Statistical Learning and Emotion Category Representations

• In Event: 2-005 - How Cognitive Science Can Inform a New Perspective for Understanding Emotional Development

Integrative Statement

Although the configurations of facial muscles that humans perceive vary continuously, we represent emotions as categories. As in other domains of categorical perception (such as for speech and color), individuals must sort through vast amounts of data to develop categories that are meaningful in a given environment. Further, emotion categories may be dynamic to allow individuals to integrate variations in cultural and situational norms. We tested whether individuals use the statistical distribution of expressions in their environments to flexibly update emotion category representations. In Experiment 1, children (N = 91, 6-8-years-old) and adults (N = 105, 18-22-years-old) categorized morphed expressions of emotion as “upset” or “calm.” The expressions presented ranged from 0% angry (i.e., 100% neutral) to 100% angry. During a training phase, we gave participants feedback regarding their categorization choices such that morphs greater than 50% angry were defined as “upset” and those less than 50% angry were defined as “calm.” Both children and adults learned this explicitly taught category boundary. Next, the feedback was removed and participants were randomly assigned to one of three between-subjects conditions: 1) anger shifted (i.e., over-sampled angry morphs, midpoint shifted toward angry anchor), 2) neutral shifted (i.e., over-sampled neutral morphs, midpoint shifted toward neutral anchor), 3) unshifted (i.e., sampled morphs from same distribution as the training phase). Rather than maintaining the explicitly taught boundary between angry and neutral, participants rapidly shifted their category boundaries according to the statistical distribution observed (χ2(2) = 453.66, p < .001; Figure 1). While children and adults shifted their category boundaries to the same extent (χ2(2) = 3.09, p = .214), children’s category boundaries were less steep that those of adults (χ2(1) = 27.52, p < .001). That adults had somewhat steeper category boundaries compared to children likely reflects that adults have more precise category representations of facial expressions. It may be that categorization becomes more robust and efficient with experience, particularly as adults prioritize facial cues, whereas children are learning how to interpret faces and divide their attention equally between faces and other contextual information (Leitzke & Pollak, 2016). In Experiment 2, we directly compared flexibility of emotion category boundaries against those of other, non-emotional, categories across early development (N = 31, children ages 4-11-years-old). Children updated category boundaries for both emotion and non-emotion category representations; however, the direction of the change varied according to domain (χ2(1) = 357.90, p < .001); in particular, children became desensitized to extreme emotion expressions when they occurred frequently in the stimuli. These experiments demonstrate that children change their category boundaries based on exposure to a particular statistical distribution of facial expressions. Such flexibility allows individuals to adjust emotion concepts across contexts and organize appropriate behavioral responses based on the available socio-cultural cues.

Authors

• Rista C Plate, University of Wisconsin-Madison
  Presenting Author

• Adrienne Wood, University of Wisconsin-Madison
  Non-Presenting Author

• Kristina Woodard, University of Wisconsin-Madison
  Non-Presenting Author

• Seth D Pollak, University of Wisconsin-Madison
  Non-Presenting Author