Objective: A concentric spheres model was used in an earlier paper to estim
ate the effects of volume conduction, reference electrode and spatial filte
ring on different EEG coherence measures. EEG data are used here to verify
theoretical predictions.
Methods: Three EEG data sets were: (1) 64 channel, recorded during 7 altern
ating periods of resting and mental calculation. (2) 128 channel, for compa
rison of eyes open versus eyes closed coherence. (3) 128 channel, recorded
during deep sleep (stages 3 and 4) and REM.
Results: The directions of large scale (lobeal) coherency changes between b
rain states are relatively independent of coherence measure. However, coher
ence between specific electrode pairs is sensitive to method and frequency.
Average reference and digitally Linked mastoids provide reasonable semi-qu
antitative estimates of large-scale neocortical source coherence. Close bip
olar, Laplacian, and dura image methods remove most reference electrode and
volume conduction distortion, but may underestimate coherence by spatial f
iltering.
Conclusion: Each EEG coherence method has its own potential sources of erro
r and provides coherence estimates for different neural population sizes lo
cated in different locations. Thus, studies of coherence and brain state sh
ould include several different kinds of estimates to take full advantage of
information in recorded signals. (C) 1999 Elsevier Science Ireland Ltd. Al
l rights reserved.