1. The kinetics of action of 17 structurally related NMDA receptor competitive antagonists were measured under voltage clamp in mouse hippocampal neurones. Analysis of the response to rapid changes in antagonist concentration during constant application of agonist was used to estimate microscopic association (kon) and dissociation (koff) rate constants for antagonist binding, assuming a two-equivalent site model for competitive antagonism. Dose-inhibition curves were analysed to estimate antagonist equilibrium dissociation constants. 2. For a series of 11 omega-phosphono, alpha-amino acids kon and koff varied 26 and 107 fold respectively. Rapid association and dissociation rate constants were obtained for flexible antagonist molecules such as D-2-amino-7-phosphonoheptanoic acid (D-AP7): kon 1.4 x 10(7) M-1 s-1; koff 20.3 s-1. For conformationally restrained molecules such as 3S,4aR,6S,8aR-6-phosphonomethyl-decahydroisoquinoline- 3-carboxylic acid (LY 235959), association and dissociation rate constants were much slower: kon 1.1 x 10(6) M-1 s-1; koff 0.2 s-1. For the D- and L-isomers of 3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP) estimates for kon were similar, but for the L-isomer koff was 10 fold faster than for the D-isomer. 3. For 2-amino-5-phosphonopentanoic acid (AP5) and its piperidine derivative cis-4-(phosphonomethyl)piperidine-2-carboxylic acid (CGS 19755), an increase in chain length of two methylene groups between the omega-phosphono and alpha-carboxylate moieties caused a 1.6 to 1.8 fold decrease in kon with little change in koff. In contrast, for AP5, CPP and its omega-carboxylate analogue, addition of a double bond close to the phosphonate moiety caused a 1.3 to 1.6 fold increase in kon. 4. For antagonists with an omega-tetrazole moiety, kon and koff were 2.8-4.6 times faster than for the parent omega-phosphono compounds. A similar, but smaller increase in kon and koff was observed for antagonists with an omega-carboxylate moiety. 5. The slow kinetics of action of potent NMDA receptor antagonists were not an artefact of buffered diffusion. In neurones equilibrated with 200 microM D-AP7, 2 microM LY 235959 and 10 microM NMDA, a transient agonist response was recorded following a rapid switch to D-AP7-free solution. This can only be explained by differences in the binding kinetics of AP7 and LY 235959, since at equilibrium, with these concentrations, either antagonist essentially eliminates the agonist response to 10 microM NMDA. 6. For all antagonists studied, the ratio koff/k0. was consistent with equilibrium Ki values obtained under similar experimental conditions, over a 40 fold range of potency. Comparison of these values with Ki estimates determined from both agonist ([3H]-glutamate), and antagonist ([3H]-CGS 19755 and [3H]- CPP) radioligand competition studies revealed good correlation between data from voltage clamp and binding experiments. However, Ki values obtained in antagonist binding assays showed on average 6.5 fold higher affinity than those obtained in voltage clamp experiments; in contrast Ki values obtained in agonist binding assays showed only 1.4 fold higher affinity. 7. The insights gained from our experiments may be of use for predicting the structural features required to generate more potent NMDA receptor antagonists, and suggest that novel acyclic compounds will have greater potential for high potency than derivatives of conformationally rigid compounds with piperazine, piperidine or bicyclic ring structures.