Hippocampus encoding memory engrams as stable heteroclinic network.

This article focuses on modeling memory encoding and storage in the hippocampal CA3–CA1 synaptic network using stable heteroclinic channels (SHCs) and stable heteroclinic networks (SHNs) in phase space. Memory engrams are represented as trajectories visiting saddle points, which correspond to segmented information blocks; short-term memory capacity is characterized by the number of saddle points accessed within an SHC, while long-term memory capacity is greatly expanded by combining multiple SHCs into an SHN. The model highlights that asymmetric mutual inhibition between adjacent postsynaptic neurons—modulated by neurotransmitters such as acetylcholine—determines short-term memory capacity and participates in long-term memory encoding. Numerical simulations demonstrate that noise in the hippocampal circuit can both disrupt and facilitate memory encoding by influencing trajectory transitions among saddle points. This dynamical systems approach provides a theoretical framework for understanding the spatiotemporal encoding and consolidation of memory without relying on attractor states.