Allometric scaling of brain activity explained by avalanche criticality

Allometric scaling laws, such as Kleiber’s law for metabolic rate, highlight how efficiency emerges with size across living systems. The brain, with its characteristic sublinear scaling of activity, has long posed a puzzle: why do larger brains operate with disproportionately lower firing rates? Here we show that this economy of scale is a universal outcome of avalanche dynamics. We derive analytical scaling laws directly from avalanche statistics, establishing that any system governed by critical avalanches must exhibit sublinear activity-size relations. This theoretical prediction is then verified in integrate-and-fire neuronal networks at criticality and in classical self-organized criticality models, demonstrating that the effect is not model-specific but generic. The predicted exponents align with experimental observations across mammal species, bridging dynamical criticality with the allometry of brain metabolism. Our results reveal avalanche criticality as a fundamental mechanism underlying Kleiber-like scaling in the brain.

💡 Research Summary

The authors address a long‑standing puzzle in neuroscience: larger brains fire at disproportionately lower rates, a phenomenon reflected in the sublinear scaling of neuronal activity with brain size. While metabolic allometry such as Kleiber’s law explains how whole‑body energy consumption scales with mass, the dynamical origin of the brain’s own sublinear activity‑size relationship has remained unclear. In this work the authors propose that the economy of scale observed in brain activity is a generic consequence of avalanche dynamics at a critical point.

Starting from the finite‑size scaling form of avalanche size (S) and duration (D) distributions,
(P(X)=A_X X^{-\alpha_X} g(X/X_c)) with (X_c\propto N^{\beta_X}) (X = S, D), they derive analytical expressions for the average avalanche size ⟨S⟩ and duration ⟨D⟩ in the large‑N limit: ⟨X⟩∼N^{\beta_X(2‑α_X)} for 1 < α_X < 2. The mean number of active sites per time step, ⟨n_a⟩, can be written as the ratio ⟨S⟩/⟨D⟩ multiplied by the fraction of time the system spends in avalanches. When the quiet‑time between avalanches, ⟨τ⟩, approaches a constant for large systems, the quiet‑time factor becomes irrelevant and the scaling reduces to
(\langle n_a\rangle\propto N^{\beta_S(2‑α_S)-\beta_D(2‑α_D)}).
Because the exponent on the right‑hand side is generically smaller than one, any critical avalanche system necessarily exhibits sublinear activity‑size scaling (η < 1).

To test the theory, the authors first simulate a directed, scale‑free integrate‑and‑fire (IF) network with N ranging from 5 × 10³ to 10⁵ neurons. They explore both fully excitatory networks (0 % inhibition) and networks with 20 % inhibitory neurons, tuning a recovery parameter δ_u to keep each system at criticality. External drive is applied very slowly (δv = 0.1 per timestep) to satisfy the separation‑of‑timescales condition of SOC. The simulations reveal ⟨n_a⟩∝N^{η} with η ≈ 0.46 for the purely excitatory case and η ≈ 0.37 when inhibition is present, confirming sublinear scaling. Simultaneously, avalanche size and duration distributions collapse onto universal curves with exponents α_S≈3/2, α_D≈2 and cutoff exponents β_S≈β_D≈0.5, exactly as predicted by mean‑field branching‑process theory.

Next, the authors examine two classic sandpile models – the deterministic Bak‑Tang‑Wiesenfeld (BTW) model and the stochastic Manna model – implemented on square lattices with periodic boundaries. For system sizes up to N≈1.6 × 10⁷, both models display ⟨n_a⟩∝N^{η} with η ≈ 0.52 (BTW) and η ≈ 0.60 (Manna). The avalanche statistics again obey the expected finite‑size scaling forms; notably, in the Manna model α_D>2, causing ⟨D⟩ to saturate, which leads directly to the simplified exponent η = β_S(2‑α_S) derived in the analytical section. The BTW model, being at the mean‑field critical dimension, shows a logarithmic correction (⟨D⟩∝ln N) consistent with the theory.

Finally, the authors compare their predictions with empirical data from mammals. Indirect measurements of glucose utilization indicate that the average firing rate per neuron scales with gray‑matter volume as ρ∝V_G^{‑0.15}. Since gray‑matter volume itself scales with the number of cortical neurons as V_G∝N^{γ} (γ≈1.59 for rodents and γ≈0.92 for primates), the total firing rate scales as ⟨n_a⟩∝N^{1‑0.15γ}, yielding exponents ≈0.76 (rodents) and ≈0.86 (primates). These values are in close agreement with the η obtained from the IF simulations and the sandpile models, supporting the claim that avalanche criticality can account for the observed brain‑wide allometry.

In discussion, the authors emphasize that the sublinear scaling is not a consequence of metabolic constraints alone but emerges inevitably from the statistical structure of critical avalanches. Because the scaling law depends only on the universal exponents (α, β) of the avalanche distributions, it holds across a wide class of models and, by extension, across real cortical tissue that operates near criticality. The work therefore provides a unifying dynamical principle linking neuronal criticality, energy efficiency, and the allometric laws that govern brain evolution. It also suggests that deviations from critical avalanche behavior—such as those potentially occurring in neurological disorders—could manifest as altered scaling of metabolic demand with brain size, opening new avenues for both theoretical and experimental investigation.