4 General Conclusions
4.3 MOTIVATING THE APPROACH
4.3.2 b Representations and their development
In order to explain how abstract concepts are constructed, we must
characterize the representations and computations that underlie mature reasoning in adults, in order to posit mechanisms that bridge from some initial state to that mature state (Carey, 2009). The research described in Chapter 3 provides a first pass characterization of the content and dimensionality of adults’ intuitive
knowledge of emotion. Both the behavioral and neural results suggest that space of emotions people attribute is best characterized by abstract properties of different emotions’ antecedent causes. Moreover, this work provides an initial sketch of a
13Of course, the theory of mind network is made up of several regions that presumably
differ in their precise computational roles. Future research might aim to understand how different components of the ToM network differentially contribute to the process of emotion perception and inference.
particular set of event properties that are reliably diagnostic of different emotional states.
Could this work provide a useful framework for investigating the development of human emotion knowledge? For example, although the simpler affective spaces like the circumplex model or the basic emotions model failed to capture the range of emotions that are attributed by human adults, it is possible that a simpler basis may be sufficient to capture the space of emotions that are represented earlier in development. Indeed, basic emotion theory was highly motivated by evolutionary, developmental and cross-cultural perspectives (Ekman et al., 1987; Ekman & Friesen, 1971; Izard, 1992; Shariff & Tracy, 2011), and might be a valid account of the representational resources available innately. The results of Chapter 3 suggest that the six basic emotions do not provide an adequate set of conceptual primitives for mature emotion inference, in that they do not provide the building blocks
necessary to explain the full range of mature emotion concepts. These basic emotions might nonetheless be important developmental primitives that are combined with other social-cognitive representations in the construction of more complex theories of emotion.
The approach used in Chapter 3 could be useful for characterizing the emotion representations available innately, but also for understanding the
emergence of new emotion knowledge over development. A large body of research has studied when different emotion concepts emerge in childhood, documenting stark differences in the developmental trajectories for different emotions (Banerjee, 2002; Ferguson, Stegge, & Damhuis, 1991; Nunner-Winkler & Sodian, 1988; Russell, 1990; Saarni, 1999; Wellman & Banerjee, 1991; Widen & Russell, 2010; Yuill, 1984). However, one limitation of much of this research is that it is not clear whether
children fail to attribute an appropriate emotion to a situation because they lack the social-cognitive or semantic knowledge necessary to parse the relevant properties of the event, or because they do not understand the emotion that those event properties elicit, or both.
The 38-dimensional feature space constructed in Chapter 3 provides a basis for teasing apart these alternatives: when children attribute inappropriate emotions to a situation, do they do so because they fail to ascribe the relevant properties to the event, or because they lack the theoretical knowledge that maps between abstract event properties and different emotional states? For example, it is not until age 8 that children reliably recognize that a person will feel embarrassed in
prototypical scenarios (e.g. “Joan spilled grape juice all over her white dress. All the kids laughed at her”; Widen & Russell, 2010). Using its appraisal profile from
Chapter 3, we can articulate the event regularities that are diagnostic of an embarrassed state. For example, compared to events that elicit devastation or disappointment, subjects rated the embarrassing events as more likely to be caused by an agent, more likely to be within control of the target character, more likely to involve the target character directly, and more likely to involve other people. With an explicit model of the event properties that are diagnostic of particular emotions (according to adult subjects’ intuitive theories), we could then ask whether
developmental differences in the attribution of embarrassment reflect change in the identification of these properties (e.g. understanding that spilling the juice was Joan’s fault and she might have avoided it by being more careful; recognizing that a purple stain on a white dress will be visually salient and draw the attention of people around Joan), or change in children’s understanding of the emotion that follows from these more abstract conditions. In previous studies on children’s
recognition of different facial expressions, researchers have made progress in understanding children’s perceptual confusions (e.g. fear with surprise, disgust with anger) by analyzing the similarity in facial action units between different pairs of expressions (Camras, 1980; Wiggers, 1982). Having an explicit account of the eliciting conditions that are similar and dissimilar between different emotions could provide a similar framework for understanding the errors children make in
attributing emotions based on situational information. In general, by decomposing our knowledge of emotions into a set of simpler (but still abstract) regularities in the eliciting conditions, we can make headway in understanding which aspects of this intuitive knowledge of emotions are available early in development, and what mechanisms of change might be necessary for expanding the scope of this early system into a full-fledged theory of others’ emotions14.