Computational and neural signatures of volatility learning in infancy: Links with temperament and the early caregiving environment

Le résumé fourni par la source

Infants demonstrate a remarkable ability to learn from statistical patterns. Yet, in probabilistic environments, successful learning requires distinguishing random variability from true changes in the underlying contingency (i.e. volatility). However, the extent to which successful volatility learning in infancy is modulated by environmental and temperamental differences is still unclear, as well as the extent to which infant neural processing reflects learning from changes. To answer these questions, we investigated the computational and neural basis of volatility learning in 11-month-old infants and examined how temperament (Infant Behavior Questionnaire-Revised; Very Short Form) and features of the early caregiving environment (Comprehensive Early Childhood Parenting Questionnaire) shape individual differences in this process.Infants (N=77, Mage=331.13±15.25 days; 32 girls, 45 boys) observed a probabilistic environment in which the most likely location of a target periodically changed. Infants learned these contingencies, as reflected in their anticipatory looking, demonstrating sensitivity to environmental volatility. To characterize the computational processes underlying this learning, we used a Hierarchical Gaussian Filter model, which captures how learners update their beliefs about both the current state of the environment and its volatility. Prediction errors derived from this model, in combination with anticipation accuracy, predicted EEG activity 588–714 ms after feedback across a broad fronto-central region, providing evidence that the cognitive mechanisms captured by the model are represented in the infant brain. Additionally, we found that individual differences in volatility learning were primarily captured by model-derived change sensitivity: infants with a stronger change sensitivity detected reversals in the volatile environment better and adapted their behavior more rapidly. Change sensitivity was positively related to IBQ Effortful Control, suggesting infants with stronger regulatory capacities also updated their beliefs about volatility in a more optimal fashion. infants with higher Effortful Control detected environmental reversals more rapidly and showed more optimal belief updating. Finally, Positive Discipline was associated with higher Effortful Control, although its indirect association with change sensitivity was not significant. Finally, in a Structural Equation Model, we found that higher Effortful Control was predicted by stronger Positive Discipline, although the indirect pathway to change sensitivity was not significant. Together, these findings identify early computational and neural signatures of hierarchical volatility learning and highlight emerging temperament and caregiving influences on adaptive learning in infancy.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.