Page 69 - Demo

Physiological Resonance and Interpretation of Emotional Expressions673Skin conductance. The electrodes measuring changes in SCL were attached to the index finger and the ring finger of the participant%u2019s non-dominant hand. Data was recorded with the EDA 100C Biopac Systems module from (2000 Hz sampling rate, Gain: 5%u00b5V, 10Hz low-pass filter) and event triggers were sent from the presentation software via parallel port. Within the PhysioData Toolbox (Sjak-Shie, 2019), the recorded data was filtered with a 2Hz low-pass filter (Ch%u00eanes et al., 2013). Skin temperature. A fast response thermistor (TSD202A, Biopac) was placed below the participants%u2019 right cheekbone to record changes in cheek temperature. Data was acquired with the SKT100C Biopac Systems module (2000 Hz sampling rate,: Gain 2%u00b0F/V, 10Hz low-pass filter). Similar to the other measures, the PhysioData Toolbox (Sjak-Shie, 2019) was used for further filtering (1Hz low-pass; Ch%u00eanes et al., 2013). Data analysisIn order to shed light on different aspects of the processing of emotional expressions, we defined three different analyses aiming at the investigation of (1) subjective interpretation, (2) physiological signal changes and (3) the linkage between the two levels, see Fig. 2 for a visualization and further explanation. Since the study was not specifically designed to perform the third analysis, it should be considered as a pilot test and further information can only be found in Online Resource 4. Prior to the analysis of the data, we looked for irregularities in each dependent variable. Importantly, for the physiological measures, we integrated information from a repeated visual inspection with statistical and literaturebased thresholds. An overview of the outlier criteria can be found in the Online Resource 1. In addition, missing trials in the EMG, SKT and SCL recordings were replaced with missing values (subject 8: 3 trials and subject 21: 2 trials). The data for all physiological channels within the windows of interest was downsampled by exporting average values within five 100ms time bins prior to stimulus onset for the baseline window and 75 100ms time bins after stimulus onset for the response window. Lastly, a baseline correction was performed by subtracting the baseline from all data points of the corresponding response window for each trial. While the entire response window (4 seconds stimulus presentation and 3.5 seconds blank screen) was used in the analysis of the relatively slowly changing SCL and SKT signals (M. E. Dawson et al., 2016; Shearn et al., 1990), EMG activity was only examined during stimulus presentation (Kret, Roelofs, et al., 2013; Kret,