1. Excitatory postsynaptic currents (EPSCs) were recorded under whole-cell voltage clamp from granule cells in slices of rat cerebellum. EPSCs from individual mossy fibre inputs were identified by their all-or-none appearance in response to a graded stimulus. Excitatory synaptic transmission was investigated at room temperature (approximately 24 degrees C) and at near-physiological temperature (approximately 34 degrees C) by analysing current fluctuations in the peak and decay of the non-N-methyl-D-aspartate (non-NMDA) component of EPSCs. 2. In a subset of synapses the mean EPSC amplitude remained unchanged as the probability of transmitter release was substantially lowered by raising the extracellular [Mg2+] and lowering [Ca2+]. These synapses were considered to have only one functional release site. Single-site synapses had small EPSCs (139 +/- 16 pS, n = 5, at 24 degrees C) with a large coefficient of variation (c.v. = 0.23 +/- 0.02, n = 5) and an amplitude distribution that was well fitted by a Gaussian distribution in four out of five cases. The EPSC latency had a unimodal distribution and its standard deviation had a temperature dependence with a temperature coefficient (Q10; range, 24-35 degrees C) of 2.4 +/- 0.4 (n = 4). 3. Peak-scaled non-stationary fluctuation analysis of single-site EPSCs indicated that the mean conductance of the underlying non-NMDA channels was 12 +/- 2 pS (n = 4) at 35 degrees C. Upper and lower limits for mean channel open probability (Po), calculated from fluctuations in the EPSC peak amplitude, were 0.51 and 0.38, respectively. These estimates, together with the open probability of the channel when bound by transmitter, suggest that only about 50% of the non-NMDA channels were occupied following the release of a quantum of transmitter. 4. At some multi-site synapses EPSCs had a low c.v. (0.4 +/- 0.01, n = 5) at 34 degrees C and non-stationary fluctuation analysis gave a parabolic variance-mean current relationship. This suggests that practically all of the non-NMDA receptors were occupied by glutamate at the peak of EPSC. The channel open probability (Po = 0.84 +/- 0.03, n = 5) at these 'saturated' multi-site synapses will therefore equal the open probability of the channel when bound by transmitter (Po,max). 5. Non-stationary fluctuation analysis of EPSCs from 'saturating' multi-site synapses indicated that 170 +/- 40 postsynaptic non-NMDA channels were exposed to transmitter at the peak of the EPSC. The mean conductance of the synaptic channels was 10 +/- 2 pS (n = 5) at 34 degrees C. 6. At synapses with multiple release sites the EPSC decay time became faster when release probability was lowered (by reducing the external [Ca2+]/[Mg2+] ratio), indicating that the transmitter concentration profile depended on release probability. No such speeding of the EPSC decay was observed at single-site synapses. 7. Our results suggest that release of a packet of transmitter from a single release site does not saturate postsynaptic non-NMDA receptors at cerebellar mossy fibre-granule cell synapses. However, at multi-site synapses transmitter released from neighbouring sites can overlap, changing the transmitter concentration profile in the synaptic cleft. We conclude that the level of postsynaptic receptor occupancy can depend on the probability of transmitter release at individual multi-site synapses.