Language-related brain activity revealed by independent component analysis

Objective: When an individual engages in a cognitive task, a multitude of diverse processes are activated in his/her brain and it is reasonable to assume that multiple brain sources are simultaneously active at any one time. Magnetoencephalographic (MEG) data recorded in such circumstances provide a picture of spatial distribution and time course of the sum of the magnetic fields generated by all these sources. Thus, the experimenter faces the challenge of separating the multiple contributions to the total recorded signal before attempting a localization of their sources and studying their functional roles. Methods: We describe in this paper how independent component analysis of MEG data collected in a word/pseudo-word reading experiment elegantly solves this problem. Results: Using a few statistical assumptions, independent component analysis resolved simultaneously active brain sources in the right-frontal, left-parietal and left-frontal areas, all showing well defined dipolar field distributions. Discussion: We describe the characteristics of these contributions and discuss the language-related functional roles that appear to be associated to some of the independent sources. We report in particular on one source, localized near Broca's area, which showed to be affected by reading words but not pseudo-words. (C) 2003 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.