Poster #120 - Statistical learning of facial cues based on observed evidence in young children.

• In Event: Poster Session 14

Abstract

Young children rationally learn about the world. For example, infants can learn language statistically from voice patterns (e.g., Saffran et al., 1996). Recently, some researchers have suggested that statistical learning plays an important role in social cognition (e.g., emotion reasoning). Schulz & Gopnik (2004) showed that 4-year-olds could use statistical learning to identify the cause of others' emotions from the observed evidence. However, the social world evidence may be more uncertain than in the experiment. For example, a smiley face may be unpleasant, or an ambiguous expression may be happy. Therefore, this study examines whether young children can track probabilistic differences in observed evidence and use them for emotion reasoning.

Methods. One hundred and thirteen children aged 4- to 6-year-olds (63 girls, M = 67.41 months) participated. Children were randomly assigned to one of the conditions: 25% condition (n = 25), 50% condition (n = 29), 75% condition (n = 31), 100% condition (n =28). All children judged whether the four neutral expressions were happy or not (pre-test). Then, the children watched evidence composed of four positive and four negative events. In each condition, the frequency (i.e., the statistical data) of the protagonist's neutral face shown in each event was different (Table 1). Finally, children judged whether the four protagonist's neutral expressions were happy or not (post-test). If young children learn about others' emotions in a rational way, their reasoning should be different by evidence. We counted the number of neutral expressions as happy and the variation of judgment for each child. The variation in each child's judgment was classified into three categories and converted into ordinal data: decrease (-1), no change (0), and increase (+1) in the number of neutral facial expressions judged as happy.

Results. An ordinal logistic regression analysis was performed with the condition as predictor variables, and month age was entered as a covariate. There was a significant main effect of conditions, Waldχ2 (3) = 8.44, p = .038. The results showed that more likely "increase (+1)" in 100% condition (B = 1.30, SE = 0.57, Waldχ2 (1) = 5.18, p = .023), and 75% condition (B = 1.31, SE = 0.55, Waldχ2 (1) = 5.56, p = .018), than in 50% condition; see Figure 1. There was no significant difference between the 50% and 25% conditions, B = 0.21, SE = 0.56, Waldχ2 (1) = 0.14, p = .711.

Discussion. This study suggests that the evidence children observed influenced emotion reasoning in young children. Overall, children made stable inferences (i.e., the influence of evidence was small). Given that children can understand facial expressions early, the evidence they observed in this experiment may have been insufficient to change their beliefs. Nevertheless, we believe that the results point to the importance of individual differences in experience and statistical learning in young children's emotion reasoning.

Authors

• Minori Takahashi, The United Graduate School of Education Tokyo Gakugei University
  Presenting Author

• Keito Nakamichi, Chiba University