Hippocampus Functions as Predictive System Updating Internal World Model

Recent preclinical research is fundamentally reframing the role of the hippocampus, the brain's primary memory center, suggesting its core function is to anticipate future occurrences rather than solely serving as an archive of past events. This paradigm shift, detailed in a study published in Nature, posits that memory actively constructs and refines an internal model of the world based on prediction errors—the discrepancies between expected and actual outcomes.

Senior author Mark Brandon, an Associate Professor in Psychiatry at McGill University and a researcher at the Douglas Research Centre, stated that this model is not static but is continuously updated as the brain learns from these errors. The investigation, spearheaded by the Brandon Lab at McGill University in collaboration with Harvard University, utilized advanced calcium imaging techniques to track neuronal activity in mice during reward-based learning tasks. Researchers observed a temporal displacement in neural firing patterns over several weeks: activity that initially peaked upon reward delivery gradually shifted backward in time to occur before the animal reached the goal.

This forward shift in neuronal activation provides a visible neural correlate, demonstrating that memory encoding prepares the organism for forthcoming events, moving reward-based learning from an abstract concept to a tangible mechanism involving synaptic plasticity. This research builds upon prior work suggesting the hippocampus encodes future trajectories, segregated by theta rhythm cycles, in conjunction with the prefrontal cortex.

This forward-looking function of memory has significant clinical ramifications, particularly concerning neurodegenerative conditions that severely affect the hippocampus, such as Alzheimer's disease. The impairment observed in these diseases may stem not only from an inability to recall the past but crucially from a diminished capacity to learn from experience and accurately forecast subsequent events. Disrupting this predictive mechanism could account for early deficits in decision-making and learning associated with the disease progression.

The Brandon Lab, founded in 2015, investigates how memory mechanisms break down in Alzheimer's to devise strategies for restoration. Understanding the hippocampus as a predictive engine offers novel avenues for therapeutic intervention aimed at restoring the brain's ability to update its world model and minimize prediction errors. The findings align with reinforcement learning principles, suggesting hippocampal dynamics follow a Temporal Difference learning model, where the representation of reward decreases while cues predicting the reward strengthen over time.

Furthermore, prior research indicates that prediction errors, or surprise, prompt the hippocampus to abandon ongoing event representations and switch to an externally-oriented processing mode to support memory updating. This mechanism allows the brain to adapt rapidly to changing environments without needing to run computationally expensive future simulations. Dr. Brandon's research portfolio at McGill includes deciphering the neural code for spatial location and investigating early Alzheimer's effects on spatial memory networks, positioning this predictive coding discovery within a broader context of hippocampal function. This work solidifies the concept of memory as an essential, dynamic mechanism for navigating and successfully interacting with the future environment.