The "phosphoryl-enzyme" prepared from phosphoglycerate kinase and adenosine 5'-triphosphate in the presence of an adenosine 5'-diphosphate trap is shown to contain stoichiometric amounts of 3-phosphoglycerate. This "phosphoryl-enzyme" is chemically competent, but is probably just a tight complex between 1,3-bisphosphoglycerate and the enzyme. The two partial exchange reactions (between adenosine 5'-diphosphate, and adenosine 5'-triphosphate, and between 3-phosphoglycerate and 1,3-bisphosphoglycerate) can both be observed, but their rates are very much slower than the rate of overall catalysis. No substrate analogue was found that accelerated the partial exchange reactions. Catalysis of each of the two exchange reactions and of the kinase reaction coincides after isoelectric focusing of purified enzyme, but the amount of cosubstrate necessary to cause the observed partial exchange rates is so small that these reactions may well be artifactual. The balance of evidence does not support a ping-pong pathway via phosphoryl-enzyme, and the reaction may be a sequential one in which the phosphoryl group is transferred between substrates in a ternary complex. The results point to the dangers in the interpretation of experiments where very small amounts of contaminating cosubstrate can lead to large kinetic effects, and to the possibility of mistaken deductions about the identity of reaction intermediates.