Two rhesus monkeys were implanted with silicon arrays of 96 microelectrodes. Neural activity was recorded periodically over a period of weeks to months. We have developed a method to determine whether single units in two separate recording sessions represent the same neuron. Pairwise cross-correlograms, the autocorrelogram, waveform shape, and mean firing rate were used together as identifying features of a neuron. When two units recorded on separate days were compared using these features, their similarity scores tended to be either high, indicating two recordings from the same neuron, or low, indicating different neurons. Although these metrics are individually weak, together they produce a strong classifier. Some neurons were recorded for >100 days. These monkeys performed a center-out reaching task, and we found that the firing properties of chronically recorded neurons were stable over time.