Coronal sections of rat brain (500 micron thick) were trimmed to form 'wedges' of tissue consisting of cerebral cortex and corpus callosum. When these slices were placed in a two-compartment bath, the cortical tissue could be depolarized, relative to the corpus callosum, by superfusions of high K+, or by amino acids such as L-glutamate, L-aspartate, quisqualate, kainate and N-methyl D-aspartate (NMDA). Responses to NMDA were reduced by magnesium ions, by the organic antagonists (-)-2-amino 5-phosphonovalerate (APV) and 2-amino 7-phosphonoheptanoate (APH), and by the dissociative anaesthetic ketamine. In this preparation, all these antagonists shifted the NMDA dose-response curve to the right in a parallel manner. A Schild plot for Mg2+ had a slope significantly less than unity, indicative of a non-competitive action, whilst Schild plots for (-)-APV, APH and ketamine appeared linear and had slopes of approximately 1. Analysis of the results of combination experiments suggested that the presumed competitive antagonists, (-)-APV and APH, share a common site of action as NMDA antagonists, and that this site is distinct from that at which ketamine exerts its action. The action of Mg2+ is clearly different from that of either (-)-APV or ketamine. It is concluded that ketamine is a non-competitive antagonist of NMDA and may act at an allosteric site on the NMDA receptor complex to influence its function.