1. The whole-cell variant of the patch recording method has been used to obtain voltage recordings from trigeminal motoneurones in tissue slices (500 microns thick) taken from rats aged 8 days. Membrane properties (input resistance, membrane time constant and rheobase, i.e. threshold current required to elicit an action potential) of the motoneurones were determined and recordings made of the (untriggered) EPSP activity. 2. Untriggered EPSP activity was recorded in standard artificial cerebrospinal fluid (ACSF), ACSF with added tetrodotoxin (TTX) and in nominally Ca(2+)-free ACSF with added TTX. In each case the amplitude distributions of single EPSPs were peaky and could be fitted by a model consisting of the sum of equidistant Gaussians (n = 7/9 cells). In contrast, the amplitude distribution of the noise was always unimodal. 3. All EPSP activity recorded in the presence of TTX was abolished by addition of 6-cyano-7-nitroquinoxaline-2-3-dione (CNQX; 10 microM), suggesting the activity was all mediated by glutamate acting primarily at AMPA/kainate receptors. 4. In the majority of cases, there was no correlation between the amplitude of EPSPs underlying each Gaussian and the EPSP rise time but there was a positive correlation between the EPSP half-width and EPSP rise time. The rise times of EPSPs underlying the first, and all, fitted Gaussians were similar to that for the total sample of EPSPs in each motoneurone. Taken together, this suggests that the EPSPs underlying each Gaussian arise from inputs to different dendritic compartments, and that the range of compartments is similar for EPSPs underlying successive Gaussians. 5. Two conclusions are drawn. First, EPSPs of different dendritic origin have similar amplitudes at the soma. Second, the multimodal distribution of EPSP amplitudes recorded in the presence of TTX raises the possibility that individual boutons may contain multiple release sites, with each perhaps operating on a separate functional group of postsynaptic receptors.