Multivariate analysis of EEG: predicting cognition on the basis of frequency decomposition, inter-electrode correlation, coherence, cross phase and cross power

This paper describes analysis of EEG data collected while participants performed a gauge-monitoring task which simulated an industrial process. Participants performed the task on two separate occasions, and the mean interval between sessions was 13 weeks. Analysis of EEG involved derivation of 5166 separate dependent variables, and these included measures of inter-electrode correlation, spectral power, coherence, cross phase and cross power. A central aim was to identify those EEG measures which provided the most reliable prediction of task demand. In particular, a major question was whether each participant might have unique aspects of their EEG which predicted cognitive load. This stemmed from a concern that attention to individual differences might provide a means for improving prediction. Results indicated that there were idiosyncratic aspects of physiological response which were highly predictive of task load. Furthermore, the predictive power of these variables survived across sessions despite the mean 3 month interval between them. Analyses also indicated the presence of EEG predictors which were common to all participants. It is concluded that idiosyncratic aspects of EEG patterns reflect genuine and reproducible individual differences. Such differences may prove a valuable tool for improving prediction. Furthermore, exploration of these variables may result in a deeper understanding of different types of cognitive style.