CBSU bibliography search
To request a reprint of a CBSU publication, please click here to send us an email (reprints may not be available for all publications)
Multiple regression analysis of ERPs reveals early processing stages in visual word recognition
HAUK, O., FORD M.A., DAVIS, M.H. , MARSLEN-WILSON, W.D., PULVERMüLLER, F.
Society for Psychophysiological Research, Abstracts of the 45th Annual Meeting, Psychophysiology 42: , Lisbon, 2005
Year of publication:
Despite the fact that visual word recognition is a rapid process in normal adults, literature on ERP/ERF effects at early latencies (i.e. before 200ms after word onset) is scarce and partly inconsistent. We studied the effects of several lexical variables on ERP responses in a visual lexical decision task. Linear multiple regression was applied instead of a classical factorial approach. This increased sensitivity, and allowed us to investigate multiple predictor variables without unrealistic constraints on item selection. Four orthogonal predictor variables were entered into the regression analysis, each of them expected to tap into different stages of the word recognition process: Frequency, Length/Neighbourhood Density, N-gram Frequency, as well as “Semantic Coherence”, a corpus-based morpho-semantic coherence measure that quantifies the semantic relatedness within a family of morphological relatives. The “classical” factor Lexicality (words vs. pseudowords) was included as well. Multiple regression was applied to individual ERP data sets and regression coefficients for the four variables were entered into group statistical analyses. Word form related variables Length and N-gram Frequency affected the ERP already around 90ms, followed by lexico-semantic variables such as word frequency around 110ms and Semantic Coherence and Lexicality around 160ms. We also found effects at later latencies for several variables, consistent with previous literature. Our results provide evidence for serial but cascaded processing in visual word recognition before 200ms after