Aims: There remains no clear consensus on whether mixed emotions represent a unique affective state or are simply the co-occurrence of discrete positive and negative emotions. This study introduces an experimental protocol for the multimodal measurement and modelling of psychophysiology (i.e., electrodermal activity [EDA], skin temperature, heart rate [variability]), state and trait valence, and cognitive appraisal as components constituting emotion dynamics across time. Through this study, more information on how mixed emotions manifest and fluctuate in the short-term can be derived, enabling a better understanding of how mixed emotions differ from discrete emotions, and the time-course of emotional experiences in general.
Methods: A convenience sample of 58 students (44 female) enrolled as part of their research participation scheme. The cost-effective yet high-resolution EmotiBit wearable was used to collect physiological data through its photoplethysmography, EDA, and temperature sensors. Readings were synchronised to emotion-eliciting films presented through PsychoPy with LabStreamingLayer. Participants also continuously self-reported positive and negative affect using a PS4 game controller: left and right triggers corresponding to positive negative affect respectively. Crucially, the intensity of the emotion experience was directly proportional to the trigger depth, and participants were able to trigger both buttons at the same time to indicate mixed emotions with variable intensity. A longitudinal, repeated-measures, within-participants, experimental design was employed. Four-condition (positive, negative, mixed, neutral) emotion category was the within-subjects independent variable. Self-reported bivariate valence and physiological measures were the intensive-longitudinal, repeated-measures dependent variables.
Results: MANCOVA with post-hoc univariate ANCOVAs found that emotion condition only had significant effects on self-reported valence and not any physiological measures. However, idiographic network modelling with Graphical VAR revealed meaningful intra-individual and gender differences. For instance, males displayed denser contemporaneous networks whilst females displayed stronger cross-lagged effects in temporal networks, suggesting stronger co-activation of affect and physiology but weaker persistence of effects for males than females. Contemporaneous networks for mixed condition were notably sparser than all other conditions, suggesting possible mutual inhibition between opposing affective systems rather than additive co-activation.
Conclusions: Findings indicate that mixed emotions are unique from their constituent discrete states and highlight moderating effects of gender on interactions between affect and physiology. Contrasting results further underscore the limitations of traditional group-level, univariate/modal analyses in examining inherently dynamic and complex psychological processes. By combining multimodal measurement with idiographic network modelling, this study demonstrates a novel framework for investigating the structure and temporal dynamics of emotional experiences holistically.
