Reinforcement learning increasingly relates to memory specificity from childhood to adulthood.

In some contexts, abstract stimulus representations can effectively promote reward pursuit, whereas in others, detailed representations are needed to guide choice. Here, we ask how, across development, the reward statistics of the environment influence the specificity of both value-guided learning computations and recognition memory. Across two experiments (N = 224), we show that participants ages 8 - 25 years adaptively up- and down-weight detailed versus broader stimulus representations and that these learning computations relate to mnemonic specificity. When participants place greater weight on granular representations during learning, they better remember stimulus details, whereas when they place greater weight on broader representations, they show enhanced memory only for categorical information. Moreover, the strength of the coupling between learning and memory specificity increases with age. We demonstrate that from early in life, reward shapes the granularity with which the world is partitioned, and increasingly across development, the specificity with which experiences are remembered.