The P300 ERP was measured in 10 subjects each for 9 days. The selection of instructions for subjects, the recording technique, the elimination of a few single trials significantly contaminated by eye movements, and the use of a correction procedure for ocular artifacts with calculable reliability and validity resulted in a set of data, in which 94% of the single trials were suitable for further analysis. The correction procedure relies on regression analysis. To reduce coherence between eyeblink activity and ongoing EEG, VEOG and EEG are averaged on eyeblinks. This yields a high reliability and validity of regression factors, determined per day, subject, and lead. In addition, this correction procedure allows for an estimation of the maximal error that must be taken into account. The efficiency of the procedure is demonstrated for single trials and averaged potentials.