The direct and Brazilian tension tests are the most popular methods for measuring the tensile strength of rocks. While the values of tensile strength obtained from these test are theoretically expected to be identical, the accumulated experimental data show that the direct tensile strength is different from the Brazilian tensile strength in most cases. In an attempt to explore the potential reasons for such discrepancy, the direct and Brazilian tension tests are compared in the light of size effect and bimodularity. The problem of size effect is investigated using the two approaches of statistical theory of strength and linear elastic fracture mechanics. The effect of bimodularity of the material on the induced stresses in the tension tests is also examined. Based on the effect of size and bimodularity, equations have been derived to relate the direct and Brazilian tensile strength. The results of the direct, Brazilian and fracture toughness tests on Lac de Bonnet granite were used to validate the equations.