The Danish MFA probe and the Ulster probe (UP) automatic recording instruments developed recently for predicting carcass composition in pig classification and grading schemes, were tested in comparison with the optical probe (OP) used at present for this purpose in Great Britain. A series of four fat and two muscle measurements, including the principal ones used in the schemes operated in Denmark, Ulster and Great Britain, was taken on 158 hot carcasses covering the pork, bacon and heavy weight ranges. The standard deviation of carcass weight ( W) and percentage lean in carcass were, respectively, 18·9 kg and 4·09 (computed over all data) and 4·51 kg and 3·51 (pooled within weight ranges). There was little difference between instruments in the precision with which they predicted percentage lean, either within weight groups or overall. Residual standard deviations computed over all data for the multiple regression of W and the best individual measurement (fat depth over the M. longissimus at the 3rd/4th last rib) were 2·24 (MFA), 2·44 (UP) and 2·23 (OP). Addition of other measurements to the regression provided only a major improvement in precision. Residual standard deviations for W and all six measurements were 2·15 (MFA), 2·35 (UP) and 2·05 (OP). The relative merits of the three instruments for use in British abattoirs are discussed.