Electroencephalographic (EEG) data is often plagued by artifacts due to the muscle activity associated with unwanted eye movements. Current strategies for dealing with ocular artifacts focus on three approaches: rejection, correction via regression techniques, or correction via independent component analysis (ICA). While regression (e.g. Gratton, Coles, and Donchin, 1983) and ICA do a reasonable job of removing the effects of eye movements from EEG, these techniques are only able to provide an estimate of the underlying data. Here, we tested a trial-rejection strategy in the context of a reward-processing task in which participants received feedback after estimating the length of one second. Using a custom MATLAB script we developed to display summaries of EEG data surrounding participant responses, a research assistant accepted or rejected each trial based on peak voltage, peak voltage change, and a visual inspection of the waveform. Using this rejection protocol, our data was free of ocular artifacts and, after subsequent analysis, resulted in event-related potentials that were less variable in both peak magnitude and temporal stability, compared to when ocular artifacts were corrected using traditional regression or ICA methods. Interestingly, we noted that our trial rejection strategy resulted in a slight bias towards correct trials (trending towards significance) suggesting that participants made more eye movements following error feedback compared to correct feedback. This bias, if present, would impact any EEG analysis, regardless of how ocular artifacts were handled. The proposed technique, however, offers three advantages: fewer ocular artifacts in the resulting EEG, less variability in the resulting ERPs, and an elimination of the need for ocular correction.
Meeting abstract presented at VSS 2013