Switching States: Heteroclinic Cycles as Organising Centres of Neuronal Dynamics

TL;DR

This paper identifies a universal bifurcation structure, based on heteroclinic cycles, that organizes state transitions in neuronal networks across various models, providing a fundamental principle for understanding brain dynamics.

Contribution

It introduces necessary conditions for a novel bifurcation structure as a universal mechanism governing neuronal state transitions across multiple mean-field models.

Findings

The bifurcation structure emerges robustly across different models.

External input and connectivity interplay converges on this shared mechanism.

Mathematical structure of nonlinear input-output relationships preserves the organizing centre.

Abstract

Neuronal networks alternate between high- and low-activity regimes, known as up and down states. They also display rhythmic patterns essential for perception, memory consolidation, and sensory processing. Despite their importance, the principles behind such state transitions remain elusive. We propose necessary conditions for the existence of a novel bifurcation structure as a universal organising centre governing these transitions. Bifurcation analysis and simulations of canonical mean-field network models, including Wilson--Cowan, Tsodyks--Markram, and Jansen--Rit frameworks, show that this bifurcation structure emerges robustly across models. We demonstrate that the interplay between external input and (synaptic) connectivity converges onto this shared mechanism, providing a fundamental principle for understanding how diverse brain states arise and are regulated. Beyond phenomenological mean-field models, we show that a shared mathematical structure of nonlinear input-output relationships, rather than model-specific details, preserves the organising centre across frameworks, revealing a general mechanism for dynamic state transitions.