averaged into 3 standard bands, theta (4–7 Hz), alpha (8–13 Hz) and beta (13–18 Hz) determined in prior studies to be sensitive to alcohol's effect on the EEG [9]-[12], [18], [25]. The 3 banded powers and their standard deviations constituted a total set of 6 variables for each of the 7 electrode sites. For each study, the power and standard deviation values were transformed to z-scores within subjects across alcohol and placebo or pre-drinking conditions to reduce scale differences between the variables and between subjects [26], [27]. The values of the 6 variables at the 7 electrode sites were then averaged over all sites to produce 6 final candidate variables characterizing alcohol’s effect over the entire cerebral cortex [14]. Differences in the variables between alcohol and placebo/pre-drinking conditions were assessed for each study by ANOVAs with condition and variable as within-subjects factors. Degrees of freedom in the ANOVAs were adjusted when appropriate using the Greenhouse-Geisser technique to correct for violations of the sphericity assumption.
