Some mathematical formulas in this article can’t be displayed in your browser.
I want to be careful here and stay on one thing: what Gustavo Deco and Morten L. Kringelbach actually accomplished in Whole-brain modelling: Cartography of the Dynamics of Mind. This is not a review of the field, and it is not an excuse to talk about my own systems. It is an attempt to explain, in plain but precise language, why their approach to modelling the brain is a genuine scientific achievement.
Let me start with the problem they set out to solve. The brain has roughly eighty-six billion neurons. You cannot model each one and expect to learn anything useful — the model would be as incomprehensible as the brain itself. For decades, neuroscience swung between two extremes: exquisitely detailed models of single neurons that explained nothing about cognition, and abstract cognitive theories that had no contact with biology. Deco and Kringelbach's central move is to find the right level of description in between.
The core idea: the brain as a dynamical system
Their foundational claim is that the brain is best understood not as a computer running algorithms, but as a dynamical system — a collection of interacting parts whose state evolves over time according to a few governing rules. More specifically, they treat the whole brain as a network of coupled oscillators.
Here is the recipe, stated simply:
- Divide the brain into regions. Using a brain atlas (a parcellation), the cortex and subcortical structures are split into a few hundred regions. Each region becomes a single node in the model.
- Wire the nodes together using real anatomy. The physical fibre tracts connecting these regions are measured in living humans using diffusion imaging (DTI). This gives a connectivity matrix — essentially a map of who talks to whom, and how strongly.
- Give each node its own local dynamics. Each region is modelled as a small oscillator that, left to itself, either sits quietly or rhythmically fluctuates. The transition between those two behaviours is governed by a single parameter — a bifurcation point.
- Couple them and turn one global knob. A single number, the global coupling , controls how strongly the regions influence each other through the anatomical wiring.
The local dynamics are often written with the Hopf model. For a single node, the equation reads:
Do not let the symbols intimidate you. is just the state of region . The parameter decides whether that region is quiet or oscillating — it is the bifurcation knob. is the region's natural rhythm. The middle term is the influence flowing in from all connected regions, scaled by the anatomy and the global coupling . The last term is biological noise. That is the whole model: simple local rules, real anatomical wiring, a little randomness.
Why this is a real achievement
The remarkable thing — and this is the heart of their work — is that when you tune the single global coupling to the right value, the simulated network spontaneously reproduces the large-scale activity patterns that we measure in real brains with fMRI. The functional connectivity, the resting-state networks, the slow fluctuations: they emerge from the model without being put in by hand.
That is the payoff. A handful of regions, real anatomy, simple oscillators, and one global parameter are enough to recreate the brain's spontaneous orchestration. It tells us that much of brain organisation is not a list of special-purpose circuits but an emergent property of a dynamical system poised near a critical point.
Deco and Kringelbach push this further with the idea of criticality. The model fits real data best when the system sits right at the edge between order and disorder — the bifurcation point. This is not a coincidence. A system at criticality maximises its sensitivity to inputs and its capacity to carry information. The brain, their models suggest, deliberately operates at this knife's edge.
Turbulence: information that cascades across scales
The most original part of the book is their import of a concept from fluid dynamics: turbulence. When you stir a fluid hard enough, energy injected at large scales cascades down to smaller and smaller scales in a structured, statistically predictable way. Kolmogorov described this cascade mathematically in the 1940s.
Deco and Kringelbach show that healthy brain activity has the same signature. Information is not broadcast all-to-all; it cascades across spatial and temporal scales the way energy cascades in a turbulent fluid. They built quantitative measures of this brain turbulence and showed that it is reduced in disease and altered states. This reframes a deep question — how does the brain move information efficiently across scales? — as a measurable physical property rather than a metaphor.
From description to intervention
What makes their cartography more than a pretty picture is that it is generative. Because the model is built from causal, mechanistic ingredients, you can perturb it and ask what happens. This is where the book becomes clinically meaningful.
They define brain states — sleep, wakefulness, anaesthesia, the psychedelic state, disorders of consciousness — as distinct regimes of the same dynamical system. A pathological state, in this view, is a brain trapped in the wrong region of its dynamical landscape. The therapeutic question then becomes concrete: what perturbation moves the system back into a healthy regime?
They go on to enrich the model with neurochemistry, layering receptor density maps (for example serotonin 5-HT2A) onto the structural connectome. This lets them simulate, in silico, what a drug like a psychedelic does to whole-brain dynamics — and the simulations match what is observed empirically. That is a striking validation: a model assembled from anatomy, simple oscillators and receptor maps can predict the global effect of a pharmacological intervention.
The engineering reality behind the science
I cannot read this work without noticing what it costs to compute. Each node is a stochastic differential equation; there are hundreds of them; and fitting the global parameters to a specific person's imaging data means running the simulation thousands of times to search the parameter space. This is genuinely heavy computation — the kind that lives on GPUs and runs for hours.
This matters for honesty about the science. The elegance of the equations hides a real numerical effort. Reproducing their results requires careful, optimised solvers, not naive loops. But this is a feature, not a bug: the model is cheap enough to be tractable on modern hardware yet rich enough to be predictive. That balance — between a model so simple it is empty and one so detailed it is uncomputable — is itself part of what they got right.
Beyond the black box
The deepest lesson of Whole-brain modelling is methodological. Deco and Kringelbach refuse to treat the brain as an inscrutable black box. They insist that the laws of statistical physics, thermodynamics and nonlinear dynamics apply to it, and they show that these laws are enough to build models that are both predictive and explanatory — models that tell you not just what the brain does, but why.
That is the achievement worth naming clearly. They have produced a working cartography of the dynamics of mind: a map in which states of consciousness are regions, transitions are trajectories, and disease is a system stuck in the wrong place. Whether or not every detail survives, the framework is the right kind of object — mechanistic, generative, and grounded in physics. They have drawn the map. The rest of us now have something real to navigate by.
