Abstract Representations of Attributed Emotion: Evidence From Neuroscience and Development

Abstract

Humans can recognize others’ emotions based on overt cues such as facial expressions, affective vocalizations, or body posture, or by recruiting an abstract, causal theory of the conditions that tend to elicit different emotions. Whereas previous research has investigated the recognition of emotion in specific perceptual modalities (e.g. facial expressions), this dissertation focuses on the abstract representations that relate observable reactions to their antecedent causes. A combination of neuroimaging, behavioral, and developmental methods are used to shed light on the mechanisms that support various forms of emotion attribution, and to elucidate the core features or dimensions that structure the space of emotions we represent. Chapter 1 identifies brain regions that contain information about emotional valence conveyed either via facial expressions and or via animations depicting abstract situational information. These data reveal regions with modality-specific representations of emotional valence (i.e. patterns of activity that discriminate only positive versus negative facial expressions), as well as modality-independent representations: in medial prefrontal cortex (MPFC), the valence representation generalizes across stimuli, indicating a common neural code that abstracts away from specific perceptual features and is invariant to different forms of evidence. Building on evidence that young infants discriminate and respond to the emotional expressions of others, Chapter 2 investigates whether infants also represent these expressions in relation to the situations that elicit them. The results of several experiments demonstrate that infants within their first year of life have expectations about how facial and vocal displays of emotion relate to the valence of events that precede them. Whereas Chapters 1 and 2 focus on a simple binary distinction between positive and negative affect, Chapter 3 investigates a space of more fine-grained discriminations (e.g. someone feeling proud vs. grateful). A combination of multi- voxel pattern analyses and representational similarity analyses reveal brain regions containing abstract and high-dimensional representations of attributed emotion. Moreover, a set of causal features (encoding properties of eliciting events that vary between different emotions) outperforms more primitive dimensions in capturing neural similarities within these regions. Together, these studies provide a newly detailed characterization of the representations that structure emotion attribution, including their development and neural basis.