Restoring Latency-Variable ERP Components from Single Trials: A New Approach to ERP Analysis with Residue Iteration Decomposition (RIDE)

Abstract

Electroencephalography (EEG) is widely used in cognitive neuroscience as a brain signal with high temporal resolution. Strong latency variability pervades cognitive EEG responses across single trials, but is not taken into consideration by the conventional averaging method yielding event-related potentials (ERPs). This trial-to-trial variability may strongly smear and mix ERP components and diminish their amplitudes, impeding proper identification of the spatiotemporal representation of brain activities reflecting specific cognitive subprocesses. Furthermore, rich dynamic information about single trials is lost in averaged ERPs. Here we propose a model of ERPs as consisting of temporally overlapping components locked to different external events or varying in latency from trial to trial as a foundation for a new ERP decomposition and reconstruction method, residue iteration decomposition (RIDE). RIDE obtains latency-corrected waveforms and topography of the components, and retrieves the latencies and amplitudes of the separated components in single trials. RIDE was tested with real data and provides new perspectives for investigating brain–behavior relationships using EEG data in latency-corrected reconstructed ERPs, separated components, and information about variability in single trials.