Aldose reductase (AR) has been implicated in the etiology of the secondary complications of diabetes, and enzyme inhibitors have been proposed as therapeutic agents. While effectively preventing the development of diabetic complications in animals, results from clinical studies of AR inhibitors have been disappointing, possibly due to poor potency in man. To assist in the design of more potent and specific inhibitors, crystallographic studies have attempted to identify enzyme-inhibitor interactions. Resolution of crystal complexes has suggested that the inhibitors bind to the enzyme active site and are held in place through hydrogen bonding and van der Waals interactions formed within two hydrophobic pockets. To confirm and extend these findings we quantified inhibitor activity with single, site-directed, mutant, human AR enzymes in which the apolar active-site residues tryptophan 20, -79, -111 and phenylalanine 115 were replaced with alanine or tyrosine, decreasing the potential for van der Waals interactions. Consistent with molecular models, the inhibitory activity of Tolrestat, Sorbinil and Zopolrestat decreased 800-2000-fold when tested with the mutant enzyme in which Trp20 was replaced with alanine. Further, alanine substitution for Trp111 decreased Zopolrestat's activity 400-fold, while mutations to Trp79 and Phe115 had little effect on the activity of any of the inhibitors. The alanine mutation at Trp111 had no effect on Tolrestat's activity but decreased the activity of Sorbinil by about 1000-fold. These latter effects were unanticipated based on the number of non-bonded interactions between the inhibitors, Tolrestat and Sorbinil, and Trp20 and Trp111 that have been identified in the crystal structures. In spite of these unexpected findings, our results are consistent with the hypothesis that AR inhibitors occupy the enzyme active site and that hydrophobic interactions between the enzyme and inhibitor contribute to inhibitor binding stability.