The carbon monoxide (CO) binding constants of human myoglobin (Mb) and several single-site mutants have been determined using two different methods. In the kinetic method, which is commonly used for this ligand, the overall association (k(on)) and dissociation (k(off)) rates of CO were measured by flash photolysis and NO replacement, respectively, and the ratio k(on)/k(off) was calculated. In the equilibrium method, the binding constant Keq was measured directly using a thin-layer technique. These two measurements yield similar results for human wild-type Mb but differ significantly for some of the mutants. Possible reasons for these discrepancies are analyzed. A model assuming the presence of interconverting conformers with different association and dissociation rates is considered in light of infrared measurements on the CO stretching frequency in the MbCO forms of the same proteins [Balasubramanian et al. (1993a) Proc. Natl. Acad. Sci, U.S.A. 90, 4718]. It is suggested that in the case of some mutants which exhibit multiple conformations, this model may lead to nonequilibrium kinetics, which could produce the observed discrepancies between the kinetic and equilibrium determinations of the binding constant. These results suggest that both equilibrium and kinetic data should be obtained, even for a monomeric protein such as Mb, before the relative stabilities of mutants can be meaningfully compared.