Further pre-processing steps were performed in Matlab 2014b using the open-source Fieldtrip toolbox v20161024 (Oostenveld et al., 2010; script: 1_preprocessing_elektra_frontiers_PAC.m). Firstly, for each participant the entire recording was band-pass filtered between 0.5 and 250 Hz (Butterworth filter, low-pass order 4, high-pass order 3) and band-stop filtered (49.5–50.5 Hz; 99.5–100.5 Hz) to remove residual 50 Hz power-line contamination and its harmonics. Data were then epoched into segments of 4,000 ms (1,500 ms pre, 1,500 ms post-stimulus onset, with 500 ms of padding either side) and each trial was demeaned and detrended. Trials containing artifacts (SQUID jumps, eye-blinks, head movement, muscle) were removed if the trial-by-channel (magnetomer) variance exceeded 8 × 10−23, resulting in an average of 63.5 trials per condition, per participant. Indices of removed trials are included in the Supplementary Materials. Site-specific MEG channels containing large amounts of non-physiological noise were removed from all analyses (MEG channels: 0111, 0322, 2542, 0532).
