1. Voltage and current recordings were made from visually identified non-pyramidal neurones in slices of layer IV of rat primary visual cortex using the whole-cell configuration of the patch clamp technique. These neurones are characterized by a high input resistance (0.5-2 G-OMEGA) and a non-adaptive behaviour of action potential frequency following depolarizing current injection, which suggests that they are stellate cells. 2. Excitatory postsynaptic currents (EPSCs) were recorded from these neurones during focal stimulation of neighbouring cells by a second patch pipette, the tip of which was placed on the soma of the stimulated cell. The response amplitude as a function of stimulus strength showed a sharp increase at a critical stimulus strength suggesting that stimulus-evoked currents represent unitary EPSCs. 3. In most cases the latencies of stimulus-evoked EPSCs were unimodally distributed with means in the range of 2.1-3.6 ms. In some experiments two peaks were seen in the distribution of latencies. The EPSC rise times, measured as the time from 20 to 80% peak amplitude, fell into a distribution ranging from 0.1 to 0.8 ms with a peak at 0.2 ms. The EPSC decay time course at -70 mV membrane potential was fitted by a single exponential with a time constant of 2.39 +/- 0.99 ms (mean +/- S.D.). The rise and decay times were independent of EPSC peak amplitudes. 4. The peak amplitude of successive unitary EPSCs, elicited by a constant stimulus, fluctuated at random. At a holding potential of -70 mV the peak amplitudes varied between 5 and 90 pA. In two out of ten cells the histogram of peak amplitudes could be well fitted by the sum of several equidistant Gaussians with a peak distance of around 10 pA. This suggests that the quantal conductance change underlying the peak current fluctuations is of the order of 100 pS. 5. At membrane potentials more positive than -70 mV the decay of stimulus-evoked EPSCs showed two components with very different time courses. In standard extracellular solution the current-voltage (I-V) relation for the fast component was almost linear whereas the slow component showed a J-shaped I-V relation with a region of negative slope conductance between -30 and -70 mV. 6. The two components of stimulus-evoked EPSCs had different sensitivities to blockers of alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA)/kainate(KA) and N-methyl-D-aspartic acid (NMDA) subtypes of the glutamate receptor (GluR) channel, respectively. The fast EPSC component was blocked by 2-3-mu-M-6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), the slow component by 20-mu-M-DL-2-amino-5-phosphono-pentanoic acid (APV). This suggests that EPSCs are mediated by both the AMPA/KA and the NMDA subtypes of the GluR channel. 7. The I-V relationship of the slow component of EPSCs changed when Mg2+ was omitted from the solution. In Mg2+-free solution the slope conductance of the slow component showed a linear voltage dependence suggesting that the region of negative slope conductance arises from the voltage-dependent Mg2+ block of NMDA receptor channels. 8. The rise time and the decay time of the slower EPSC component, mediated by NMDA receptors, were considerably longer than those of the AMPA/KA receptor-dependent component. The rising phase showed a sigmoidal time course with rise times of 5.4-6.2 ms. The decay could be described by the sum of two exponentials with time constants of 16-79 and 108-307 ms, respectively. 9. Under experimental conditions where both the fast and the slow components of fluctuating unitary EPSCs could be observed (nominally Mg2+-free extracellular solution; -70 mV membrane potential) their relative sizes were almost independent of the EPSC amplitude. This suggests that AMPA/KA and NMDA receptor subtypes are co-activated and that the ratio of the currents mediated by the two receptor subtypes is relatively constant for a particular synapse. EPSCs in different cells showed different ratios of AMPA/KA to NMDA receptor-mediated currents. 10. It is concluded that unitary EPSCs in non-pyramidal (stellate) cells of layer IV of rat visual cortex are mediated by both the AMPA/KA and the NMDA subtypes of the GluR channel. The estimated number of AMPA/KA receptor channels activated by a quantum of transmitter is in the order of ten to twenty and the number of co-activated NMDA receptor channels varies between one and ten, depending on the synapse studied.