Neuronal Avalanches¶

Neuronal avalanches are cascading bursts of neural activity whose sizes follow a power-law distribution -- the signature of a system operating at or near the critical point.

In 2003, John Beggs and Dietmar Plenz discovered something unexpected in slices of rat cortex. When they recorded the spontaneous activity of neural networks, the cascading bursts of firing did not follow a normal distribution (most events medium-sized, extremes rare). Instead, small cascades were overwhelmingly common, medium cascades less so, and large cascades rare -- but not impossibly rare. The sizes followed a power law, the statistical fingerprint of a system poised at criticality.

The Discovery¶

Beggs and Plenz recorded local field potentials from organotypic cortical cultures using multielectrode arrays. They defined an avalanche as a sequence of activations that propagates across electrodes without temporal gaps: activity at one electrode triggers activity at neighbors, which triggers their neighbors, and so on until the cascade dies out.

The critical observation was the distribution of avalanche sizes. If the cortex operated in an ordered regime, cascades would die out quickly -- almost all avalanches would be tiny. If it operated in a chaotic regime, every perturbation would spread everywhere -- avalanches would be uniformly large. What they found was neither: a power-law distribution with exponent approximately -3/2, meaning that an avalanche twice as large was roughly 2.8 times less likely. This specific exponent is the theoretical prediction for a system at the critical point of a branching process.

Power Laws: The Signature of Criticality¶

A power law describes a relationship where one quantity varies as a power of another: P(s) ~ s^(-α), where s is avalanche size, P(s) is the probability of observing that size, and α is the scaling exponent. On a log-log plot, power laws appear as straight lines -- and that linearity spans orders of magnitude.

Why does this matter? Normal distributions have a characteristic scale (the mean). Power laws do not. A system producing power-law-distributed events has no preferred scale -- the same statistical pattern governs events from the smallest to the largest. This scale-free behavior is the mathematical consequence of operating at a phase transition, where correlations extend across the entire system. A power law on its own does not prove criticality, though. Non-critical processes can produce apparent power-law distributions (Touboul & Destexhe, 2017), and avalanche power laws, their scaling relations and long-range temporal correlations can all arise across an extended phase of long-range order with no critical point (Sipling, Zhang & Di Ventra, 2026). That is why the Four-Model Theory looks for the convergence of several signatures rather than any single one.

From Slices to Living Brains¶

The original finding was in cortical slices -- a preparation far removed from the intact brain. Subsequent work extended the result dramatically. Power-law neuronal avalanches have been observed in:

Awake behaving monkeys (Petermann et al., 2009) -- confirming that avalanche dynamics are not an artifact of isolated tissue.
Human MEG recordings (Shriki et al., 2013) -- demonstrating avalanche dynamics at the whole-brain scale.
Developing cortex (Gireesh & Bhatt, 2008) -- showing that criticality emerges during neural development.
Multiple species across mammals and birds -- suggesting that criticality is a conserved organizational principle, not a quirk of one species' cortex.

Avalanche dynamics also shift predictably with brain state. Under propofol anesthesia, the dynamics are pushed subcritical. (Not every anesthetic abolishes the power law: avalanches persist in anesthetized cat visual cortex under a non-propofol agent, Hahn et al., 2010.) During sleep, avalanche statistics change in ways consistent with periodic departure from and return to criticality. These state-dependent changes indicate that the power law is not a static property of neural tissue but a dynamic signature of the brain's operating regime.

Figure¶

graph LR
    subgraph "Neuronal Avalanche"
        direction TB
        N1["Neuron\nfires"] --> N2["2 neighbors\nactivated"]
        N2 --> N3["4 neurons\nactivated"]
        N3 --> N4["Cascade\ndies out"]
    end

    subgraph "Size Distribution"
        direction TB
        S["Power Law\nP(s) ~ s⁻³ᐟ²"]
        SM["Many small\navalanches"]
        ML["Few medium\navalanches"]
        LG["Rare large\navalanches"]
        SM --> ML --> LG
    end

    style N1 fill:#4CAF50,color:#fff,stroke:#333
    style N2 fill:#66BB6A,color:#fff,stroke:#333
    style N3 fill:#81C784,color:#000,stroke:#333
    style N4 fill:#A5D6A7,color:#000,stroke:#333
    style S fill:#FFF3E0,stroke:#E65100,color:#000
    style SM fill:#E3F2FD,stroke:#1565C0,color:#000
    style ML fill:#E3F2FD,stroke:#1565C0,color:#000
    style LG fill:#E3F2FD,stroke:#1565C0,color:#000

A neuronal avalanche begins when one neuron fires and triggers a cascade through its neighbors. The sizes of these cascades follow a power law -- many small, few large, with no characteristic scale -- the hallmark of criticality.

Key Takeaway¶

Neuronal avalanches are among the most direct empirical evidence that the brain operates near criticality. Their power-law size distribution -- observed across species, preparations, and recording methods -- is a signature of near-critical dynamics, though not proof of a critical point on its own. For the Four-Model Theory it is one of three convergent biological signatures of the Class 4 regime that consciousness requires.