A hypothesis is put forward to explain the apparent monovalency of human IgG4. It is based upon the known instability of the IgG4 hinge. IgG4 is secreted as a regular bivalent antibody, but after secretion interacts with another IgG4 molecule. This interaction results in the exchange of half molecules (a combination of one heavy chain and one light chain) between the two IgG4 molecules. The postulated bispecific IgG4 antibodies can indeed be found in selected human sera following repeated immunization with two non-crossreacting antigens.