Shearing of rock joints is a critical failure mode in rock masses. The shear strength of rock joints must, therefore, be considered in the design of structures in rock masses. Several criteria for prediction of shear strength have been proposed over the years. However, the possible effect of scale on shear strength is an issue. One possible reason for this is that previous experimental studies on the scale effect contain various sources of uncertainties (mixed test methods, multiple testing of same specimen, application of results to other materials than tested, and omitted handling of statistical dispersion). In this paper, the results from a uniquely comprehensive experimental laboratory program, that handles these uncertainties and also extends the range of previously tested conditions, is presented. 46 direct shear tests on two joint types, natural and tensile induced granite rock joints, have been performed under the constant normal stress and the constant normal stiffness boundary condition at 5 MPa initial normal stress applied over three specimen sizes (35 mm × 60 mm, 70 mm × 100 mm and 300 mm × 500 mm). Analysis of variance shows no effect of the specimen size on the shear strength, whereas the joint type and boundary condition has. Quantitative estimates of the influence of the joint type and boundary condition on the shear strength are presented. A consistent approach for determination of shear strength from the point of time associated with a shear stiffness change of the test system is also presented.