The brain’s ability to store and remember memories is at its core, but its mechanism is a field under continuous investigation. In a new work, MIT researchers introduce a new computational model for how the loop between the hippocampus and the entorhinal region stores both episodic and spatial memories, and it introduces a new theory for creating memories, one that discredits current hypotheses in proposing that neural structures involved in navigating through a space can form a basis for uniting together episodic memories.

Building Blocks of Encoding Memories: Place Cells and Grid Cells

Almost five decades ago, neuroscientists first discovered place cells in the hippocampus—nerve cells that become active when an individual moves through a specific location. Although these cells have subsequently been associated with storing episodic memories, how they encode events in a manner divorced of spatial context was not understood. In a new theory, an MIT research group, guided by Professor Ila Fiete, speculates that place cells, in concert with grid cells in the entorhinal cortex, contribute to a gridlike structure that enables encoding of episodic memories.

Grid cells are a specific group of neurons that discharge in a grid-shaped manner, producing a triad-shaped grid for an individual’s environment. In concert, grid and place cells form an effective mechanism for moving about in a location. Yet, according to the new model, such cells form a network that can serve to index non-space memories, allowing for the use of the brain’s system for representing spaces in forming memories about events.

A Computational Model for Episodic Memory Creation

To explore how memories for events become stored in the spatial memory system, the researchers developed a computational model of hippocampal function that involves grid cells and extends grid cell-based frameworks for representing spaces to include memories for events, namely, episodic memories. Under its most significant hypothesis, grid cells and interactions between grid cells and hippocampal cells generate a sequence of wells of neural activations that form a grid of locations for stored memories, but not with rich contents, but with indices for stored episodic information in synaptic contacts between hippocampus and sensory cortex.

When a memory is later recalled from partial cues, the hippocampal-entorhinal network directs neural activity toward the nearest well, which then activates the relevant sensory cortical regions to reconstruct the full memory. This mechanism explains how fragmented inputs can lead to coherent memory retrieval, a phenomenon well-documented in cognitive neuroscience.

Also, such a model can involve describing recall in terms of sequence. Each well in a network holds information for recalling a successive memory, and a causally linked sequence of episode memories is produced. Sequential recall mirrors naturally, in a continuous narrative, the manner in which humans recall experiences, and not in discrete snippets.

Overcoming Challenges of Traditional Model Shortfalls

Previous models, such as Hopfield networks, have failed to simulate neural encoding of memories but suffer a fundamental flaw: a finite information capacity, in that new memories can induce catastrophic forgetting of stored information when added in a continuous, ongoing manner. Biological memories, in contrast, decay over a period and not in an all-or-nothing, catastrophic manner. The MIT model succeeds in encoding this biological feature through its simulation of memories becoming weakened over time with new ones being added in a constant flux. The grid cell-hippocampal model naturally averts overloading memories through its distribution in a organized yet adaptable grid, such that older memories decay over a period of time and not in a single go.

Human Brain and Its Neural Basis

Perhaps most exciting about the study is its theory for why such techniques function with such effectiveness. For example, memory competition will apply such a technique in recalling massive amounts of information through projecting each detail onto a location in a mental or real environment. In such a new computational model, such a technique exploits a default mechanism in the brain for projecting memories onto structures in space. By projecting new information onto a stored map of a location, one can maximize recall efficiency and accuracy.

Moreover, such techniques for memory not only enhance recall through association but actually exploit the brain’s native encoding structure. Long-term memories for spaces in the sensory cortex build an additional structure, allowing retrieval of new information through use of native spatial-memory structure.

Implications and Future Directions

The results of this work reveal new avenues for investigating memory function in both living and artificial intelligence networks. By knowing about the processes involved in encoding in episodic memory, it will become easier to develop new therapies for memory impairments such as Alzheimer’s disease, in which hippocampal loss profoundly impairs recall in an episodic manner. The work sets a basis for AI representations that closely resemble human memory, specifically in machine learning with requirements for recall in an associative and sequential manner.

The future work of the group involves extending their model to investigate in detail how memories in an episode transition towards semantic memory—disembodied information unassociated with its first encounter. Besides, integration of such brain-inspired model of memory with AI-facilitated neural networks can make artificial machines for learning smarter and efficient in terms of adaptability.

Conclusion

The MIT study introduces a new model that brings together grid and episodic encoding in terms of encoding memories. By explaining how grid and place cells collaborate in creating a frame for storing memories, the study puts a strong case for both spatial and episodic memories employing similar neural circuits. Not only does the model shed new information about memories forming, but it paves the way for future neurological disease, AI memories, and cognitive enhancing techniques work. As neurosciences try to grasp the complexity of memories in humans, such observations represent a significant step towards knowing humans remember and navigate through life.

Dr. Prahlada N.B
MBBS (JJMMC), MS (PGIMER, Chandigarh). 
MBA in Healthcare & Hospital Management (BITS, Pilani), 
Postgraduate Certificate in Technology Leadership and Innovation (MIT, USA)
Executive Programme in Strategic Management (IIM, Lucknow)
Senior Management Programme in Healthcare Management (IIM, Kozhikode)
Advanced Certificate in AI for Digital Health and Imaging Program (IISc, Bengaluru). 

Senior Professor and former Head, 
Department of ENT-Head & Neck Surgery, Skull Base Surgery, Cochlear Implant Surgery. 
Basaveshwara Medical College & Hospital, Chitradurga, Karnataka, India. 

My Vision: I don’t want to be a genius.  I want to be a person with a bundle of experience. 

My Mission: Help others achieve their life’s objectives in my presence or absence!

My Values:  Creating value for others. 

Leave a reply