A study of 38 beef sides (total dissected fat 16.5%-33.0%) showed that densities of forequarter, hindquarter, and side, each predicted percentage carcass components with less accuracy than 12th rib fat thickness. Hot carcass weight added to density of side improved the predictions of fat and bone only slightly (standard errors of estimate of 3.45% and 0.99%, respectively) over density of side alone. Density of primal cuts or combinations of primal cuts predicted carcass composition with lower standard errors of estimate than any of the above covariates.The addition of fat thickness at the 12th rib and carcass weight to the most accurate density measurements (loin and rump) did not significantly improve the prediction of any carcass component.Hot carcass weight was not related to the density of any carcass part. Side fat percentage was related to the density of all carcass parts except shin, and covariance analysis showed that breed effects remained in forequarter, side, and rump. Density of hindquarter, loin, and topside were free from breed effects.Density of side alone, or together with hot carcass weight, did not predict carcass composition with sufficient accuracy to be considered valuable for commercial adaptation.