Alternative models for analyses of liver and mammary transorgan data were formulated and fitted to liver and mammary data sets respectively. The models considered metabolite inputs to and effluxes from an extracellular pool. In general, fits were greatly improved over previous efforts using other models (Miller et al. 1991a; Hanigan et al. 1992; Wray-Cahen et al. 1997). Errors of prediction were generally less than 15% for liver and less than 20% for mammary glands. With the possible exception of glutamine for the udder, all metabolites exhibited linear responses to extracellular concentrations within the observed ranges of inputs. However, prediction biases were evident for beta-hydroxybutyrate, acetate, and propionate by liver and for arginine, histidine, citrulline and glycerol by mammary tissue. These biases were hypothesized to be caused by the existence of additional regulatory complexity. With the exception of histidine, parameter estimates for essential amino acid removal by liver were 2-3-fold lower than for mammary gland. Infusion of an amino acid mixture into the mesenteric vein did not alter parameter estimates for removal of amino acids by the liver. Treatment of cows with bovine somatotropin resulted in changes in mammary parameter estimates for aspartate, glutamate, leucine, phenylalanine, glucose, and glycerol.