A new off-line procedure for dealing with ocular artifacts in ERP recording is described. The procedure (EMCP) uses EOG and EEG records for individual trials in an experimental session to estimate a propagation factor which describes the relationship between the EOG and EEG traces. The propagation factor is computed after stimulus-linked variability in both traces has been removed. Different propagation factors are computed for blinks and eye movements. Tests are presented which demonstrate the validity and reliability of the procedure. ERPs derived from trials corrected by EMCP are more similar to a 'true' ERP than are ERPs derived from either uncorrected or randomly corrected trials. The procedure also reduces the difference between ERPs which are based on trials with different degrees of EOG variance. Furthermore, variability at each time point, across trials, is reduced following correction. The propagation factor decreases from frontal to parietal electrodes, and is larger for saccades than blinks. It is more consistent within experimental sessions than between sessions. The major advantage of the procedure is that it permits retention of all trials in an ERP experiment, irrespective of ocular artifact. Thus, studies of populations characterized by a high degree of artifact, and those requiring eye movements as part of the experimental task, are made possible. Furthermore, there is no need to require subjects to restrict eye movement activity. In comparison to procedures suggested by others, EMCP also has the advantage that separate correction factors are computed for blinks and movements and that these factors are based on data from the experimental session itself rather than from a separate calibration session.