The usefulness of event-related brain potentials (ERPs), both as s research tool for the study of cognitive processes and for applications in clinical and operational settings, is limited by the need to signal average over multiple occurrences of the event of interest and by the consequent need for precise knowledge of the timing of the eliciting event. The proposed work will explore the applicability of several pattern recognition techniques for identifying and quantifying single ERPs in segments of ongoing EEG with minimal knowledge of the timing of the eliciting event. Modifications of techniques that have been used for latency- correction average ERPs having trial-to-trial latency """"""""jitter"""""""" and techniques for classifying single trial ERPs of known timing will be evaluated using simulated EEG data with known ERPs embedded. The focus will be on determining the signal to noise conditions under which the various techniques, alone or in combination, are reliable. Approaches which prove most successful for quantifying simulated data will be validated on a limited sample of real data collected from human subjects in a simple task.