Past experience influences new learning through memory reactivation



Journal Title

Journal ISSN

Volume Title



How we learn and remember new information is influenced by our past experiences. The impact of prior knowledge on learning is at the heart of many computational and theoretical models of memory, which posit that our past is able to influence the present by reactivating memories during learning. Neuroimaging work in humans has shown that memory reactivation occurs through the reinstatement of patterns of activation from encoding in neocortex, a process that is driven by the hippocampus. Because memory reactivation allows for new experiences and related memories to be coactive during learning, studying memory reactivation presents an opportunity to explore how the past and present interact in the brain. The research in this dissertation uses behavioral testing and high-resolution functional magnetic resonance imaging (fMRI) to test the hypothesis that learning behavior and memory representation in the human brain are influenced by the reactivation of memories that are related to new learning experiences. The relationships between reactivation, behavior, and memory representation were investigated in four experiments. Experiment 1 tested whether hippocampal-cortical interactions underlie the reactivation of memories in new contexts to support generalization behavior. As with retrieval, hippocampus guided the reactivation of relevant memories in cortex during generalization. Experiment 2 examined how memory reactivation during learning influences representations for related events in hippocampal subfields dentate gyrus/CA₃ (DG/CA₃) and CA₁. Reactivation led to dissociable memory signatures across subfields, with stronger reactivation leading DG/CA₃ to differentiate related memories and CA₁ to integrate related memories. Experiment 3 tested whether perceptual similarity between related events further modulates subfield representations. Consistent with prior work in humans and animals, DG/CA₃ representations were sensitive to the degree of perceptual similarity across learning. Finally, Experiment 4 tested whether reactivating a memory through retrieval prior to learning related information influences the ability to infer indirect relationships across learning episodes. Retrieval before learning led to faster and more accurate across-episode inference decisions. Overall, the findings of this dissertation exemplify the importance of memory reactivation for how new information is learned and represented in the brain and reveal the conditions that allow the past to influence how we learn in the present.



LCSH Subject Headings