Brain criticality debate

Lens adapted to this topic: Evidence for critical or near-critical brain dynamics

This perspective holds that neural systems often operate near a critical transition, where activity can support flexible information processing. It rests on observations of neuronal avalanches and scale-invariant or long-range correlations, theoretical links between near-criticality and computation, and proposals that homeostatic mechanisms maintain an optimal setpoint. Supporters acknowledge objections but argue that several can be addressed by appropriate modeling and analysis.

• Neural networks may operate near a critical phase transition, a regime associated in theory and experiments with efficient information processing.
• Neuronal avalanches and related scale-invariant activity patterns motivate interpreting cortical dynamics as critical or near-critical.
• Homeostatic regulation may maintain criticality as a functional setpoint, and departures from it have been associated with disorders, anesthesia, or impaired states.
• Some objections can be answered by refining models: for example, mutual-information benefits may appear near transition points when external random drive is not too strong.
• Recent experimental work reports that critical-dynamics measures predict cognitive performance and are disrupted by epileptic spikes, medication, and slow-wave activity.

Lens adapted to this topic: Alternative explanations for critical-like neural activity

This perspective accepts that brains can display critical-like signatures but disputes that they demonstrate operation at a critical point or that criticality is required for cognition. It emphasizes that power laws and other criteria can be generated by simple models, that cortical dynamics may be reverberating and subcritical, and that memory or self-regulation can reproduce the observed patterns without a critical computational setpoint.

• Scale-invariant correlations and other criticality signatures may be produced by memory, so their presence does not establish that the brain is at a critical point.
• Cortical dynamics may be better described as reverberating and subcritical, while criticality may characterize some in-vitro neuronal networks rather than the whole brain.
• Simple systems such as coin-flip models can satisfy multiple criteria used to diagnose criticality, raising concerns about false positives.
• Power-law avalanche distributions and related observations have alternative interpretations, so they do not by themselves establish a phase transition.
• Self-regulation toward stable activity could explain proximity to criticality; computational benefits of near-criticality may be side effects rather than evolutionary drivers.