Emotion Experience from Stories, Videos and Everyday Life: Structure and Individual Differences

Citation

Han, Yanting (2022) Emotion Experience from Stories, Videos and Everyday Life: Structure and Individual Differences. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/yhmt-9t69. https://resolver.caltech.edu/CaltechTHESIS:05272022-181044242

Abstract

Most studies of emotion have as their subject matter the emotion experiences that people can describe and rate. By contrast to this approach from psychology, studies in animals, and some biological studies in humans, focus on behavior and its adaptive function. These two literatures typically use very different corresponding features by which to characterize emotion: categories or dimensions describing feelings for which we have convenient words, for the former (e.g., happiness, pleasantness), and functional properties for the latter (e.g., persistence, generalizability, approachabil- ity). In this thesis I use both sets of ratings, and I ask whether the latter, biologically inspired features could also be used to characterize people’s emotion experiences, and might reveal novel dimensions of variability. They also typically use different sets of stimuli to induce the emotions: lexical stimuli in which participants are asked to imagine something hypothetical are common in human studies; ecologically valid stimuli that at least the subjects cannot distinguish from the real world are common in animal studies. Here I used three domains of stimuli: stories, videos, and real-life experiences, in the same set of participants, permitting a unique comparison.

I took advantage of a sample of approximately 1000 Americans who were surveyed longitudinally over the internet during the COVID-19 pandemic. I collected ratings of emotion experiences evoked by three classes of stimuli: a validated set of short stories, a validated set of short videos, and actual experiences in real life across multiple waves. I found that all three types of emotion experiences could be characterized by low dimensional spaces, with the first two factors that accounted for most of the variance in people’s ratings corresponding to the dimensions of valence and arousal, in line with prior work. However, I discovered additional novel factors related to generalizability (the extent to which an emotion experience is shared across many different situations and occurrences) or modularity (the extent to which an emotion experience is unique to specific situations). The findings show that emotion features not usually assessed in humans can be recovered from subjective ratings of their experiences. I argue for a revision of current dimensional theories of emotion: they have been incomplete because they were restricted to ratings entrenched in how we think of our conscious experience, and the typical English words we use to describe it. The new dimensions validate some theories of emotion and offer hope for linking psychological studies in humans with behavioral or neurobiological work across species. I also characterized the distributions of the three types of emotion experiences and found that emotions were distributed along continuous gradients, with no well-separated clusters even for emotions belonging to the six basic emotion categories.

My thesis presents two additional topics that capitalize on my unique sample: the emotions experienced during the COVID pandemic, and individual differences. For example, I also found that resilience buffered individuals against the effect of loneliness on depression, and that people who had tested positive for COVID felt more morally disgusted towards acts of violating social norms. I also explored the association between psychological traits and differences in emotion experiences both in terms of the magnitudes of the ratings and the overall correlation structure across scales. Again, the richness of my dataset reveals a number of associations that are theoretically interesting and that will be of relevance to understanding mood and anxiety disorders as well.

All of the data will be made publicly available, and the core parts of many of the investigations were pre-registered.

Thesis Files