Multivariate Pattern Analysis of Electroencephalography Data in a Demand-Selection Task

Cognitive effort is costly and partly aversive, and thus humans usually avoid it if given the chance. In Demand-Selection Tasks (DST), participants tend to choose the easy option over the hard one. The neural underpinnings of this effect, however, are not well understood. The current study is an initial approximation to adapt a DST to a format that allows measuring concurrent high-density electroencephalography. We used multivariate pattern analysis (MVPA) to decode conflict-related neural processes associated with congruent or incongruent events in a time-frequency resolved way and determined how different frequency bands contribute to the overall decoding accuracy. The decoding analysis involved the use of Support Vector Machines, a supervised learning algorithm that provides a theoretically elegant, computationally efficient, and very effective solution for many practical pattern recognition problems. Preliminary results show significant differences in activation patterns for congruent and incongruent trials, yielding 80% of decoding accuracy 400 ms after the stimulus onset. The results of frequency bands contribution analysis suggest that context-dependent proportion of congruency effect may rely on neural processes operating in Delta and Theta-band frequencies.