The best means to utilize X-ray computed tomography (CT) and ultrasound to predict carcass lean, fat and bone weights in vivo in terminal sire sheep were tested. Data on 160 lambs from three breeds were considered: 50 Suffolk males, 50 Suffolk females, 40 Texel males and 20 Charollais males. One-fifth of the lambs within each breed and sex group were slaughtered at each of 14, 18 and 22 weeks of age and the remaining two-fifths at 26 weeks. Carcasses were dissected into lean, fat and bone weights. Prior to slaughter all lambs were CT scanned, with cross-sectional scans taken at seven sites along the body (ischium, hip, mid shaft of femur, 2nd and 5th lumbar vertebrae and 6th and 8th thoracic vertebrae), and ultrasound scanned at the 3rd lumbar vertebra and 13th rib.A set of three CT scans that reliably predicted carcass lean, fat and bone weights was identified which included a scan in each of the three main carcass regions: ischium in the hind leg, 5th lumbar vertebra in the loin and 8th thoracic vertebra in the shoulder. Breed and sex affected the intercepts of the prediction equations but not their slopes. Therefore, a minimal set of equations is likely to be sufficient to predict tissue weights, at least within terminal sire sheep breeds. Equations derived showed high degrees of fit to the data with R2 values of 0·924, 0·978 and 0·830 for lean, fat and bone weights, respectively, when predicted using CT alone, and 0·589 and 0·857 for lean and fat weights, respectively, when predicted using ultrasound alone. Using live weight in addition to CT information only improved prediction accuracy slightly for lean (0·966) and fat (0·986) although more substantially for bone (0·925). Where live and tissue weights are considered contemporaneously in genetic evaluations, excluding live weight from prediction may therefore be preferable to avoid colinearity among weight measures.