Unmasking Word Processing with ERPs: Two Novel Linear Techniques for the Estimation of Temporally Overlapped Waveforms

Abstract

Masked priming experiments are frequently used to study automatic aspects of word processing. Direct measures of such processing obtained with functional neuroimaging techniques (ERPs, fMRI, etc.) need to isolate the neural activation related to relevant events when they are rapidly followed by others (a situation found in other popular paradigms such as the attentional blink and repetition blindness). Here we examine the assumption of "simple insertion", which underlies the use of subtraction to isolate components of temporally overlapping waveforms. We propose two novel linear methods and illustrate how they extract temporal and spatial ERP components that the subtraction method fails to detect. We show this through the analysis of ERP data from a masked semantic priming procedure. The new techniques reveal activation generated by unconscious (masked) prime words as early as 100 ms and 200 ms post stimulus-onset; a pattern which simple subtraction fails to detect.