Data were analyzed with custom-made Matlab routines, in addition to relying on publicly available packages including FieldTrip (Oostenveld et al., 2011) and EEGLAB (Delorme and Makeig, 2004). Epochs extending −800 to 1250 ms relative to stimulus onset were extracted and detrended. Longer epochs were selected to prevent contamination of time window of interest (−300 to 800 ms) with edge artifacts during wavelet analysis. A combination of automatic threshold rejection, careful visual inspection, and independent component analysis (Delorme and Makeig, 2004) was used to remove epochs with eye-blink and other artifacts. In order to eliminate potential statistical bias, the tallies of artifact-free correct trials were equalized across beverage and task conditions by removing superfluous trials at random.
