EEG bands during wakefulness, slow-wave and paradoxical sleep as a result of principal component analysis in man

Human electroencephalogram (EEG) has been divided in bands established by v isual inspection that frequently do not correspond with EEG generators nor with functional meaning of EEG rhythms. Power spectra from wakefulness, sta ge 2, stage 4 and paradoxical sleep of 8 young adults were submitted to Pri ncipal Component Analyses to investigate which frequencies covaried togethe r. Two identical eigenvectors were identified for stage 2 and stage 4. 1 to 8 Hz and 5 to 15 Hz (87.95 and 84.62 % of the total variance respectively) . Two eigenvectors were extracted for PS: 1 to 9 Hz and 10 to 15 Hz (81.62% of the total variance). Three eigenvectors were obtained for W: with frequ encies between 1 to 7 Hz, 7 to 11 Hz, and 12 to 15 Hz (78.32% of the total variance). Power for all frequencies showed significant differences among v igilance states. These results indicate that slow wave activity can oscilla te at higher frequencies, up to 8 Hz, and that spindle oscillations have a wider range down to 5 Hz. No theta band was independently identified, sugge sting either that delta and theta oscillations are two rhythms under the sa me global influence, or that the traditional division of theta band in the human cortical EEG is artificial. Alpha as a band was identified only durin g wakefulness. Principal component analysis upon spectral densities extract ed broad bands different for each vigilance state and from traditional band s, consistent with functional significance of EEG and with frequencies of g enerators of rhythmic activity obtained in cellular studies in animals.