1. Intracellular recordings were obtained from 112 supraoptic nucleus magnocellular neurosecretory cells (MNCs) in superfused explants of rat hypothalamus maintained in vitro. The effects of glutamate receptor agonists and antagonists were examined at 32-34 degrees C. 2. In control solutions, spontaneously active (> 5 Hz) phasic or continuous neurones showed interspike interval distributions slightly skewed toward short intervals, but did not feature pauses in the 0.4-2 s range. Current injection to alter the rate of cell discharge shifted the histograms according to the mean firing rate, but failed to induce intermittent pauses in the 0.4-2 s range. 3. Application of N-methyl-D-aspartate (NMDA) induced a mode of firing in which bimodal interspike interval distributions reflected a high incidence of clusters of short interspike intervals (0.5-1.5 s) recurring every 1-3 s. In contrast, firing evoked by application of D,L-alpha-amino-3-hydroxy-5-methyl-4-isoxalone propionic acid (AMPA) was not associated with a clustering of impulse discharge. 4. The putative endogenous excitatory amino acid transmitters L-glutamate, L-aspartate and quinolinate all mimicked the effects of NMDA. Clustered spiking responses to these agents were reversibly blocked by D,L-2-amino-5-phosphono-valerate (APV), but not by 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). In contrast, the non-NMDA receptor ligands kainate and quisqualate caused CNQX-sensitive increases in firing rate, but these responses were not associated with the appearance of clustered activity. 5. When applied to cells showing negative resting potentials (< -70 mV), or to neurones hyperpolarized by current injection, responses to NMDA consisted of rhythmic (approximately 1 Hz) voltage oscillations associated with bursts of spike discharge. In the presence of TTX, NMDA could induce subthreshold voltage oscillations in the absence of action potentials. 6. Application of a voltage clamp to potentials between -75 and -55 mV during rhythmic bursting responses failed to reveal any rhythmic oscillation of the membrane current. In all cases, rhythmic bursting activity resumed upon returning to the current-clamp mode. 7. Rhythmic bursting responses to NMDA application were abolished in Mg(2+)-free solutions, suggesting that the voltage dependence of NMDA channels served to promote regenerative voltage changes throughout the cycle. The NMDA-induced current itself, however, did not appear to decrease with time, suggesting that a distinct, outward current, was necessary to initiate the repolarizing phase of each cycle.(ABSTRACT TRUNCATED AT 400 WORDS)