In this study, a cylinder embedded within cemented soils was used to cause directly tensile failure of cemented soils. An existing dumbbell type direct tensile test and a split tensile test that is most general indirect tensile test were also carried out to verify the developed built-in cylinder tensile test. Testing specimens with two different sand/cement ratios (1:1 and 3:1) and two curing periods (7 and 28 days) were prepared and tested. Total 10 specimens were prepared for each case and their average value was evaluated. Unconfined compression tests were also carried out and the ratio of compressive strength and tensile strength was evaluated. The tensile strength determined by built-in cylinder tensile test was slightly higher than that by dumbbell type direct tensile test. The dumbbell type test has often failed in joint part of specimen and showed some difficulty to prepare a specimen. Among three tensile testing methods, the standard deviation of tensile strength by split tensile test was highest. It was shown that the split tensile test is applicable to concrete or rock with elastic failure but not for cemented soils having lower strength.