I ntact human erythrocytes were cross-linked with glutaraldehyde (GA) or dimethyladipimidate (DMA) and tested for their ability to bind [125I]-IgG anti-Rh0(D) and to undergo antibody-mediated hemagglutination. There was no decrease in antibody binding after treatment with GA concentrations up to 1.25% and DMA concentrations up to 1%. Red cells treated with these concentrations of GA and DMA did not agglutinate. The techniques employed to induce agglutination of the cross-linked red cells involved "incomplete" IgG anti-Rho (D) in albumin, "complete" IgM anti-D Rho (D) in saline, and the antiglobulin (Coombs) reaction. The agglutinability of the chemically modified red cells was inversely correlated with the extent of fixation. The dissociation of antibody binding from agglutinability in cross-linked erythrocytes suggests that Rho (D) antigen mobility is required for red cell agglutination. Antigen mobility was manifested by the transition from a relatively monodisperse distribution pattern of Rho (D) antigen sites to one of large aggregates or clusters when agglutination was induced by IgM anti-Rho (D), IgG anti-Rho (D) agglutination of protease modified red cells, and by anti-IgG agglutination of IgG anti-Rho (D)-coated red cells. Antigen clustering was not as prominent in red cells agglutinated by IgG anti-Rho (D) in the presence of albumin. Even though antigen mobility is a prerequisite for antibody-mediated hemagglutination, clustering does not appear to be an absolute requirement. The degree of antigen clustering differs with varying types of agglutination.