The hypothesis that insulin is internalized into the hepatic Golgi apparatus was tested by the diaminobenzidine-shift protocol of Courtoy et al. (1984, J. Cell Biol. 98, 870). Highly purified Golgi fractions were isolated after the coinjection of [125I]insulin and the synthetic ligand, galactose-bovine serum albumin-horseradish peroxidase. Golgi fractions were subsequently reacted in the presence or absence of diaminobenzidine, then subjected to Percoll gradient centrifugation. For incubations carried out in the absence of diaminobenzidine, [125I]insulin-containing components were found at a low density (peak density congruent to 1.042) identical to that of the Golgi marker enzyme galactosyltransferase. However after incubations carried out in the presence of diaminobenzidine, the majority of [125I]insulin-containing components was shifted to a higher density of greater than 1.06 while that of galactosyltransferase remained unchanged (peak congruent to 1.042). These observations indicate that the majority of internalized insulin is not located in galactosyltransferase-containing Golgi components.