1. Single channel recording techniques were used to study the ion channel openings resulting from activation of N-methyl-D-aspartate (NMDA) receptors by the agonist glutamate. Patches were from cells acutely dissociated from adult rat hippocampus (CA1). Channel activity was studied at low glutamate concentrations (20-100 nM) with 1 microM-glycine, in the absence of extracellular divalent cations. 2. Channel openings were to two main conductance levels corresponding to 50 pS and 40 pS openings in extracellular solution with 1 mM-Ca2+. Around 80% of openings were to the large conductance level. The single channel conductances increased as extracellular Ca2+ was reduced. 3. Distributions of channel open times were described by three exponential components of 87 microseconds, 0.91 ms and 4.72 ms (relative areas of 51, 31 and 18%). Most long openings were to the large conductance level. 4. The channel closed time distribution was complex, requiring five exponential components to describe it adequately. Of these five components, at least three, with time constants of 68 microseconds, 0.72 ms and 7.6 ms (relative areas of 38, 12 and 17%) represent gaps within single activations of the receptor. The presence of a component with a mean of 7.6 ms is notable because gaps of this length have not previously been identified as being within single NMDA receptor channel activations. 5. Channel activations were identified as including gaps underlying at least the first three closed time components. Activations consisted of clusters of channel openings. Distributions of the length of these clusters had mean time constants of 88 microseconds, 3.4 ms and 32 ms (relative areas of 45, 25 and 30%). Long clusters contained short, intermediate and long duration openings as well as subconductance openings. The open probability within clusters averaged 0.62. Three components were evident in distributions of the number of openings per cluster. These had mean values of 1.22, 3.2 and 11 openings per cluster. 6. An inverse correlation was evident between the length of adjacent open and closed times. When open intervals were separated into groups based on the length of adjacent gaps, the time constants of the exponential components in these conditional open time distributions were independent of the length of the adjacent gap. This supports the idea that the NMDA receptor channel gating has the properties of a discrete Markov process. 7. The long duration of NMDA receptor channel clusters suggests that they contribute to the slow time course of the NMDA receptor-mediated synaptic current.