A multi-component decomposition algorithm for event-related potentials

Event-related potentials (ERPs) are evoked activities of the brain related to specific events. They can be estimated by averaging across many trials aligned to a specific event onset time point, such as the stimulus, the response, or other behaviorally significant markers. If a single trial includes more than one such event (marker), as in all reaction-time tasks, the cross-contamination of components related to different events may mislead the explanation of ERPs. In order to recover event-related components, Zhang [Zhang J. Decomposing stimulus and response component waveforms in ERP. Journal of Neuroscience Methods 1998;80:49-63] provided a method for decomposing of ERPs according to two markers (stimulus and the behavioral response). Here we extend this formulation to deal with three or more markers in a single trial, and recover individual ERP components that are time-locked to those markers. As an application, we analyzed a cuing experiment with three events: cue, stimulus and response. The elapse between cue and stimulus was varied from trial to trial by the experimenter, and the time between stimulus and response was determined by the subjects (reaction-time variation). Our decomposition results show that the cue-dependent component waveform turns out to flatten out 500 ms after cue-onset, a finding consistent with our experimental paradigm.