topography was repeatedly high-pass filtered and the theory compared with experiment at each stage to determine the effective scale over which the surfaces were mismatched. (3) Ground surfaces of marble were subjected to the same loading conditions. Here the elastic assumption was no longer appropriate. These experiments are an example that the limit of elastic deformation is reached for all materials for surface asperities below a given dimension. This dimension is related to the indentation hardness of the material, which is lower for calcite than for slicate minerals.