Infants track the statistics of emotion transitions in the home

• In Event: How do the statistics of the environment shape the infant mind across domains?

Abstract

Children experience highly dynamic emotions embedded in a rich statistical structure. Prior research shows that adults follow predictable transitions between emotions, e.g., positive emotions more often follow other positive (vs. negative) emotions (Thornton & Tamir 2017; 2020). Further, adults accurately estimate the likelihood of a given transition (Thornton & Tamir, 2017; Zhao et al., 2020), and they predict upcoming emotions (Thornton et al., 2019). How do infants build toward this mental model of the statistics of emotion transitions? How do individual differences in infants’ early experiences shape their expectations about emotion?
To answer these questions, we measured the pupillary responses of 54 infants (4-10 months; pending full preregisted sample) to isoluminant videos of transitions between emotions (Figure 1a), accompanied by a corresponding emotional vocalization.
We first tested whether infants track the naturally occurring statistics of emotion transitions. We measured the pairwise temporal alignment in pupillary responses across participants, using dynamic time-warping distance (Giorgino, 2009). This measure has been linked to heightened engagement and similar interpretation of a stimulus (Nencheva et al., 2021; Kang & Wheatley, 2017). We predicted that infants would be more aligned in their pupillary responses to transitions that were on average more (vs. less) frequent among caregivers. This prediction was confirmed by a mixed effect model with random intercepts by emotion and participant (𝛽=0.05, t(5694.18)=3.52, p=0.0004), controlling for total pixel change during the video. This suggests that infants track the statistical regularities of emotion transitions, and converge in their processing of familiar transitions.
Next, we tested whether individual differences in caregivers’ emotion dynamics correspond to differences in infants’ emotion processing. To do so, we asked primary caregivers to report how likely they are to experience each transition in everyday life (Figure 2a; Thornton & Tamir, 2017). This allowed us to compute the average likelihood of each transition and to measure individual differences in the emotion transitions that infants may observe at home. We explored links between individual differences in caregiver emotion transitions and infant processing. We found that infants who observe more "typical" emotion transitions at home process emotion transitions in a more typical way. In a linear model, the average alignment of the pupillary responses of a given infant was predicted by how different their caregiver’s transitions were from other caregivers in the sample (𝛽=-0.48, t(44)=-3.13, p=0.003). Further, pupillary alignment between any two infants in the sample marginally correlated with the difference in their primary caregivers’ reported emotion transitions, as revealed by a mixed effect model with random intercepts by participant (𝛽=-0.05, t(1050.59)=-1.69, p=0.09).
Together, these findings suggest that infants’ emotion processing is tightly linked to the emotion statistics they observe at home. Using a statistical learning approach, this investigation provides a new window into the cognitive mechanisms that support infants in integrating the complex statistics of their environment to form predictions, and the origins of individual differences in emotion development.

Authors

• Mira Nencheva, Princeton University
  Presenting Author

• Diana I. Tamir, Princeton University

• Casey Lew-Williams, Princeton University