Psychophysiology, 37:S35, 2000.
Principal components analysis (PCA) as a tool for identifying EEG frequency bands: II. Dissociation of resting alpha asymmetries
Stefan Debenera, Jürgen Kayserb, Craig E. Tenkeb, André Beauducela
a Dresden University of Technology, Department of Psychology II, Germany
b Department of Biopsychology, New York State Psychiatric Institute, New York, NY 10032, USA
Abstract
Activity in the EEG alpha (8-13 Hz) band is considered to be an inverse measure of brain activation. However, the interpretation of changes in EEG alpha is complicated by the boundaries selected for measuring alpha, and the nature and impact of artifacts. Using a methodology analogous to that applied for the identification and measurement of ERP components, a covariance-based frequency domain PCA (fPCA), followed by unscaled Varimax rotation, was applied to resting EEG recorded from 138 subjects during two conditions (eyes open/closed) at 29 sites (nose reference). Amplitude spectra (sqrt power) spanned 0-25 Hz (0.25 Hz resolution). Six factors (89.9% variance) were extracted using subjects (138) x channels (29) x conditions (2) as cases. Factors were clearly distinguishable, and included four factors with alpha loading peaks (i.e., at 9.5, 10.25, 11.0, and 11.5 Hz), and a low (< 5 Hz) and high (> 14 Hz) frequency factor. Split-half analyses supported factor stability. While the 9.5-Hz factor also extended into theta range (below 8 Hz), the remaining alpha factors had no secondary loadings. Factor score topographies revealed distinct regional characteristics, and prominent condition effects for the alpha factors. Factor score asymmetries (R minus L) calculated for each site were not systematically correlated among alpha factors. For instance, 9.5-Hz alpha asymmetry was unrelated to 11.5-Hz alpha asymmetry at midfrontal sites (r = -.05, eyes closed; r = .12, eyes open). Implications for alpha asymmetry research are discussed.
Keywords: EEG methodology; frequency PCA; alpha asymmetry