We have studied the role of restrictions to lateral mobility in the segregation of proteins to apical and basolateral domains of MDCK epithelial cells. Radioimmunoassay and semiquantitative video analysis of immunofluorescence on frozen sections showed that one apical and three basolateral glycoproteins, defined by monoclonal antibodies and binding of beta-2-microglobulin, were incompletely extracted with 0.5% Triton X-100 in a buffer that preserves the cortical cytoskeleton (Fey, E. G., K. M. Wan, and S. Penman. 1984. J. Cell Biol. 98:1973-1984; Nelson, W. T. and P. J. Veshnock. 1986. J. Cell Biol. 103:1751-1766). The marker proteins were preferentially extracted from the "incorrect" domain (i.e., the apical domain for a basolateral marker), indicating that the cytoskeletal anchoring was most effective on the "correct" domain. The two basolateral markers were unpolarized and almost completely extractable in cells prevented from establishing cell-cell contacts by incubation in low Ca++ medium, while an apical marker was only extracted from the basal surface under the same conditions. Procedures were developed to apply fluorescent probes to either the apical or the basolateral surface of live cells grown on native collagen gels. Fluorescence recovery after photobleaching of predominantly basolateral antigens showed a large percent of cells (28-52%) with no recoverable fluorescence on the basal domain but normal fluorescence recovery on the apical surface of most cells (92-100%). Diffusion coefficients in cells with normal fluorescence recovery were in the order of 1.1 x 10(-9) cm2/s in the apical domain and 0.6-0.9 x 10(-9) cm2/s in the basal surface, but the difference was not significant. The data from both techniques indicate (a) the existence of mobile and immobile protein fractions in both plasma membrane domains, and (b) that linkage to a domain specific submembrane cytoskeleton plays an important role in the maintenance of epithelial cell surface polarity.