Controversy exists concerning the localization of the enzyme Na+,K(+)-ATPase to canalicular membranes in hepatocytes. Most studies find enzyme activity only at the basolateral plasma membrane domain of the hepatocyte. However, Na+,K(+)-ATPase activity has been detected recently in a canalicular membrane fraction prepared by Mg++ precipitation, suggesting that differences in membrane domain fluidity account for these discrepancies. To reinvestigate this question, we used free-flow electrophoresis to further purify canalicular liver plasma membranes originally separated by sucrose density centrifugation. With this technique, canalicular membranes devoid of Na+,K(+)-ATPase activity by routine assay were separated into six subfractions. More than 80% of the activities of canalicular marker enzymes was recovered in two subfractions closest to the anode, which were totally devoid of Na+,K(+)-ATPase activity. However, Na+,K(+)-ATPase activity could now be detected in the four other fractions that contained only small amounts of canalicular marker enzymes. The basolateral marker enzyme, glucagon-stimulated adenyl cyclase, comigrated with this cryptic Na+,K(+)-ATPase activity. Furthermore, addition of 6 mumol/L [12-(2-methoxyethoxy)-ethyl-8-(cis-2-n-octylcyclopropyl)-octanoate ], a membrane-fluidizing agent, to the original canalicular membrane preparation and to all subfractions did not stimulate or unmask latent Na+,K(+)-ATPase activity. Finally, when canalicular membranes isolated by Mg++ precipitation were subjected to free-flow electrophoresis, they could not be separated from the more positively charged Na+,K(+)-ATPase-containing fractions, probably because of alterations in surface charge. Together these findings suggest that Na+,K(+)-ATPase is a basolateral enzyme, that represents a small contaminant when present in canalicular liver plasma membranes and that methodological differences may account for previous discrepancies.