Results obtained with the in vitro Wilson-Wiseman method [1954), J. Physiol. 123, 116-125) and with an in vivo perfusion model are compared. The in vitro method suffers from insufficient oxygenation. Five plasma membrane enzymes, five cytosol and two lysosomal enzymes were estimated. The medium consisted of Ringer solution without and with 10 mmol/l taurocholate. The greatest difference between both models is shown by the lysosomal enzymes, which are amply released in vitro but not in vivo. The in vitro to in vivo ratios of the release rates of cytosol enzymes vary from about 10 to less than one suggesting deterioration by lysosomal cathepsins in certain cases. The removal of plasma membrane enzymes, particularly those that are characterized by superficial localization, is inhibited by hypoxia. This implies that the normal release of these enzymes, at least in part, is energy-dependent.