Multivariate EEG Approaches, Working Memory, and Contingency Learning
Date:
This talk provided an overview of how multivariate EEG decoding approaches are transforming ERP research and enabling new insights into cognitive processes that are difficult to capture through traditional averaging techniques. I began by reviewing recent advances in the use of multivariate methods (e.g., SVM decoding, cross-validated Mahalanobis distance) to extract information from scalp voltage patterns.
In the first part of the talk, I compared the statistical power of univariate versus multivariate analyses across standard ERP-eliciting paradigms. The second part focused on applying these methods to a novel P3-based paradigm in order to address longstanding questions about information content in working memory that average ERPs could not reveal. Finally, I presented a two-armed bandit decision-making task used to explore the role of working memory in linking unsustained stimulus representations with feedback-based learning outcomes.
Together, the talk showcased how multivariate EEG tools can deepen our understanding of attention, memory encoding, and cognitive flexibility in dynamic learning contexts.