Cognitive Brain Research 15 (2003) 137–153 www.elsevier.com / locate / cogbrainres Research report Parametric analysis of event-related potentials in semantic comprehension: evidence for parallel brain mechanisms a, ,1 b a b,c * Joseph Dien , Gwen A. Frishkoff , Arleen Cerbone , Don M. Tucker a Department of Psychology, 2007 Percival Stern Hall, Tulane University, New Orleans, LA 70118, USA b Department of Psychology, Institute of Cognitive and Decision Sciences, University of Oregon, Eugene, OR, USA c Electrical Geodesics, Inc., Eugene, OR, USA Accepted 8 March 2002 Abstract In event-related potential (ERP) studies of cognitive processes, the electrophysiological responses are typically contrasted between experimental conditions that are taken to represent discrete categories (e.g. attended vs. unattended stimuli, or real vs. nonsense words). Because categorical variation is less powerful than continuous or parametric variation, a more effective method may be to relate continuous variation in the cognitive process with matching variation in the electrophysiological responses. We assessed continuous variation in the expectancy and meaningfulness of words in different sentence contexts by having subjects rate the words along these two dimensions. ERP averages were then created for each word by averaging the ERP across all subjects’ responses to that word. A parametric principal components analysis was then conducted by multiplying the factor topographies from the temporal PCA by the parameter correlation maps for each rating parameter. This analysis showed that both expectancy and meaningfulness begin to influence lexical processing around 200 ms. Source localization of the expectancy N2 (recognition potential) pointed to a source in the left fusiform gyrus region (visual word form area). Source localization of the meaningfulness N2 (meaning recognition potential) suggested a right inferior posterior source, such as in the right cerebellum or right fusiform area. Further research with parametric analysis of dense array ERPs may clarify the multiple neural mechanisms of word recognition.  2003 Elsevier Science B.V. All rights reserved. Theme: Neural basis of behaviour Topic: Cognition Keywords: EEG; Priming; Event-related potential; Fusiform gyrus; Language; N2 1. Introduction channels. Principal component analysis (PCA) can be applied to separate sources of variation in the time course Given recent advances in dense array (64- to 256- of the scalp electrophysiological responses that may reflect channel) measurement of brain electrical activity, there are specific neural processes [16]. An important problem is new opportunities for applying multivariate analysis to identifying the psychological processes that may be rel- electroencephalographic (EEG) and event-related potential evant to an electrophysiological component. In the present (ERP) data. The multivariate approach provides objective paper, we describe the use of continuous ratings of and rigorous tools for extracting hidden signals from the psychological properties of words in order to extract the measurement variables, in this case the recorded scalp specific electrophysiological variation in the principal components that is relevant to those ratings. Applied to a conventional semantic incongruity visual ERP paradigm, *Corresponding author. Tel.: 11-504-862-3300; fax: 11-504-862- this method suggests that the visual word form areas of 8744. both left and right fusiform areas may respond to semantic E-mail address: jdien@tulane.edu (J. Dien). 1 information by 200 ms. http: / / www.tulane.edu / |jdien / dien.html 0926-6410 / 03 / $ – see front matter  2003 Elsevier Science B.V. All rights reserved. PII: S0926-6410(02)00147-7