Carcasses of 154 hogs representing seven genotypes with substantial variation in carcass composition and percentage of lean were completely dissected and analyzed. Measurements from a ruler, Hennessy probe, and electromagnetic scanner were each used to predict wholesale and lean boneless carcass value. Error, defined as dissected value minus predicted value, due to the omission of sex, genotype, weight, and their interactions was estimated for each model. The errors were significantly different from zero for the models using ruler and electromagnetic scanning measurements separately (P < .01). Errors due to sex, genotype, weight, and their interactions were greatest for the less lean barrows. A combination of probe and electromagnetic scanner measurements resulted in the least error. The value of barrows with low percentage of lean was consistently overpredicted, whereas the value of leaner gilts was underpredicted for the models using ruler and electromagnetic scanning separately (P < .001).