In numerical modeling of uncemented soils, Rowe's stress-dilatancy relation for frictional materials was widely used and supported by a large body of experimental results. Even though the derivation of the theory was questioned because the principle of minimum energy no longer applies when friction is involved, De Josselin de Jong (1976) did prove, with an alternative approach based on the laws of friction, that Rowe's final conclusions and his stress-dilatancy relation were valid. Besides frictional materials, Rowe (1962) also proposed a stress-dilatancy relation for cohesive-frictional materials. Recently, researchers applied this relation for cemented sands. Although the Rowe's relation for frictional materials was proved to be correct, the validity of his stress-dilatancy relation for cohesive-frictional materials was never checked. This must be done before the application of that relation to model cohesive-frictional materials. This paper will show that the Rowe's stress-dilatancy relation for cohesive-frictional materials is not correct. A correct stress-dilatancy relation for cemented sands will be proposed, which was derived using the friction laws used by De Josselin de Jong (1976).