A microprocessor-based device was constructed to retrieve single unit activity from nerve recordings contaminated by other units and EMG activity. The microneurographic signal is sampled at 10 kHz and an algorithm applied to identify impulses from a single nerve fibre. On line, a TTL pulse is delivered when an event, i.e. a provisional nerve impulse, is selected. The wave form and clock time of events are stored. Moreover, the latest selected event and the actual selection criteria are continuously displayed on a standard oscilloscope. Off line, the wave form and clock time of events as well as an instantaneous frequency plot can be displayed on the oscilloscope. The final selection of events is done with a combination of a second algorithm, which essentially is a wave form comparator, and a manual check. The device is controlled either by hardware, with knobs on the front panel, or by software through a data bus connected to a microcomputer. Clock times and wave forms of the events, which are stored in the microprocessor memory, may also be presented on the data bus for later off-line analysis and coordination with other related signals collected during the experiment, e.g. transducer and electromyography records, whether these were stored on analog or digital tape or computer disc. Compared to other available techniques, the device has a superior discriminative power when electromyographic artefacts are present.