TL;DR
This study reveals how children's value-based learning and brain development change over two years. It shows that while the striatum supports learning generally, the hippocampus becomes increasingly important for learning with delayed feedback as children mature, suggesting developing memory systems are less specialized than in adults.
What Problem Does This Paper Address?
Understanding how children learn and make decisions is crucial for education and development. While we know that in adults, different brain regions like the hippocampus and striatum play distinct roles in learning based on when feedback is received (immediate vs. delayed), it's unclear how these systems develop and contribute to learning during childhood. Specifically, how do children's learning strategies and the underlying brain structures change over time, and how do these changes relate to each other? This paper aims to fill this gap by longitudinally tracking children's value-based learning and brain volume over two years.
Methods
The researchers used a combination of behavioral experiments, computational modeling, and neuroimaging.
- Reinforcement Learning Task: Children performed a task where they learned to associate cues with choices that led to rewards. The key manipulation was the timing of feedback: either immediate or delayed after their choice.
- Computational Modeling: They used reinforcement learning models (specifically, a Rescorla-Wagner model) to quantify learning behavior. This involved estimating parameters like learning rate (how much new information influences decisions) and inverse temperature (how much decisions are guided by learned values vs. randomness).
- Longitudinal Brain Imaging: Structural MRI scans were used to measure the volume of the striatum and hippocampus at two time points, two years apart.
- Latent Change Score (LCS) Models: These statistical models were used to analyze how changes in brain volume and learning parameters over time were related.
Datasets
The study used data collected from children at two time points, approximately two years apart.
- Participants: 140 children aged 6-7 at the first wave, with 126 returning for the second wave. A subset also underwent MRI scans.
Key Insights and Ideas
- Developing Brain, Developing Learner: Children's learning abilities improve significantly over two years, becoming more accurate and faster. This is mirrored by changes in their computational learning parameters, moving towards more "optimal" strategies.
- Hippocampus Catches Up for Delayed Feedback: While the striatum seems to support learning regardless of feedback timing, the hippocampus's role becomes more pronounced for learning with delayed feedback as children mature. This suggests a developmental trajectory where the hippocampus takes on a more specialized role.
- Less Specialized Brains in Childhood: Compared to adults, children's memory systems (hippocampus and striatum) appear less differentiated. They show more overlap in their contributions to learning, suggesting a more cooperative, less specialized neural architecture during middle childhood.
- Brain Structure Predicts Learning Trajectory: Larger striatal volume was associated with better learning overall, both immediate and delayed. However, larger hippocampal volume was specifically linked to better delayed feedback learning, highlighting its developing role.
Key Results
- Behavioral Improvements: Children showed increased accuracy, faster reaction times, increased win-stay behavior, and decreased lose-shift behavior over the two-year study.
- Computational Model: The value-based learning model best described children's behavior. Learning rate and inverse temperature increased over time, moving towards optimal values.
- Brain Volume Changes: Both striatal and hippocampal volumes increased, with the hippocampus showing more protracted maturation.
- Brain-Cognition Links:
- Larger striatal volume was associated with better immediate and delayed learning scores.
- Larger hippocampal volume was specifically associated with better delayed learning scores.
- These brain-cognition links were longitudinally stable.
Limitations and Future Work
- Episodic Memory Nuance: While there was a trend for enhanced episodic memory with delayed feedback, it wasn't consistently significant in the full dataset, suggesting further investigation is needed. The authors suggest this might be due to the protracted development of hippocampal maturation.
- Striatal Specificity: Contrary to predictions, striatal volume was associated with both immediate and delayed learning, not just immediate learning as hypothesized. This suggests a more general role for the striatum in value-based learning during childhood.
- Future Directions: The authors suggest future studies could use functional neuroimaging to clarify memory system differentiation and explore how factors like stress influence hippocampal development and learning. They also propose modeling reaction times and using tasks that allow for independent manipulation of different phases to better disentangle model parameters.
Summary updated