Engram Memory Encoding and Retrieval: A Neurocomputational Perspective

TL;DR

The paper surveys how memories are encoded, stored, and retrieved through engram ensembles, drawing from cellular neuroscience and computational modeling. It argues that memory efficiency, capacity, and stability emerge from the interaction of temporally sensitive plasticity and multi-level sparsity, supported by frameworks such as Hebbian learning, STDP, Sparse Distributed Memory, and engram gating. Its key contributions include a synthesis of biological findings with computational principles, highlighting how sparse high-dimensional representations and dynamic pruning can yield robust, interference-resistant memory traces. The work offers a roadmap for integrating neurobiological detail with tractable models to advance memory theory and inform therapies for memory-related disorders.

Abstract

Despite substantial research into the biological basis of memory, the precise mechanisms by which experiences are encoded, stored, and retrieved in the brain remain incompletely understood. A growing body of evidence supports the engram theory, which posits that sparse populations of neurons undergo lasting physical and biochemical changes to support long-term memory. Yet, a comprehensive computational framework that integrates biological findings with mechanistic models remains elusive. This work synthesizes insights from cellular neuroscience and computational modeling to address key challenges in engram research: how engram neurons are identified and manipulated; how synaptic plasticity mechanisms contribute to stable memory traces; and how sparsity promotes efficient, interference-resistant representations. Relevant computational approaches -- such as sparse regularization, engram gating, and biologically inspired architectures like Sparse Distributed Memory and spiking neural networks -- are also examined. Together, these findings suggest that memory efficiency, capacity, and stability emerge from the interaction of plasticity and sparsity constraints. By integrating neurobiological and computational perspectives, this paper provides a comprehensive theoretical foundation for engram research and proposes a roadmap for future inquiry into the mechanisms underlying memory, with implications for the diagnosis and treatment of memory-related disorders.

Figures (7)

• Figure 1: Lifecycle of an engram: An initial experience activates a sparse set of neurons, which undergo synaptic modifications during consolidation, forming a stabilized engram that can later be reactivated by partial cues. Excitatory (solid, blue) and inhibitory (dashed, red) connections guide engram refinement and stability.
• Figure 2: Structural plasticity during engram consolidation. Initially, many neurons are weakly connected after encoding. During refinement, unused neurons are pruned and relevant connections are strengthened. The final engram is sparse and robust, with strong internal connectivity supporting memory recall.
• Figure 3: Illustration of Hebbian plasticity and Spike-Timing-Dependent Plasticity. Left: Coactivation of pre- and postsynaptic neurons strengthens the synapse (Hebb's Law; i.e., "fire together, wire together"). Center: STDP curve showing how the relative timing of spikes controls potentiation (LTP) or depression (LTD). Right: Calcium concentration-based thresholds underlying synaptic change --- moderate calcium leads to LTD; high calcium leads to LTP.
• Figure 4: Comparison of dense (left) and sparse (right) memory coding. In dense coding, memories activate many overlapping neurons, increasing metabolic cost and interference. Sparse coding activates small, distinct neuronal subsets, improving efficiency and separability.
• Figure 5: Illustration of engram gating in a metaplastic binarized neural network. A task-specific encoder projects a gating vector that selects a sparse ensemble of neurons (a stochastic “engram”). Only this subset participates in forward pass and learning, preventing interference with other tasks and mitigating catastrophic forgetting.