Page 77 - Demo

Physiological Resonance and Interpretation of Emotional Expressions753indicated by adding a bracket (red = between categories within bodily expressions, blue = between categories within facial expressions, grey = within category across modalities OR between subtle cue types). Straight line = p < .001, dashed line = p < .01, dotted line = p < .05Subtle facial cues. A separate model on the perceived intensity of the subtle facial cues revealed that the presence of a cue was a significant predictor of the intensity rating, F(3, 2097) = 669.31, p < 0.001. Crucially, faces with dilated pupils were rated equally intense as the same expressions with average pupil sizes (neutral). In contrast, stimuli with a blush received higher ratings than both neutral faces and faces with dilated pupils. Faces with tears were rated as significantly more intense than faces with the two other cue types and compared to neutral (see Fig. 3c and Tables 7 and 8 in Online Resource 2).Physiological results (Analysis 2)Skin conductancePrototypical facial expressions. In the LMM, the linear polynomial was a significant predictor of the changes in SCL, Flinear (1,181345) = 9.457, p = 0.002. Further, all interactions between emotion category and the three polynomials were significant, Flinear*category (4,181345) = 5.596, p < 0.001; Fquadratic*category(4,181345) = 12.274, p < 0.001; Fcubic*category (4,181345) = 15.145, p < 0.001, indicating that the shape of the signal differed for emotional as compared to neutral expressions. Looking at the t-statistics (Table 1 in Online Resource 3) as well as the predicted value graphs (Fig. 4a) for distinct emotion categories, the presentation of angry, happy and sad facial expressions were more strongly associated with an initial peak at around 2s and a decline over time which was strongest for happy expressions. A cubic component in the signal was observed following fearful faces, which however was not as strong as the other categories and without the pronounced peak at the beginning. Notably, only the interaction between angry facial expressions and the cubic trend did not include 0 in the bootstrap confidence intervals for the model coefficients, indicating that exclusively this effect was robust. Bodily expressions. As for the model on facial expressions, the linear polynomial significantly predicted SCL measurements, Flinear (1,181420) = 9.981, p = 0.002. In addition, the linear and cubic polynomials were involved in significant interaction terms with emotion category ,Flinear*category (4, 181420) = 22.935, p < .001; Fcubic*category (4, 181420) = 5.541, p < .001, suggesting that the expression of emotion via the body also had an effect on the shape of SCL measurements. In this modality, however, only happy and, to lesser degree, fearful expressions were related with