EEG-recording and processing
We recorded EEG from 128 scalp electrodes, using the Active Two Biosemi
system, arranged according to the 10-5 system (Oostenveld & Praamstra,
2001). We recorded electrooculogram vertically, with one electrode just
above the eyebrow and one approximately 3 cm below the eye. EEG was
recorded with a sampling rate of 512 Hz.
We recorded eyes-closed resting state EEG for two minutes both prior to
starting, and after completion of the computerized perceptual task.
Participants were asked to relax and sit still during this period. The
EEG recordings were then cut into 1-second, non-overlapping epochs.
Pre-stimulus-EEG data was epoched from -700 – 0 ms relative to stimulus
onset. Epoched data was then detrended and we spherically interpolated
an average of 2.9% of electrodes for eyes-closed data and an average of
3.4% of electrodes for pre-stimulus data.
We used Matlab (Mathworks, 2022) for the pre-processing of all
EEG-recordings, using custom Matlab scripts. Current source density
(CSD) transformation is used in many EEG studies in order to reduce
volume conduction effects and signal mixing (Kayser & Tenke, 2006;
2015). This step was not carried out in Samaha & Postle’s original
study. We therefore opted to analyse the data both with- and without
applying CSD transformation in order to determine its impact.
For the CSD-transformed eyes-closed EEG data, we rejected epochs when
voltages exceeded a CSD-transformed amplitude threshold of 500 µV
m2 for both scalp and eye electrodes. An average of 47
(standard deviation (SD) 53.5) epochs were rejected per participant. For
CSD-transformed pre-stimulus EEG data, we rejected epochs when voltages
exceeded a threshold of 500 µV m2 for scalp electrodes
or 100 µV for eye electrodes (eye electrodes were not included in the
CSD-transformation and therefore required a lower threshold). We
rejected an average of 125 (SD 108.2) epochs per participant.
For the non-CSD-transformed eyes-closed EEG-data, we rejected epochs
when voltages exceeded a threshold of 100 µV for scalp electrodes and
500 µV for eye electrodes. We rejected an average of 32 (SD 28)
eyes-closed EEG-data epochs per participant. For non-CSD-transformed
pre-stimulus EEG-data, we rejected epochs when voltages exceeded 100 µV
for all electrodes. We rejected an average of 52 (SD 58.8) epochs of
pre-stimulus EEG-data per participant.
Epochs were then multiplied by a hamming window and fast Fourier
transformed with a frequency resolution of 0.1 Hz. Individual alpha peak
frequency was defined as in Samaha & Postle and was extracted from the
highest amplitude in the power spectrum within the frequency range 8 –
13 Hz. All participants showed a clear peak within this range at all
scalp electrodes for eyes-closed data. Similar to the original study,
not all electrodes showed a peak in the 8 – 13 Hz frequency range and
no electrode showed a peak for all participants for pre-stimulus data.
We therefore followed the original methodology and chose whichever
electrode had the highest power amplitude within the alpha band range
for each participant.
Because it has been shown that peak alpha frequency can change over time
within an individual (Cohen, 2014), we computed instantaneous alpha to
assess if pre-stimulus alpha frequency is higher prior to correct task
responses compared to incorrect responses. Instantaneous alpha
corresponds to the first temporal derivative of the phase angle time
series. To calculate instantaneous alpha, we replicated Samaha &
Postle’s method, and chose the electrode that had the highest amplitude
in the alpha band range for the pre-stimulus recordings at the group
level, which we interpreted as the electrode that had the highest alpha
amplitude for most participants: electrode C16. Following our same
reasoning as with the previous analysis, we again also computed
instantaneous alpha frequency for electrode B7. We epoched the
pre-stimulus EEG-recordings from -700 to 0 ms relative to stimulus
onset, and then copied, flipped and appended each epoch to avoid edge
artifacts. Using code developed by Cohen (2014), we then defined
frequency boundaries of 8 – 13 Hertz, and applied a zero-phase bandpass
filter with 15% transition zones. We extracted the analytic signal with
a Hilbert transform and median filtered using default parameters set by
Cohen: we applied the median filter ten times with different window
sizes ranging from 10 to 400 ms, in steps of 10 ms. As Samaha & Postle
only reported instantaneous frequency for the 500 ms prior to stimulus
onset, we followed their method and only calculated instantaneous alpha
for the timepoints ranging from -500 – 0 ms.