Derivatives of D-xylose and D-glucose, in which the hydroxy groups at C-5, and C-5 and C-6 were replaced by fluorine, hydrogen and azide, were synthesized and used as substrates of the NAD(P)H-dependent aldehyde reduction catalysed by aldose reductases isolated from the yeasts Candida tenuis, C. intermedia and Cryptococcus flavus. Steady-state kinetic analysis showed that, in comparison with the parent aldoses, the derivatives were reduced with up to 3000-fold increased catalytic efficiencies (kcat/Km), reflecting apparent substrate binding constants (Km) decreased to as little as 1/250 and, for D-glucose derivatives, up to 5.5-fold increased maximum initial rates (kcat). The effects on Km mirror the relative proportion of free aldehyde that is available in aqueous solution for binding to the binary complex enzyme-NAD(P)H. The effects on kcat reflect non-productive binding of the pyranose ring of sugars; this occurs preferentially with the NADPH-dependent enzymes. No transition-state stabilization energy seems to be derived from hydrogen-bonding interactions between enzyme-NAD(P)H and positions C-5 and C-6 of the aldose. In contrast, unfavourable interactions with the C-6 group are used together with non-productive binding to bring about specificity (6-10 kJ/mol) in a series of D-aldoses and to prevent the reaction with poor substrates such as D-glucose. Azide introduced at C-5 or C-6 destabilizes the transition state of reduction of the corresponding hydrogen-substituted aldoses by approx. 4-9 kJ/mol. The total transition state stabilization energy derived from hydrogen bonds between hydroxy groups of the substrate and enzyme-NAD(P)H is similar for all yeast aldose reductases (yALRs), at approx. 12-17 kJ/mol. Three out of four yALRs manage on only hydrophobic enzyme-substrate interactions to achieve optimal kcat, whereas the NAD(P)H-dependent enzyme from C. intermedia requires additional, probably hydrogen-bonding, interactions with the substrate for efficient turnover.