The origins of shear strength are reviewed in the context of direct shear testing of rock discontinuities. A method of analysing data to correct for dilation over different increments of horizontal displacement throughout a test is demonstrated and various forms of presenting data are discussed. It is argued that the corrected strength envelope for the test as a whole can be used to derive appropriate lower bound strength parameters for design. Additional strength may be allowable to account for roughness at the field scale, but as an alternative approach, it is recommended that the dilation-corrected strength be used directly for design with a lower factor of safety on shear strength than would normally be adopted.