The transition-state theory of exchange-only membrane transport is applied to experimental results in the literature on the anion exchanger of red cells. Two central features of the system are in accord with the theory: (i) forming the transition state in translocation involves a carrier conformational change; (ii) substrate specificity is expressed in transport rates rather than affinities. The expression of specificity is consistent with other evidence for a conformational intermediate (not the transition state) formed in the translocation of all substrates. The theory, in conjunction with concepts derived from the chemistry of macrocyclic ion inclusion complexes, prescribes certain essential properties in the transport site. Separate subsites are required for the preferred substrates, Cl- and HCO3-, to account for tight binding in the transition state (Kdiss congruent to 1 microM). Further, the following mechanism is suggested. A substrate anion initially forms a loose surface complex at one subsite, but in the transition state the subsites converge to form an inclusion complex in which the binding forces are greatly increased through a chelation effect. The conformational change at the substrate site, which is driven by the mounting forces of binding, sets in train a wider conformational change that converts the carrier from an immobile to a mobile form. Though simple, this composite-site mechanism explains many unusual features of the system. It accounts for substrate inhibition, partially noncompetitive inhibition of one substrate by another, and "tunneling," which is net transport under conditions where exchange should prevail, according to other models. All three types of behavior result from the formation of a ternary complex in which substrate anions are bound at both subsites. The mechanism also accounts for the enormous range of substrate structures accepted by the system, for the complex inhibition by the organic sulfate NAP-taurine, and for the involvement of several cationic side chains and two different protein domains in the transport site.