Hierarchical organization of critical brain dynamics
Mouse brain activity reveals criticality gradients that vary with hierarchy and tasks.
A new study published on arXiv (2604.21832) by Cambrainha et al. bridges two foundational concepts in neuroscience: the hierarchical organization of brain structure and the hypothesis that brain dynamics operate near a critical point. Using phenomenological renormalization group approaches on large-scale neuronal spiking activity from mouse visual cortex and hippocampus, the team discovered that signatures of criticality are not uniform across brain regions. Instead, they vary systematically along the known anatomical hierarchy in both systems.
The most striking finding is that the direction of this gradient depends on the specific criticality exponent used. Exponents based on static properties, such as spatial correlations, point to a gradient in one direction along the hierarchy, while exponents based on dynamic properties, such as temporal correlations, point in the opposite direction. This reveals a nontrivial, measure-dependent organization of criticality. Furthermore, the signatures in the visual system are strongly modulated by engagement in a visual task. The correlations among criticality markers from different brain regions during active task performance were sufficient to reconstruct the anatomical hierarchy from the dynamics alone. The scaling exponents closely followed a theoretically predicted scaling relation and covaried with hierarchical position, providing a direct link between neuronal collective dynamics and macroscopic brain architecture.
- Criticality signatures vary systematically along the anatomical hierarchy in mouse visual cortex and hippocampus.
- Static and dynamic criticality exponents point in opposite directions along the hierarchy, revealing measure-dependent organization.
- Task engagement modulates criticality signatures, and correlations among markers can reconstruct the hierarchy from dynamics.
Why It Matters
This work links brain structure to dynamics, potentially informing AI and neural network design.