Chloride reduces the oxygen affinity of mammalian haemoglobin by acting as an allosteric effector that stabilizes the quaternary deoxy (T) structure. Perutz and others showed evidence that it does so by neutralizing electrostatic repulsion by an excess of positive charges in the cavity that runs through the centre of the molecule, but without binding to any specific site. On the basis of this proposal, any amino acid substitutions in the central cavity that halve the number of excess positive charges should halve the chloride effect, neutralization of the excess positive charges should inhibit it and introduction of additional positive charges should enhance it. Charge changes on the surface of the molecule should leave it unaltered. We have tested this proposal by measuring the chloride effects in several abnormal human haemoglobins with replacements of polar residues in the central cavity or on the surface that we happened to come across. They all proved consistent with the proposal. It appears that diffusible electrolytes can modify allosteric equilibria without necessarily binding to any specific site. Our proposal also implies that amino acid substitutions that make the central cavity more electropositive should destabilize the T-structure and therefore increase the oxygen affinity, while substitutions that make it more electronegative should do the reverse. A survey of all substitutions reported in the literature shows that this is true, with a few exceptions due to special stereochemical effects.