Researchers at McGill University’s Brandon Lab have overturned a long-held assumption about the hippocampus, revealing the brain’s memory centre doesn’t just store the past – it actively predicts rewards. Published January 29, 2026, in Nature, the preclinical study used advanced imaging techniques to track brain activity in mice, demonstrating a structured reorganization of memories as animals learned a task. “The hippocampus is often described as the brain’s internal model of the world,” explains senior author Mark Brandon, Associate Professor in McGill’s Department of Psychiatry, “What we are seeing is that this model is not static; it is updated day by day as the brain learns from prediction errors.” This discovery offers a new framework for understanding learning and could unlock critical insights into the early stages of Alzheimer’s disease, where decision-making and learning are often impaired.
Hippocampal Neural Shifts Predict Rewards, Not Just Store Memories
Researchers at the Brandon Lab at McGill University, collaborating with Harvard University, have revealed a dynamic function of the hippocampus beyond simple memory storage—predicting future rewards. Published in Nature on January 29, 2026, the study demonstrates the brain region actively reorganizes memories to anticipate outcomes, a previously unobserved learning process. The team tracked neural activity in mice learning a task, utilizing novel imaging techniques allowing weeks-long observation of cellular changes.
The research challenged the assumption that shifts in hippocampal brain activity were random, instead proving they are structured and predictive. “What we found was surprising,” said Brandon, noting that neural activity initially peaking at reward delivery gradually shifted before the reward was received. This suggests the hippocampus supports a more complex form of reward learning than previously understood, building upon the principles demonstrated by Ivan Pavlov’s earlier work. Importantly, this predictive capacity offers a new lens through which to examine the early cognitive deficits seen in Alzheimer’s disease, potentially illuminating pathways for future restoration efforts.
Novel Brain Imaging Tracks Weeks-Long Neuronal Activity Changes
This innovation, first utilized in Canada by the Brandon Lab founded in 2015, allowed scientists to observe how the brain’s memory center adapts to predictable rewards in mice—tracking neuronal “glow” over extended periods. The team discovered that initial neural peaks coinciding with reward delivery gradually moved earlier, ultimately occurring before the reward was received, demonstrating the hippocampus isn’t merely storing memories but actively predicting outcomes. This structured reorganization challenges the long-held assumption of randomness in hippocampal activity shifts and offers a new perspective on sophisticated reward learning, moving beyond the simple cue-reward associations established by Pavlov’s experiments.
Predictive Hippocampus Dysfunction Linked to Early Alzheimer’s Impact
This predictive capacity, exceeding simpler reward associations demonstrated by Pavlov, positions the hippocampus as a sophisticated learning center integrating memory and context. The research suggests Alzheimer’s disease impacts this predictive function early on, explaining difficulties with learning and decision-making observed in patients. The Brandon Lab, founded in 2015, hopes this work will illuminate how this predictive signal may fail and be restored, potentially leading to new therapeutic strategies.
To achieve this unprecedented spatial and temporal resolution, the researchers employed advanced two-photon microscopy coupled with genetically encoded calcium indicators (GECIs). This complex setup allows for the chronic, long-term monitoring of individual neuronal ensembles within the dentate gyrus and CA1 regions of the hippocampus. Unlike previous methods that provided only generalized recordings of regional activity, this technique permits the measurement of single-cell firing rates across multiple days, providing the granular data necessary to distinguish true predictive encoding from mere post-hoc correlation. The subsequent analysis required advanced machine learning algorithms to deconvolute the signal from the noise inherent in chronic *in vivo* recording.
From a computational neuroscience perspective, the observed shift suggests the brain is leveraging predictive coding mechanisms, a concept where cortical areas attempt to minimize prediction error. Instead of waiting for the reward signal, the hippocampus appears to generate an internal forward model of the expected sensory and chemical cascade that precedes reward acquisition. This framework suggests that learning is less about association (A leads to B) and more about continuous hypothesis testing: the brain calculates the probability of an outcome based on current partial information, thereby making the prediction itself a biologically active state rather than a passive memory retrieval.
Translating these findings into human biomarkers presents significant challenges, necessitating the development of non-invasive imaging techniques capable of measuring the functional connectivity and temporal dynamics of hippocampal networks. Future efforts will likely focus on identifying quantifiable neurophysiological markers—such as specific patterns of prefrontal-hippocampal synchrony—that reflect the efficiency of prediction encoding. Confirming these circuits in human subjects, particularly in early cognitive decline, could establish new diagnostic metrics that detect predictive failure long before overt memory deficits manifest.
By showing that the healthy hippocampus helps turn memories into predictions, the study offers a new framework for understanding why learning and decision-making are affected early in Alzheimer’s disease and opens the door to research into how this predictive signal may fail and be restored.
