Complex Brain Hypothesis Overview

• Complex Brain Hypothesis is a framework positing that brain functions emerge from collective dynamics and intricate balances between order and disorder.
• Methodologies span measuring scale-free neural avalanches, phase transitions, and transient synchronizations using quantitative metrics like power laws and order parameters.
• Diverse formulations of CBH link neurodynamics, cognitive evolution, computational modeling, and consciousness, underscoring a unified yet pluralistic view of brain complexity.

The Complex Brain Hypothesis (CBH) designates a family of research programs that treat the brain as a complex, self-organizing system whose distinctive capacities arise from collective dynamics rather than from isolated components. In its best-known neurodynamical form, CBH holds that the brain operates at or very near the critical point of an order–disorder phase transition, thereby exhibiting scale-free fluctuations, divergent correlation length, and maximal susceptibility. Other formulations preserve the emphasis on complexity while shifting the explanatory center to metastability, higher-order interactions, grammatical capacity, evolutionary trade-offs, or inferential complexity in consciousness. The result is not a single doctrine but a heterogeneous theoretical landscape united by the claim that brain function depends on structured balances—order and disorder, integration and segregation, redundancy and efficiency, or entropy and complexity (Tagliazucchi et al., 2011, Tagliazucchi et al., 2012, Kelso, 2023, Mago et al., 15 May 2026).

1. Conceptual scope and terminological plurality

The expression “Complex Brain Hypothesis” is used in several distinct literatures. In statistical and systems neuroscience, it most often refers to the proposal that the brain is poised near criticality. In cognitive-evolutionary work, it can refer to claims that enlarged brains enabled recursive recall and contextual focus. In computational neurobiology, it can denote formal models in which neural systems solve implicit parametric equations or implement grammar-like computations. In recent consciousness research, it has been reformulated to mean that phenomenal richness is better indexed by complexity than by entropy alone. A further source of ambiguity is acronymal: some papers use “CBH” for the “Cognitive Buffer Hypothesis,” not for the “Complex Brain Hypothesis” (Tagliazucchi et al., 2011, Gabora et al., 2013, Schad, 2019, Heesom-Green et al., 25 Nov 2025, Mago et al., 15 May 2026).

Usage	Core proposition	Representative paper
Critical-dynamical CBH	Brain operates near a second-order critical point	(Tagliazucchi et al., 2011)
Connectome–cognitome CBH	Dynamical complexity emerges near criticality	(Tagliazucchi et al., 2012)
Metastable reformulation	Brain occupies a “sea of metastability”	(Kelso, 2023)
Cognitive-evolutionary CBH	Brain enlargement enabled recursive recall and contextual focus	(Gabora et al., 2013)
Consciousness-oriented CBH	Richness of experience tracks complexity, not entropy alone	(Mago et al., 15 May 2026)
Acronymal alternative	“CBH” = Cognitive Buffer Hypothesis	(Heesom-Green et al., 25 Nov 2025)

This plurality matters because disagreements in the literature often concern not only evidence but also the target explanandum. Some versions seek a general principle of brain dynamics; others address hominin cognitive evolution, conscious-state taxonomy, or computational expressivity. A recurring misconception is therefore that “CBH” names a single, universally accepted framework. The published record instead shows a cluster of partially overlapping hypotheses with different mathematical objects, empirical targets, and standards of confirmation (Tagliazucchi et al., 2012, Tognoli et al., 2013).

2. Criticality as the canonical neurodynamical formulation

The canonical formulation begins from the claim that the brain, as “a very large network of nonlinear excitable units,” operates at or near the critical point of an order–disorder phase transition. At that point three hallmark features appear simultaneously: scale-free fluctuations, divergent correlation length, and maximal susceptibility. In this view, those features reconcile two demands of adaptive behavior that would otherwise seem incompatible: stable large-scale coordination and flexible switching among many patterns (Tagliazucchi et al., 2011).

The theoretical apparatus is drawn directly from critical phenomena. Event sizes obey a power law,

$P(s)\propto s^{-\tau},$

where in neural terms $s$ can be the number of neurons involved in an avalanche of synchronous firing. Correlation length obeys

$\xi\propto |t-t_c|^{-\nu},$

with $t$ a control parameter such as synaptic gain. Scaling relations link avalanche sizes and durations; in branching-process language, criticality corresponds to a branching ratio of $1$, so that on average one spike produces exactly one offspring spike (Tagliazucchi et al., 2011).

A large empirical program is organized around this framework. At small scales, Beggs and Plenz reported neuronal avalanches in organotypic cortical slices with

$P(s)\propto s^{-3/2},$

together with finite-size scaling, separation of timescales, stationarity of the size distribution, Omori-like laws for foreshocks and aftershocks, and fractal spatial spread. At mesoscales, EEG and MEG spectral densities scale like $1/f^\beta$ with $\beta\approx 1$, and detrended-fluctuation analysis yields exponents $>0.5$, indicating long-range temporal correlations. At large scales, resting-state fMRI functional connectivity graphs obtained by thresholding pairwise BOLD correlations are scale free with $P(k)\sim k^{-2}$, small-world, and assortative; spatial coarse-graining leaves the normalized two-point correlation function invariant, and empirical correlation length grows with cluster size (Tagliazucchi et al., 2011).

Tagliazucchi and Chialvo sharpened this program by introducing explicit order and control parameters for a brain-wide second-order transition. Their control parameter is instantaneous global activity,

$s$0

and the order parameter is the normalized size of the largest co-activation cluster,

$s$1

The signatures of criticality are then expressed as $s$2 for $s$3, a susceptibility

$s$4

that peaks at $s$5, and a correlation length $s$6 extracted from two-point correlations of the point process. This formulation explicitly links microscopic avalanches and macroscopic resting-state networks, and it distinguishes structural complexity of the connectome from dynamical complexity of the “cognitome” (Tagliazucchi et al., 2012).

The functional significance claimed for criticality is correspondingly broad. Dynamic range is maximal at criticality in networks of excitable elements; subcritical networks damp weak stimuli, while supercritical networks saturate. Mutual information between network parts is maximized, susceptibility diverges, and divergent correlation length permits transient synchronization of distant regions without imposing a single global attractor. Trial-to-trial variability is therefore interpreted not as mere noise but as a signature of large-amplitude fluctuations in a system poised at the edge of an order–disorder transition (Tagliazucchi et al., 2011).

3. Metastability and coordination dynamics as an alternative emphasis

A major revision of the criticality-centered CBH replaces tuning to a single critical point with metastability. In Coordination Dynamics, brain complexity is defined by the continual interplay between integration and segregation. It is neither maximal synchrony nor complete independence that matters, but an intermediate regime in which neural subsystems couple and decouple over time, forming and dissolving transient coalitions. Tognoli and Kelso present metastability as the regime that reconciles information flow with synchronization and mass action (Tognoli et al., 2013).

The standard mathematical entry point is the Haken–Kelso–Bunz family of phase equations. For two coupled oscillators, the extended relative-phase dynamics can be written

$s$7

Fixed points satisfy

$s$8

and saddle-node bifurcations eliminate stable and unstable solutions in pairs. Beyond those bifurcations, no true attractors remain, yet trajectories can continue to hover near the “ghosts” of former equilibria. This is the mathematical core of Kelso’s claim that the brain–mind lives in a broad “sea of metastability” rather than at a uniquely privileged point separating order and randomness (Kelso, 2023).

This reformulation does not deny the importance of integration–segregation balance; it changes the mechanism by which that balance is realized. In metastability, relative-phase tendencies such as in-phase and anti-phase coordination coexist with drift and noise-induced switching. For larger populations one may write a Kuramoto-style system,

$s$9

and characterize collective behavior with the order parameter

$\xi\propto |t-t_c|^{-\nu},$0

The metastability index is then the temporal variance of $\xi\propto |t-t_c|^{-\nu},$1, large when coherence waxes and wanes rather than remaining fixed (Tognoli et al., 2013).

Kelso contrasts the expected statistics of metastable dynamics with those of criticality. Near ghost states, dwell-time distributions are typically exponential or stretched exponential,

$\xi\propto |t-t_c|^{-\nu},$2

whereas genuine criticality would predict power-law dwell times. Likewise, order-parameter fluctuations in metastability are moderate and non-divergent, and there is no requirement for scale-free avalanche distributions. The empirical claim is that cortical modules in EEG, MEG, and behavior repeatedly form and dissolve transient phase-locked coalitions on timescales of tens to hundreds of milliseconds, more consistent with metastable coordination than with fine-tuning to a single critical point (Kelso, 2023).

A plausible implication is that criticality and metastability need not be interpreted as mutually exclusive at all scales. The literature surveyed here shows two different ways of formalizing the same desideratum: large repertoires of coordinated yet reconfigurable states. One emphasizes second-order phase transitions and universal scaling; the other emphasizes broad parameter regions populated by coexisting tendencies and transient coordination (Tagliazucchi et al., 2012, Kelso, 2023).

4. Evolutionary and phylogenetic versions of CBH

In cognitive-evolutionary work, CBH has been used to explain how anatomical and neurological changes produced qualitatively new cognitive capacities. Gabora and Kitto distinguish two transitions. The first, associated with the large brain of Homo erectus, is “recursive recall,” or self-triggered recall and rehearsal, enabling thoughts and actions to be chained into streams. The second, associated with anatomically modern humans, is “contextual focus,” the capacity to shift between a minimally contextual analytic mode and a highly contextual associative mode. Their computational support comes from EVOC simulations, in which chaining raises mean fitness without plateauing and sustains higher cultural diversity, and from an evolutionary art system in which contextual focus yields portraits judged more creative and appealing than those from a fixed-fitness baseline (Gabora et al., 2013).

The same paper embeds these claims in a formal theory of concepts inspired by quantum formalisms. In context $\xi\propto |t-t_c|^{-\nu},$3, a concept $\xi\propto |t-t_c|^{-\nu},$4 is represented as

$\xi\propto |t-t_c|^{-\nu},$5

and different contexts induce different decompositions. Concept combination is represented by tensor products, allowing emergent interpretations of compounds such as FIRE $\xi\propto |t-t_c|^{-\nu},$6 FOOD. Within this framework, larger brains are taken to permit richer encoding of experiential details, while contextual focus changes the effective “width” of cognitive activation, alternately expanding and contracting the accessible combination space (Gabora et al., 2013).

Fialkowski offers a sharply different evolutionary account. Building on introgression of an archaic MCPH1 allele about $\xi\propto |t-t_c|^{-\nu},$7 years ago to roughly $\xi\propto |t-t_c|^{-\nu},$8 worldwide frequency, he argues that a variant associated with slightly less complex brain structure could have been strongly favored. In the structured account provided, neuroblast proliferation is modeled by

$\xi\propto |t-t_c|^{-\nu},$9

and adult brain volume is linked to progenitor dynamics through relations such as $t$0 and $t$1. The allele-frequency trajectory is represented by a logistic model with an inferred selection coefficient on the order of $t$2–$t$3 per generation. Fialkowski’s interpretation is that noncognitive selection pressure—specifically thermoregulatory reliability under endurance running—can explain much of phylogenetic brain enlargement, and that increasing complexity is not unconditionally adaptive for effective thinking (Fialkowski, 2013).

A related but acronymally distinct literature concerns the “Cognitive Buffer Hypothesis,” also abbreviated CBH. In an in silico neuro-evolution study, Artificial Neural Networks controlled RL foraging agents in environments with varying seasonality and with or without size-dependent energy costs. Under energy constraints, increasing seasonality produced smaller ANNs and lower structural complexity; network size differed across environments with Kruskal–Wallis $t$4, and the correlation between season count and ANN size was negative ($t$5, $t$6). The authors interpret this as support for the Expensive Brain Hypothesis rather than the Cognitive Buffer Hypothesis under realistic energy costs (Heesom-Green et al., 25 Nov 2025).

Complexity-driven symmetry breaking has also been formalized in models of lateralization. In a bilateral-vs-lateralized framework with reliability $t$7, metabolic cost $t$8, coordination cost $t$9, and task complexity $1$0, the utilities of pure configurations are

$1$1

The boundary

$1$2

maps when each solution is preferred. Within this framework, only fully lateralized or bilateral solutions are relevant; partially engaged solutions occur only on measure-zero boundaries. As $1$3 increases, lateralization can become the only viable strategy in parts of parameter space, though the converse transition can also occur when redundancy becomes advantageous (Seoane, 2021).

5. Formal computational and systems-theoretic realizations

Several papers translate CBH into explicit computational architectures. One influential proposal links mammalian brain size to formal language classes. Rodriguez and Granger argue that thalamocortical loops compute formal grammars; successive cortical regions describe grammar rewrite rules of increasing size; cortical–subcortical ratios determine the quantity of stacks in pushdown-like computations; and quantitative increase in stack depth yields qualitatively increased computational power. Their specific conjecture is that human brain capacity is equivalent to indexed grammars, “far short of full Turing-computable (recursively enumerable) systems” (Rodriguez et al., 2016).

The formal hierarchy is standard. Regular grammars correspond to finite-state automata; context-free grammars correspond to single-stack pushdown automata; indexed grammars permit nonterminals carrying stacks of indices, with push, pop, and unchanged-stack rewrites. The biological conjecture is that increasing cortical–subcortical volume ratios permit deeper nested calls to a hippocampal stack-like mechanism, thereby moving species across computational thresholds without introducing wholly new circuitry. This is offered as an explanation for how quantitative allometric change could produce qualitative differences in syntax and other hierarchical capacities (Rodriguez et al., 2016).

A different computational realization is given by the proposal that the brain solves “life’s complex, context-dependent equations” as implicit parametric systems. In this framework, neural activity satisfies

$1$4

with components

$1$5

Learning updates parameters by

$1$6

while activations evolve as

$1$7

Consciousness is then linked to a thresholding condition: once a sufficiently low-residual solution is reached, a broadcast-type mechanism elevates it into awareness (Schad, 2019).

A broader systems-neuroscience version treats the brain as a complex adaptive system. The formal statement given in that literature combines a weighted graph $1$8, node dynamics such as

$1$9

and an order parameter

$P(s)\propto s^{-3/2},$0

with the claim that healthy brains operate near adaptive critical points. This program emphasizes multiscale data fusion, neural-avalanche analysis, adaptive-network models, and techniques such as Generalized Phase Locking Analysis, spike–LFP coupling matrices, and surrogate-based statistical tests to bridge microscopic and macroscopic descriptions (Safavi, 2023).

Recent higher-order approaches extend CBH beyond pairwise graphs. In the combinatorial-complex framework, random variables $P(s)\propto s^{-3/2},$1 are summarized by total correlation $P(s)\propto s^{-3/2},$2 and dual total correlation $P(s)\propto s^{-3/2},$3, from which one defines

$P(s)\propto s^{-3/2},$4

Here $P(s)\propto s^{-3/2},$5 measures total multivariate dependence and $P(s)\propto s^{-3/2},$6 indicates net synergy. Cells of a combinatorial complex are then selected from subsets of brain regions when pairwise or higher-order information-theoretic criteria are met. The central claim is that graph representations systematically miss synergistic interactions involving three or more regions, whereas combinatorial complexes provide a data-driven scaffold for topological deep learning on neural data (Sánchez et al., 22 Nov 2025).

6. Consciousness, entropy, and unresolved controversies

A recent and conceptually distinct use of CBH concerns the relation between neural entropy, neural complexity, and phenomenal richness. In this formulation, introduced to address Minimal Phenomenal Experiences (MPEs), the key claim is that richness of consciousness is not indexed by entropy alone but by the complexity of the brain’s generative model. Complexity is defined as the relative entropy between posterior and prior beliefs, technically a Kullback–Leibler divergence:

$P(s)\propto s^{-3/2},$7

Within the free-energy decomposition,

$P(s)\propto s^{-3/2},$8

complexity quantifies how many effective degrees of freedom are recruited to explain data (Mago et al., 15 May 2026).

This proposal is designed to resolve the “entropy–content conundrum.” Both high-content psychedelic experiences and minimal-content meditative or pharmacological states can show elevated neural entropy. Entropy itself is defined in the usual way,

$P(s)\propto s^{-3/2},$9

with spectral entropy obtained from normalized power spectra. The CBH response is to distinguish two regimes of liberated inference. In a fine-grained, overfitting regime, loosened constraints expand the degrees of freedom of the model, producing high entropy and high complexity, characteristic of high-content psychedelic experience. In a coarse-grained, underfitting regime, a simpler model dissolves variety into content-minimal awareness, producing high entropy but low complexity, characteristic of some MPEs (Mago et al., 15 May 2026).

The empirical program proposed for this version of CBH is correspondingly differential. Beyond Lempel–Ziv complexity, it recommends Block Decomposition Method, causal large-scale models, integrated information $1/f^\beta$0, and perturbational measures such as PCI. Its explicit prediction is that MPEs and high-content psychedelic states may share elevated global entropy while diverging in perturbational signatures: MPEs should exhibit high entropy with low $1/f^\beta$1 and low response diversity to perturbation, whereas high-content psychedelic experiences should exhibit high entropy, high $1/f^\beta$2, and high PCI (Mago et al., 15 May 2026).

Across the broader CBH literature, three controversies recur. The first concerns dynamical regime: criticality emphasizes singular transitions and universal scaling, whereas metastability emphasizes broad parameter regions and transient coalitions (Tagliazucchi et al., 2012, Kelso, 2023). The second concerns what “complexity” means: structural network organization, dynamical repertoire, grammatical capacity, multivariate synergy, and inferential complexity are not interchangeable constructs (Tagliazucchi et al., 2012, Rodriguez et al., 2016, Sánchez et al., 22 Nov 2025, Mago et al., 15 May 2026). The third concerns adaptiveness: some papers present increasing complexity as functionally advantageous, while others argue that energy costs, thermal reliability, or shifted selective regimes can favor simplification, lateralization, or more efficient architectures instead of maximal complexity (Fialkowski, 2013, Seoane, 2021, Heesom-Green et al., 25 Nov 2025).

Taken together, these works support a restrained but durable interpretation of CBH. The shared thesis is that brain function cannot be reduced to local mechanisms alone; it depends on collective organization across scales. What remains unsettled is the preferred formalism for that organization—critical exponents, metastable order parameters, grammar hierarchies, higher-order topologies, or variational complexity—and the biological regimes in which each formalism is explanatory.