A technique for the derivation of digitally-averaged multiple unit activity (MUA) is described. The use of signal averaging instead of analog integration improves the temporal resolution and thus provides a clearer picture of the instantaneous MUA level. MUA recordings have been used in the identification of regions active in the generation of event-related potentials, based in part on the limited volume within which a semi-microelectrode 'sees' action potentials. However, averaged MUA waveforms may be affected by time-locked activity volume-conducted to the electrode site. A theoretical analysis of the magnitude of this effect is presented, along with experimental data in support of its assumptions and predictions. The most important factor is not the absolute size of the volume-conducted potentials, but their magnitude relative to that of the locally-generated MUA. When full-wave rectification is used, volume-conducted activity which is a considerable fraction of the MUA level will not significantly affect the averaged MUA waveform. Half-wave rectification should not be used, as it leads to a much larger effect from small far-field potentials.