Twenty-three schizophrenic patients and 26 age-matched control subjects were studied using the P300 recorded during the auditory oddball task, with counting. Our aim was to assess the most suitable method of measurement and analysis of P300 amplitude and latency for use in clinical studies of schizophrenia. The effect of high-pass filtering, peak definition method and recording electrode site were all investigated. We have developed a technique, based on a least-mean-squares approximation to data, which seems particularly well suited to dealing with multi-peak P300 complexes. We have also investigated the spectral composition of the P300 and have found some evidence to support a proposed 2-frequency model of the P300 complex.