Page 79 - Demo

Physiological Resonance and Interpretation of Emotional Expressions773Subtle facial cues. In the last SCL model, the linear polynomial was again identified as significant predictor, Flinear (1,140024) = 8.855, p = .003, as were the interactions between all three polynomials and emotion category, Flinear*category (3, 140024) = 16.339, p < 0.001; Fquadratic*category(3, 140024) = 45.746, p < 0.001; Fcubic*category (3, 140024) = 11.745, p < 0.001. Thus, the presence of facial signs of emotional involvement, without the context of prototypical emotion displays, also affected SCL properties: based on the statistics (Table 3 in Online Resource 3) and predicted time courses (Fig. 4c) for the three cue types versus neutral (no cue), the SCL signal decreased to a lesser degree for faces with an added blush and faces with dilated pupils, with even a slight late increase for the latter. Moreover, when observing faces with added tears, SCLs of participants increased steeply, with a peak around 2.5s and a fast decline. Importantly, the coefficient for the interaction between the quadratic trend and tears cue category was the only coefficient which was consistently below 0 in the bootstrap samples, pointing out the stability of the observed peak in SCL for tears. Skin temperaturePrototypical facial expressions. While only the linear polynomial and the cubic polynomial were significant predictors of the SKT signal in the response window, Flinear (1,182620) = 5.622, p = .018; Fcubic (1, 182620) = 4.909, p = 0.027, all interactions between the three polynomials and emotion category became significant model terms, Flinear*category (4, 182620) = 8.518, p < .001; Fquadratic*category(4, 182620) = 6.948,p < .001; Fcubic*category (4, 182620) = 4.757, p = .001. Emotional versus neutral facial expressions therefore also seemed to affect changes in SKT differently. Looking at the model statistics (Table 4 in Online Resource 3) and predicted value plots (Fig. 5a), there was a stronger increase in SKT following happy and fearful expressions and a diminished late increase following angry expressions compared to neutral ones. In addition, after an initial increase, cheek temperature already declined after approximately 6s for sad and fearful expressions while this was not the case for the other facial expression categories. Importantly, no coefficient for any predictor was consistently larger or smaller than 0 in the bootstrap analysis. Bodily expressions. In the model describing SKT changes associated with viewing bodily expressions of emotions, the linear polynomial as well as the three interactions between each polynomial and emotion category were significant, Flinear (1, 182845) = 4.220, p = .040; Flinear*category (4, 182845) = 9.937, p < .001; Fquadratic*category(4, 182845) = 20.160, p < .001; Fcubic*category (4, 182845) = 6.151, p < .001. Examining the effect of emotion in a body posture on the shape of the signal more closely, SKT