Page 104 - Demo

Chapter 4102was transmitted and amplified using a Biopac MP150 system combined with a BioNomadix 2 channel wireless EMG amplifier. Data recordings were made in AcqKnowledge 4.3 (Biopac Systems Inc, 2009) using a sampling rate of 1000 Hz. Event markers as defined in the E-Prime (Psychology Software Tools, 2016) tasks were sent via a parallel port and saved within an event marker channel. For data preprocessing, the EMG recordings were loaded into the PhysioData Toolbox (SjakShie, 2019) in which they were rectified and filtered with a 28 Hz high cut-off filter, a 500 Hz low cut-off filter, and a 50 Hz notch filter. For each trial, separate epochs were defined for the fixation period (1s), the first 500ms of stimulus presentation in which a neutral expression was shown (later defined as baseline), the subsequent 1500ms in which the emotional expression was shown (later defined as response), and the first 1500ms of the blank screen after stimulus presentation. Within these epochs, the EMG signal was downsampled by calculating the average signal within consecutive 100ms time bins. The combined data from all subjects was then exported into MATLAB for further preprocessing. First, an automated artifact detection, which was inspired by Dignath et al. (Dignath et al., 2019), was conducted. More specifically, for each subject and each muscle region, we checked the distribution of the EMG signal for extreme values (+/- 3.5 SDs) in the time bins regarding the absolute value of each time bin and the relative differences between subsequent bins. This was performed in relation to (1) the entire time interval of interest per trial (5 baseline time bins and 15 response time bins) and (2) the distribution of baseline time bins in the same position across trials. If more than 50% of all time bins (20) or more than 50% of the baseline time bins (5) belonging to one trial had extreme values, this trial was entirely excluded from the analysis. Otherwise, the respective time bins were replaced with missing values. Across all subjects, 17 trials (0.002%) were excluded from the corrugator analysis and 150 trials (0.02%) from the zygomaticus analysis. After excluding the marked time bins, a baseline correction was performed by subtracting the mean EMG activity of all baseline time bins belonging to one trial from the respective response time bins. The baseline-corrected EMG data was then z-scored for each participant and each muscle region. Furthermore, the data was summarized on a trial level by averaging the last second of each trial%u2019s response window (last 10 time bins) for each participant and each muscle region as well as by averaging across the two presentations of each of the 60 stimuli to end up with the same amount of observations as for the rating data (trial-averaged data; used as predictor in generalized linear mixed models). Lastly, the data was also summarized on a