1. Whole-cell patch recordings were used to examine the EPSC generated by the calyx of Held in neurones of the medial nucleus of the trapezoid body (MNTB). Each neurone receives a somatic input from a single calyx (giant synapse). 2. A slow NMDA receptor-mediated EPSC peaked in 10 ms and decayed as a double exponential with time constants of 44 and 147 ms. A fast EPSC had a mean rise time of 356 microseconds (at 25 degrees C), while the decay was described by a double exponential with time constants of 0.70 and 3.43 ms. 3. Cyclothiazide slowed the decay of the fast EPSC, indicating that it is mediated by AMPA receptors. The slower time constant was slowed to a greater extent than the faster time constant. Cyclothiazide potentiated EPSC amplitude, partly by a presynaptic mechanism. 4. The metabotropic glutamate receptor (mGluR) agonists, 1S,3S-ACPD, 1S,3R-ACPD and L-2-amino-4-phosphonobutyrate (L-AP4) reversibly depressed EPSC amplitude. A dose-response curve for 1S,3S-ACPD gave an EC50 of 7 microM and a Hill coefficient of 1.2. 5. Analysis of the coefficient of variation ratio showed that the above mGluR agonists acted presynaptically to reduce the probability of transmitter release. Adenosine and baclofen also depressed transmission by a presynaptic mechanism. 6. alpha-Methyl-4-carboxyphenylglycine (MCPG; 0.5-1 mM) did not antagonize the effects of 1S,3S-ACPD, while high concentrations of L-2-amino-3-phosphonopropionic acid (L-AP3; 1 mM) and 4-carboxy-3-hydroxyphenyglycine (4C3HPG; 500 microM) depressed transmission. 7. There was a power relationship between [Ca2+]o and EPSC amplitude with co-operativity values ranging from 1.5 to 3.4. 8. The mechanism by which mGluRs modulate transmitter release appeared to be independent of presynaptic Ca2+ or K+ currents, since ACPD caused no change in the level of paired-pulse facilitation or the duration of the presynaptic action potential (observed by direct recording from the terminal), indicating that the presynaptic mGluR transduction mechanism may be coupled to part of the exocytotic machinery. 9. Our data are not consistent with the presence at the calyx of Held of any one known mGluR subtype. Comparison of the time course and pharmacology of the fast EPSC with data from cloned AMPA receptors is consistent with the idea that GluR-Do subunits dominate the postsynaptic channels.