We advance an hypothesis which explains the presence of anti-A1 in group A2 individuals and of anti-B1 in group Bv individuals without assuming that A1 and A2 receptors are different or that B1 and Bv receptors are different. Taking into account the known fact that the receptor-density is higher in A1 and A2 erythrocytes, the hypothesis states that anti-A1 antibodies have a low affinity for A receptors and cannot form firm bridges between red cells when they use only one valency per cell. Consequently they do not agglutinate A2 cells, on which the density of receptors is too low. They do agglutinate A1 cells, on which the receptors are closer together. Thus subtype-specific antibodies must be multivalent, and bivalent antibodies such as IgG cannot distinguish between high- and low-receptor-density cells. A corresponding explanation may be valid for subtypes of B and anti-B1. Our results, as far as they go, support this hypothesis. Although normal anti-A and anti-B isoagglutinins contain a small proportion of IgG molecules, we could not demonstrate any IgG in isoagglutinins specific for subtypes A1 and B1. Anti-A1 reagents could be prepared from the purified IgM fraction of anti-A sera but not from the IgG fractions of the same sera. Oligosaccharides with blood group A or B specificity strongly inhibited agglutination when the test cells had a low density of receptors (A2B or Bv). They were much poorer inhibitors of agglutination when the test cells had a high density of receptors (A1 or normal B). Macromolecular salivary blood group substances were used as inhibitors in parallel tests. They inhibited strongly regardless of the receptor density on the erythrocytes. The use of A2B rather than A1, and Bv rather than B1, erythrocytes in agglutination-inhibition tests for blood group specific oligosaccharides improved the method by making it more sensitive. According to the above concept, anti-A1 or A2 individuals may possess an autoantibody directed against their own A receptors. The problems arising from this situation are discussed.