Artifactually high coherences result from using spherical spline computation of scalp current density.

Coherence computed from common reference montages inextricably confounds true coherence with power and phase at the recording and reference electrodes. Direct measurement of coherence requires reference-free EEG data, such as data from EEG scalp current densities (SCDs), which estimate the potential gradient perpendicular to the scalp. Perrin et al. (1989) presented a method for computing SCDs by taking the Laplacian of the scalp potential surface generated by spherical spline interpolation. When this method of computing SCDs was applied to EEG data gathered from young adults, very high values were observed for inter-electrode coherences computed from the spherical spline derived SCD data but not from coherences computed from the common reference data. These high coherences prompted further examination of the properties of the spherical spline function and of spherical spline derived SCDs. Simulated data were constructed, and coherence was computed on the simulated data and on the SCDs derived from the spherical spline procedure and from the Hjorth (1980) procedure. The results of those simulations are presented, which demonstrate that a major artifact is introduced by using the spherical spline procedure. This artifact results from the spline weighting matrix used to derive the SCDs and strongly inflates the inter-electrode coherences of the SCD transformed data.