This paper presents the results of an analytical and experimental study on the performance of reinforced concrete beams subjected to pure torsion. The main parameters of the experimental tests were amount of torsional reinforcement and the ratio of the transverse torsional reinforcement to the longitudinal torsional reinforcement. The test results indicated that the maximum amount of torsional reinforcement required in ACI 318-05 code underestimated almost twice as much as the observed maximum amount of torsional reinforcement. Comparisons between the tested and calculated torsional behaviors of the 102 beams showed that the torsional failure modes of ACI 318-05 code disagreed with the observed failure modes. In addition, the torsion provisions in ACI 318-05 code overestimate the torsional strength of the RC beams in which relatively large amount of torsional reinforcement were reinforced, while underestimate for the beams with small amount of torsional reinforcement. This discrepancy between the theoretical ultimate torsional strength as given by the ACI 318-05 code and the experimental one can be due to neglecting the tension stiffening effect and the contribution of the torsional strength by concrete.