Using temporo-spatial principal component analysis as tool to dissociate latent ERP components of episodic memory retrieval: Objectifying time-window selection for overlapping ERP components

This report details how principal component analysis (PCA) can be used as a valuable tool to dissociate latent ERP components, even when considerable temporal and spatial overlap makes it difficult to discern ERP effects in standard time windows. We illustrate our methodological approach in a data set from a recognition memory paradigm, in which event-related potential (ERP) correlates of familiarity, recollection and the late parietal negativity (LPN) were partially overlapping. By adapting standard time windows based on the results of a temporo-spatial PCA, small yet reliable ERP correlates reflecting familiarity and recollection for identical items and late recollection for changed items were identified, complementing the result pattern observed in behavioral performance. Due to similar temporo-spatial characteristics and opposing polarities in late parietal ERP correlates associated with memory retrieval, component overlap is often observed in this field of research. Hence, the complex interplay of several processes underlying higher cognitive functions such as memory retrieval may interfere with standard ERP assessment. In such instances, PCA can provide promising ways to objectively assess time window selection for subsequent ERP analyses.