1. Initial rates of oxidative deamination of L-glutamate with NAD+ as coenzyme, and of reductive aminiation of 2-oxoglutarate with NADH as coenzyme, catalysed by bovine liver glutamate dehydrogenase were measured in 0.111 M-sodium phosphate buffer, pH 7, at 25 degrees C, in the absence and presence of product inhibitors. All 12 possible combinations of variable substrate and product inhibitor were used. 2. Strict competition was observed between NAD+ and NADH, and between glutamate and 2-oxoglutarate. All other inhibition patterns were clearly non-competitive, except for inhibition by NH4+ with NAD+ as variable substrate. Here the extrapolation did not permit a clear distinction between competitive and non-competitive inhibition. 3. Mutually non-competitive behaviour between glutamate and NH4+ indicates that these substrates can be bound at the active site simultaneously. 4. Primary Lineweaver-Burk plots and derived secondary plots of slopes and intercepts against inhibitor concentration were linear, with one exception: with 2-oxoglutarate as variable substrate, the replot of primary intercepts against inhibitory NAD+ concentration was curved. 5. Separate Ki values were evaluated for the effect of each product inhibitor on the individual terms in the reciprocal initial-rate equations. With this information it is possible to calculate rates for any combination of substrate concentrations within the experimental range with any concentration of a single product inhibitor. 6. The inhibition patterns are consistent with neither a simple compulsory-order mechanism nor a rapid-equilibrium random-order mechanism without modification. They can, however, be reconciled with either type of mechanism by postulating appropirate abortive complexes. Of the two compulsory sequences that have been proposed, one, that in which the order of binding is NADH, NH4+, 2-oxoglutarate, requires an implausible pattern of abortive complex-formation to account for the results. 7. On the basis of a rapid-equilibrium random-order mechanism, dissociation constants can be calculated from the Ki values. Where these can be compared with independent estimates from the kinetics of the uninhibited reaction or from direct measurements of substrate binding, the agreement is reasonable good. On balance, therefore, the results provide further support for the rapid-equilibrium random-order mechanism under these conditions.