Neural and behavioral correlates of caregiver-specific reinforcement learning in children
Year: 2026
Type: preprint
Abstract: This project examines how children process and learn from socioemotional cues that differ in relational meaning. Children’s social development depends on their ability to detect, interpret, and learn from others’ emotional expressions and feedback. However, most studies of emotion processing and reinforcement learning in children use unfamiliar faces, generic social stimuli, or nonsocial rewards. Less is known about whether children learn differently from emotional cues provided by their own caregiver compared to an unfamiliar adult, and whether these differences are associated with clinically meaningful individual differences such as callous-unemotional (CU) traits, internalizing symptoms, externalizing/aggression, emotion dysregulation, and caregiver-child relationship quality. The study will recruit approximately 80 mother-child dyads with children ages 8–10 years from the greater Tuscaloosa area. Recruitment will use community-based convenience sampling with targeted oversampling for CU traits to ensure variability in the primary individual-difference construct. Data will be collected across two visits. Visit 1 includes consent/assent, eligibility confirmation, parent- and child-report questionnaires, clinical interviews, cognitive testing, behavioral tasks, and standardized photography of the mother and child for use in personalized task stimuli. Visit 2 includes child fMRI scanning while children complete tasks designed to measure caregiver-specific socioemotional learning, observational learning, emotion categorization, and resting-state functional connectivity. The study includes four primary task domains. First, the SERL-R task assesses direct reinforcement learning and reversal learning from mother, stranger, and nonsocial/money feedback. Second, the OSERL task assesses observational learning from mother versus stranger advisors, including how children use advisor choices and emotional feedback to guide their own decisions. Third, the MSEC task assesses rapid categorization of mother versus stranger emotional expressions, focusing on calm-versus-fearful and calm-versus-angry judgments. Fourth, resting-state fMRI assesses intrinsic connectivity among affective, reward, salience, regulatory, and social-cognitive networks that may support caregiver-specific socioemotional learning. The primary purpose of the project is to determine whether caregiver emotional cues function as uniquely salient learning signals during childhood. Behavioral and computational models will be used to estimate learning rates, prediction-error weighting, feedback sensitivity, advisor weighting, reversal learning, response bias, and emotion sensitivity. Trial-level fMRI analyses will examine neural responses to caregiver versus stranger cues, source-specific value and prediction-error signals, feedback-related updating, source-by-emotion effects, and functional connectivity among regions involved in affective salience, valuation, social cognition, and regulatory control. Resting-state analyses will test whether intrinsic network organization predicts task-based behavioral and neural sensitivity to caregiver versus stranger cues. The project also aims to clarify how CU traits and related clinical dimensions are associated with social-affective learning. We expect that children will show behavioral and neural differentiation of mother versus stranger cues, particularly when cues are emotionally meaningful or learning-relevant. We further expect that CU traits will be associated with reduced sensitivity to distress-related cues, punishment-related feedback, affiliative reward, and caregiver-specific relational value. In contrast, anxiety/internalizing symptoms may be associated with heightened sensitivity to threat-related caregiver cues, whereas aggression/externalizing symptoms and emotion dysregulation may be associated with altered anger processing, feedback sensitivity, reversal learning, and regulatory coupling. Expected outcomes include a clearer understanding of how children learn from emotionally meaningful caregiver cues compared to unfamiliar adult cues, how these processes are represented in the developing brain, and how they vary as a function of CU traits and broader psychopathology-relevant dimensions. The study is intended to advance mechanistic models of socioemotional learning in childhood and identify behavioral, computational, and neural markers that may help explain individual differences in empathy, conscience development, emotion regulation, aggression, and risk for persistent antisocial behavior.
Authors: Castagna, Peter James, Seung-Lark Lim
Language: English
Subfield: Clinical Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7163524368