Abstract Human erythrocyte ghosts retain most of the enzyme glyceraldehyde 3-phosphate dehydrogenase bound to the membrane. The bound dehydrogenase, 3 x 105 molecules per ghost, can be eluted with increasing concentrations of salt, with 50 % elution occurring at 0.16 m NaCl, pH 7.5. Exogenous glyceraldehyde phosphate dehydrogenase can bind to the membrane. Two types of binding sites have been detected. The high affinity sites bind a maximum of 3 x 105 molecules of dehydrogenase per ghost with a Kd less than 10-8 m for the enzyme obtained from either erythrocytes or rabbit muscle. The low affinity sites bind up to 63 x 105 molecules per ghost with Kd greater than 10-7 m. Enzyme bound to the membrane in this way can be re-eluted with NaCl, indicating freely reversible interaction between the enzyme and membrane. Kant and Steck (J. Biol. Chem. (1973) 248, 8457–8464) have reported reversible binding of erythrocyte glyceraldehyde phosphate dehydrogenase with ghost membranes; however, their conditions did not permit the demonstration of two types of binding sites. There is specificity in the enzyme-membrane interactions. Both the erythrocyte and rabbit muscle glyceraldehyde phosphate dehydrogenase bind to the high and low affinity sites, but yeast enzyme interacts only weakly with the membrane. A large amount of yeast enzyme (62 x 105 molecules per ghost) binds with a Kd of 10-4 m, binding which resembles the interaction of cytochrome c with this membrane. Ghosts prepared and sealed in the presence of MgSO4 bind very little glyceraldehyde phosphate dehydrogenase and no binding occurs with high affinity. This result agrees with the conclusion of Kant and Steck that the dehydrogenase specific sites are located on the cytoplasmic surface of the membrane. Ghosts incubated at 37° in NaCl become partially sealed. Such ghosts also bind the dehydrogenase both at the high and low affinity sites. However, at each type of site part of the bound enzyme is cryptic, i.e. its activity is masked unless a detergent, Triton X-100, is present. It appears that under these conditions there is an allotopic effect on some of the dehydrogenase when it interacts with the membrane. The number of high affinity glyceraldehyde phosphate dehydrogenase binding sites on the ghost membrane is similar to that of other known membrane proteins of erythrocytes and about equals the average number of the dehydrogenase molecules in the intact erythrocyte. These facts, together with the high specificity and high affinity, suggest that the interaction between this enzyme and the membrane may have a physiological function. Because both the enzyme and membrane are well characterized they form a valuable model system for examining enzyme-membrane interactions in vitro.