To be fully exploitable extracellular multi-unit data have to be sorted out into several single neuron spike trains: this particular data processing is called "spike-sorting". This work is a contribution to the development and the carrying out of an automatic spike-sorting method implementing a Markov Chain Monte Carlo (MCMC) method. The proposed method enables the experimentalist to take into account the occurrence times of spikes, in addition to the information provided by their waveforms, to perform spike-sorting. This use of temporal information makes it possible to automatically identify neurons with non-stationary spike waveforms. It also improves the separation of neurons whose spike waveforms are similar. This methodological work le...