ABSTRACTCarcass measurement and dissection data for 1176 pigs were used to examine the suitability of different measurements for predicting carcass lean percentage in commercial classification and grading. Three carcass weight groups with means of 47, 72 and 93 kg, were analysed separately. Similar results were obtained and those for the 72 kg group are given in this abstract. The s.d. of percentage carcass lean in this group was 3·84. Visual conformation score, carcass length and m. longissimus depth were valueless as predictors (residual s.d. of carcass lean percentages were 3·84, 3·77 and 3·80 respectively). Fat thickness measurements taken on the exposed surface of the split carcass provided a less precise prediction (best residual s.d.=2·89) than fat measurements taken over the m. longissimus using the optical probe (best residual s.d. = 2·20). The probe measurements were taken 6·5 cm from the dorsal mid line at fixed positions along the carcass from 4/5th cervical vertebrae to the 5/6th lumbar vertebrae. Precision of prediction showed a regular pattern with highest levels at the last rib (the P2 position currently used in the Meat and Livestock Commission classification scheme) and at the 13th rib, with lower values anteriorly and posteriorly. These two measurements in multiple regression were more precise than any other pair of measurements for predicting both percentage carcass lean (residual s.d. = 2·16) and the lean percentage in individual primal joints.