Methods have been developed for the removal of retinol from human plasma retinol-binding protein (RBP), so as to form the retinol-free apoprotein, and for the recombination of apo-RBP with retinol to again form the holoprotein. Retinol is removed from RBP by gently shaking a solution of RBP with heptane under controlled conditions. During the shaking, retinol is gradually extracted from the RBP and into the heptane phase. The reassociation of apo-RBP with retinol is achieved by exposing a solution of apo-RBP to Celite coated with a thin film of retinol, followed by isolation of the RBP by gel filtration on Sephadex G-100. This procedure results in the recombination of apo-RBP with an amount of retinol almost identical with that previously removed by extraction. The two-phase extraction procedure was used to explore some of the factors which affect the interaction of retinol with RBP. The retinol-RBP complex was most stable in the lower portion of the pH range 5.6 to 10. The rate of removal of retinol from the RBP-prealbumin complex (the form in which RBP normally circulates in plasma) was markedly less than the rate of its removal from RBP alone. The interaction of retinol with RBP appears to be stabilized by the formation of the RBP-prealbumin complex. The recombination procedure was employed to examine the specificity of the binding of retinol to RBP, by determining whether compounds other than all-trans-retinol would effectively bind to apo-RBP. Apo-RBP did not bind cholesterol, but displayed a slight affinity for phytol. The affinity of RBP for beta-carotene was minimal, whereas both retinyl acetate and retinal were bound about one-third as effectively as all-trans-retinol. In contrast, retinoic acid bound to apo-RBP almost as effectively as did retinol. Each of two isomers of retinol, 13-cis and 11,13-di-cis-retinol, bound to apo-RBP to some extent. The 13-cis isomer appeared to bind somewhat less effectively than did the 11,13-di-cis isomer. The binding of retinol to RBP is highly but not absolutely specific.