Page 182 - Demo

Chapter 6180recognition (eleven ascending values, treated as continuous) and (3) Perceived emotional intensity (eleven ascending values, treated as continuous). Happy facial expressions were recognized at ceiling performance in all three groups (overall <1% inaccurate responses, see also Table S2), resulting in little variance in Emotion recognition accuracy to predict. As a consequence, we investigated how both activity over the corrugator region and zygomaticus region, which tends to be decreased and increased, respectively, when mimicking happy expressions, would differentially relate to the confidence in the recognition of the expressions as well as their perceived emotional intensity but not emotion recognition accuracy. Hence, Corrugator activity (baseline-corrected and z-scored), Zygomaticus activity (baseline-corrected and z-scored) and Group (AS, SA, NC), as well as both twoway interaction between one muscle region and group, became predictors of (4) Confidence in emotion recognition of happy expressions and (5) Perceived emotional intensity of happy expressions in two separate LMMs. The addition of a three-way interaction (Corrugator activity*Zygomaticus activity*Group) did not improve either model fit. Lastly, as changes in skin conductance in response to emotional stimuli have been shown to reflect emotional arousal in general rather than being emotion-specific, the three final (G)LMMs of the main analysis included all emotional expressions, with the exception of the at ceiling recognized happy expressions in the Emotion recognition accuracy model. We thus predicted (6) Emotion recognition accuracy, (7) Confidence in emotion recognition and (8) Perceived emotional intensity by Emotion category (anger, fear, happiness (not in accuracy model), sadness, neutral), Skin conductance (log-transformed maximum deflection) and Group (AS, SA, NC), as well as their two-way interactions and the three-way interaction. Based on research highlighting the role of alexithymia in altered interoception (e.g., Brewer et al., 2016) and emotion processing (D. A. Trevisan & Birmingham, 2016) as well as on observing significant group differences in depressive symptoms (see Table S2), we re-fitted all models with standardized alexithymia and depressive symptom scores as covariates, which did not change our results in a meaningful way. As mentioned above, we additionally fitted Bayesian mixed models to have more certainty not only with regard to the alternative but also the null hypothesis. To evaluate the independence of observations on prior specifications, we fitted models with narrow (normal(0, 1)), medium (normal(0, 2.5)) and wide (normal(0, 5)) priors. All statistical analysis were conducted in R 4.2.2. More technical details regarding the analyses can be found in the Supplemental Materials.