Percentage of lean, fat, and bone were determined in 18 bob veal (BV) and 28 special-fed veal (SFV) carcasses. Carcasses were subjected to a set of visual conformation scores and a variety of physical measurements. No significant differences were found regarding carcass percentage of lean, fat, and bone within the three BV weight groups (P > .05). On average, SFV were 12% fatter than BV and did not have a greater percentage of lean (P > .05), except for SFV carcasses weighing 88.2 to 97.7 kg. Bob veal had less fat (internal, external, and intermuscular) and a higher bone percentage than SFV (P < .05). The round and shoulder primals had the greatest proportion of lean in both the BV and SFV carcasses. Bob veal carcasses had an average conformation score of average Good and SFV carcasses had an average conformation score of average Choice. In addition, a parsimonious subset of variables was identified for predicting total percentage of lean (TPLEAN) for both BV and SFV separately, using "stepwise" regression model building procedures. For BV, all four identified predictor variables were subjective conformation scores (i.e., muscling, appearance, leg thickness, loin-back plumpness) (R = .73, P < .03). For SFV, four predictor variables were also identified: kidney and pelvic fat, fat thickness, carcass length, and lateral thickness (R = .61, P < .03). Although both regression equations were significant predictors of TPLEAN, confidence limits for predicting future TPLEAN value were wide relative to the variation in the actual TPLEAN values. Thus, the practical utility of the regression equations is limited.