EEG recording and preprocessing
EEG data was acquired using a standard 64-channel Ag/AgCl electrode cap
(Brain Products) in accordance with the extended international 10–20
system. Vertical and horizontal electrooculograms were simultaneously
recorded, with electrodes placed below the left eye and on the outer
canthus of the right eye. The EEG signals were referenced to the FCz
electrode and sampled at 1000 Hz, and were online filtered using a
0.1-100 Hz bandpass filter. The impedance of all electrodes was kept
below 10 kΩ during the recording.
EEG data was preprocessed offline using EEGLAB (Delorme & Makeig, 2004)
and custom scripts in MATLAB (MathWorks). A zero-phase high-pass filter
at 0.3 Hz was used to eliminate slow drifts, followed byZapline-plus (Klug & Kloosterman, 2022) and CleanLine(Mullen 2012) to remove line noise. Bad channels were identified and
removed by Clean_rawdata , and the data was downsampled to 250 Hz
to conserve computation time. Independent component analysis was
conducted with ICLabel (Pion-Tonachini et al., 2019) to recognize
and eliminate components associated with eye blink, vertical eye
movement, muscle motor, and channel noise. The EEG data was then
segmented from -300 to 1500 ms with epochs locked on the onset of the
stimuli, and baseline correction was conducted using an interval from
−300 to 0 ms. Artifact rejection was used to mark and exclude epochs
with voltage values exceeding ± 100 μV at any time point. Bad channels
were interpolated, and the data was re-referenced to the common average.
Our preprocessing process was based on some recent standardized EEG
processing pipelines (Bailey et al., 2023; Monachino et al., 2022;
Pedroni et al., 2019).
Two participants had more than 40% of their trials identified as
artifacts, and were thus excluded from further analysis. Of the 28
participants included, 2.57% (SD = 3.84%) of the trials
contained artifacts, and 1.897 (SD = 1.496) channels were deemed
bad.