Neural criticality · Grey Matter

Neural criticality is the hypothesis that cortical networks run near the boundary between activity that dies out and activity that explodes, a critical point, and that this position gives them the widest range of responses and the richest repertoire of patterns.

Neural criticality. Neural criticality is the hypothesis that cortical networks run near the boundary between activity that dies out and activity that explodes, a critical point, and that this position gives them the widest range of responses and the richest repertoire of patterns.

The main evidence is the neuronal avalanche. Recording spontaneous activity on grids of electrodes in slices of rat cortex, Beggs and Plenz (2003) saw bursts that spread across a few or many electrodes and then stopped. The number of avalanches of a given size fell as a power law with exponent about −3/2, and of a given duration with exponent about −2, the values predicted for a critical branching process, in which each active unit activates on average exactly one other (a branching ratio of 1). Below 1 activity fades quickly; above 1 it grows until something stops it.

P(s)∝s−3/2,P(T)∝T−2P(s) \propto s^{-3/2}, \qquad P(T) \propto T^{-2}P(s)∝s−3/2,P(T)∝T−2

The formula says there is no typical avalanche size: small ones are common, large ones rare, but every scale occurs, from a few neurons to most of the recorded field.

It has been seen in many preparations. Avalanche-like statistics appear in cultures, in awake animals and in human MEG and intracranial recordings, though the exponents vary and some power laws can arise without criticality.

It has computational appeal. Models and slice experiments show the widest dynamic range (the span of inputs a network can distinguish) and the greatest information capacity near the critical point.

It is a balance point. The excitation-inhibition balance sets the branching ratio, so drugs that weaken inhibition push networks to the supercritical side, with large, synchronous events, and stronger inhibition pushes them subcritical.

A critical cortex responds to everything without being overwhelmed by anything.

The same closeness to the edge that gives it range also explains how a small shift can tip it into runaway synchrony.

Questions: What is a neuronal avalanche, and what does its power law say about the cortex? An avalanche is a burst of activity that spreads across a group of recording sites and then stops, preceded and followed by quiet. In slices of rat cortex Beggs and Plenz found that avalanche sizes follow a power law with exponent about −3/2 and durations one with exponent about −2, the values expected for a branching process in which each active unit triggers on average exactly one other. That means activity is balanced at the edge between dying out and growing, which is the critical point. Since the balance of excitation and inhibition sets how many units each event recruits, weakening inhibition pushes avalanches toward large, runaway events and strengthening it makes them fizzle. If running near a critical point is useful, how does it relate to a seizure? Near the critical point a network has the widest dynamic range: in slice experiments by Shew and colleagues, cortical networks distinguished the widest span of input strengths when their avalanches followed the critical power law, and responded poorly when drugs pushed them to either side. Reducing inhibition made the networks supercritical, with large, synchronised events that dominate activity, a regime that resembles epileptic activity. The criticality hypothesis therefore links the brain's sensitivity and its vulnerability to the same position. It is a hypothesis: whether seizures are best described as a shift past a critical point is still debated, and other mechanisms (failure of specific interneurons, ion changes) can produce them.