Children and adults use distinct neurocognitive mechanisms to support successful memory-based inference

Abstract

Reasoning depends on the ability to connect information across distinct experiences to derive knowledge that was never directly observed, and this capacity exhibits protracted developmental improvement that extends into emerging adulthood. Despite extensive behavioral work, it remains unclear whether developmental gains in inference reflect quantitative strengthening of a single mechanism or qualitative changes in the knowledge representations and computations that support inference decisions. Here, we tested the hypothesis that improvements in inference are linked to maturation of the hippocampus and posterior parietal cortex, resulting in age-related differences in how inference decisions are computed. We predicted that children (7-12 years) would rely on an iterative retrieval mechanism, requiring retrieval and combination of multiple distinct memories at the time of inference, whereas adults would be able to directly retrieve inferred relationships from structured representations that encode shared relations across experiences. Using functional MRI combined with computational modeling of response times, we show that hippocampal activity predicts successful inference via an iterative retrieval mechanism in both children and adults. Critically, only in adults does angular gyrus activity predict inference via a distinct, direct retrieval mechanism, consistent with access to inferred relationships represented as either integrated memories or in geometrically aligned neural spaces that organize events according to their shared relational structure. These findings identify a developmental shift in the neural mechanisms that support inference, demonstrating that maturation of posterior parietal cortex enables access to representations that capture derived linkages across experiences, fundamentally changing how inference decisions are computed across development.