Fig. 12

The process and explanation of offline data analysis. Notes: Step 1 Data re-referencing: Since the EEG recorder records the potential difference between the two electrode points, a relatively neutral electrode point (like mastoids) needs to be set as reference electrode to make the recorded voltage have a uniform standard (Luck 2014); Step 2 Data filtering: The frequency range of most relevant components in the ERP waveform is between 0.01–30 Hz (Luck 2014). However, the raw EEG data include a lower frequency signal (slow voltage changes caused by nonneural activity, such as skin electrical signals) and a higher frequency signal (some unexpected high frequency noise, such equipment noise and muscle electrical signals). Therefore, a bandpass is used to filter out the signal noise in the analysis. According to early studies (Addante et al. 2012), we used 0.1–30 Hz as the bandpass in our study; Step 3 Data segmentation: This step is to obtain the latency amplitude of interest under each condition, the segmented data is called “epoch” in an ERP study; Step 4 ICA: In this step, EEG signals are transformed into a space of independent source components, and some artifacts, such as blinks, eye movement and muscle tension, are manually identified and subtracted from the data (Changquan et al. 2018; Zhang et al. 2015); Step 5 Data averaging: The EEG signal collected in a single trial in the test is considered to be composed of ERP waveforms and random noise. Therefore, the ERP waveforms can be extracted by superposition averaged from repeated trials (Luck 2014). Thus, the same kind of stimuli should be evaluated many times in an ERP study