Laboratory tests confirm the accuracy of the compliance equations for wedge-loaded, linearly elastic, double cantilever beam test specimens used for the measurement of fracture energy GI but show that there are significant discrepancies with theory in tests on rock specimens of the same design. The dependence of the compliance on the length of the crack in the test specimen is not correctly predicted by theory for the experiments done on rock. The axial load applied to the arms of the double cantilever beam as a result of wedge loading reduces Young's modulus by as much as 44% and decreases the measured elastic anisotropy of specimens of granite. The experiments show that useful measurements of GI can be made on rock provided that the Young's modulus used in the determination of GI is measured on the same specimen under the same conditions of loading as are used in the fracture experiments. Values of GI for Sioux quartzite determined this way are within 12% of GI determined from the toal work done on the specimens in the fracture experiments. Measurements on nine different types of rock show that a tensile crack must advance as much as 50 grain diameters at a constant speed before a steady state value of GI is attained. Hence methods of measuring GI based on crack initiation gives results that are both variable and different from the steady state value of GI. The range of GI among the nine different types of rock tested with the wedge-loaded double cantilever beam is from 40 to 260 J/m2, which is large compared to any error in the reduction of the data due to the assumption of linear elasticity.