There is evidence, some of it of questionable authenticity, which suggests that phosphoglycerate kinase takes up a more compact form following the binding of substrates. Using this evidence it has been assumed that a conformational rearrangement is required for phosphoryl transfer to occur and that this is brought about by moving the enzyme's two domains towards each other. In order to test this hypothesis we have modified, by site-directed mutagenesis, an arginine residue thought to be involved in stabilising the transition-state intermediate. Although some 1.3 nm away from the site of phosphoryl transfer, as seen in the crystallographically determined structure, the substitution of arginine 168 by lysine (R168K) more than halves the specific activity of the enzyme. Substituting the arginine with a methionine (R168M) reduces activity further, but not completely, thus proving that the charge associated with this residue is not essential for catalytic activity. Both mutations raise the Michaelis constants (Km) for ATP and glycerate 3-phosphate. The largest change is observed with the triose substrate and the methionine mutant, suggesting that the primary function of arginine 168 is to influence the environment of this substrate. The effect on activity of adding sulphate to R168K and R168M mutant enzyme has also been investigated. The sulphate activation effect at low substrate concentrations is reduced for the methionine substitution but almost abolished for the lysine substitution. The most reasonable explanation of all these findings is that, in the wild-type enzyme, the guanidinium group of arginine 168 forms a hydrogen bond with one of the triose substrate's C1 oxygens. This steric arrangement would not be possible in the 'open form' of this enzyme as observed in the crystal structure.