1. Cross-correlation analysis of the discharges of individual motor units recorded from various different finger muscles has been performed during weak, isometric, voluntary contractions in man. 2. The dominant feature in 88% of the cross-correlograms studied was a narrow, central peak, the area of which significantly exceeded that expected for independent processes (P less than 0.001). The highest bin counts in these central peaks were mostly within 5 ms of time zero in the histograms, and the base of these peaks extended between 5 and 31 ms (modal value = 13 ms with 90% of the values lying between 8 and 18 ms). The width and displacement of the central cross-correlogram peaks were similar irrespective of whether the contributory spike trains were recorded from motor units active in the same finger muscle or recorded from motor units in different, co-activated finger muscles. 3. The time course of the central peaks in this study was found to be consistent with the hypothesis that it is generated by the joint occurrence of EPSPs evoked in motoneurones by branches of common stem presynaptic fibres using the theoretical model developed by Kirkwood (Kirkwood & Sears, 1978). The model parameters providing the best fit with our experimental data imply that synaptic contacts on motoneurones made by these common inputs lie on average peripherally in the dendritic tree and generate small (less than 300 microV) EPSPs superimposed on a high level of background synaptic noise. 4. Minima (troughs) were found either side of the central peak in 27% of the cross-correlograms studied, and their appearance was invariably associated with a large central peak. These secondary features could not be modelled with the same operator parameters that describe the central peaks. Their presence was particularly noticed in association with very regular discharges from the output motoneurones. 5. Smaller and broader secondary peaks symmetrically displaced 30-55 ms either side of the large, narrow central peak were observed in 7% of the cross-correlograms studied. We suggest that these secondary features which were found at lags shorter than the interspike interval of the contributory motor unit spike trains reflect the autocorrelation functions of the spike trains of common input fibres. On this basis the observed displacement of these secondary peaks from the primary feature in the cross-correlogram indicate firing rates for common input fibres in the range 18-33 impulses s-1. 6. In a small number of cases (1.4%) the cross-correlogram was flat and indistinguishable from the results of cross-correlating independent spike train data.