Thomsen-Friedenreich (TF) antibodies were prepared from human serum by different enrichment procedures. This resulted in three antibody preparations all of which agglutinated neuraminidase-treated erythrocytes. On the other hand, each of the three antibody populations showed a distinct specificity pattern. Anti-TF1 antibodies could be inhibited in the hemagglutination inhibition assay by asialofetuin, asialotransferrin, asialoglycophorin and asialomucin. The sialylated form of these glycoproteins showed no inhibition. No significant inhibition could be achieved with several mono- or disaccharides. This suggests that anti-TF1 recognizes common structures on glycoproteins normally hidden by sialic acid. Anti-TF2 antibodies showed specificity for asialofetuin, bovine submaxillary mucin, asialomucin, asialoglycophorin, the disaccharide gal-beta (1-3)N-acetyl-galactosamine (galNAc) and nitrophenyl-beta-galactoside. Because asialotransferrin or unbound lactosamine were not inhibitory, we suppose that the residual common structure of the inhibitors is (gal)-galNAc-O-Ser/Thr, which is present in high amounts in submaxillary mucin. Anti-TF3 antibodies were inhibited by asialoglycophorin but not by asialomucin or asialofetuin. Strong saccharide inhibitors were gal-beta (1-3)galNAc, nitrophenyl-beta-galactoside, as well as galactose. Therefore, both of the antibody preparations, anti-TF2 and anti-TF3, could be inhibited by gal-beta-(1-3)galNAc, but showed preference to one or the other sugar component of the disaccharide resulting in a differential recognition of glycoconjugate inhibitors. Anti-TF2 and anti-TF3 seem to recognize the carbohydrates in the context of a protein backbone, because gal-beta-(1-3)galNAc in connection with a ceramide backbone (GM1) was not inhibitory. When tested on three human breast cancer cell lines, only anti-TF2 recognized epitopes exposed on the cell surface. We, therefore, conclude that human serum contains at least three subpopulations of TF antibodies with distinct specificities. Only anti-TF2 can detect cryptic erythrocyte epitopes which are also exposed on human breast cancer cell lines.