The Thermodynamics of Falling Apart

A candle flame is not a thing. It is a process.

This sounds like a semantic trick, but it is one of the most important ideas in modern physics. A candle flame has a shape, a color, a temperature. You can point to it. You can warm your hands by it. But it is not made of any fixed material — the molecules passing through it at any given moment are different from the ones that were there a moment ago. What persists is not stuff but organization: a pattern of energy flow that maintains itself by continuously consuming fuel and exporting waste heat into the surrounding air.

Stop the fuel, and the pattern collapses. Not because the flame was damaged. Because the process that sustained it was interrupted.

Ilya Prigogine won the Nobel Prize for formalizing this. He called these patterns dissipative structures — organized systems that exist far from equilibrium, maintaining their improbable order by burning through energy and exporting entropy to their environment. A candle flame is one. A hurricane is one. A cell is one.

A human being is one.

And the thing about dissipative structures is that they don’t fail gradually. They hold, and they hold, and they hold — and then they don’t. The physics has a name for the moment of collapse: a bifurcation point. A threshold beyond which the current pattern of organization can no longer sustain itself, and the system must either find a new stable configuration or fall apart into disorder.

If you have ever watched someone — or been someone — who held it together under impossible stress for months or years and then suddenly, catastrophically, couldn’t, you have seen a bifurcation point. The last straw was never about the last straw. It was about the thermodynamics of a dissipative structure reaching the edge of what it could sustain.

This is where the story starts.

The Conversation Inside Your Body

Your body runs on coordination. Not on any single signal doing the right thing, but on many signals doing the right thing in relation to each other.

Cortisol rises in the morning and falls at night. But that rhythm doesn’t exist in isolation — it’s coupled to your sleep architecture, which is coupled to your autonomic nervous system, which is coupled to your inflammatory signaling, which is coupled to your metabolic regulation. Each of these systems has its own dynamics, its own rhythms, its own logic. But they are not independent. They talk to each other. They constrain each other. They form a web of mutual influence that keeps the whole enterprise coherent.

The Evolutionary Stress Framework calls this coherence. And coherence is not free. It is thermodynamically expensive. Maintaining all those couplings — keeping cortisol and inflammation in conversation, keeping autonomic tone and sleep architecture in sync — requires energy, intact signaling pathways, and a history of developmental calibration that taught the system how to do this in the first place.

Coherence is your body’s candle flame. It is the dissipative structure you are made of.

When the Music Stops

Here is where information theory enters the story, and it enters with a specific gift: it lets us measure what coherence is and what it means when it’s lost.

Claude Shannon, working at Bell Labs in 1948, defined a quantity called entropy — not the thermodynamic kind (though they turn out to be related), but informational entropy. It measures uncertainty. If you know exactly what state a system is in, entropy is zero. If the system could be in any of a thousand states and you have no idea which one, entropy is high.

The formula is elegant:

H(X) = −Σ p(xᵢ) log p(xᵢ)

For a regulatory system — your body’s web of cortisol, autonomic tone, immune signaling, and the rest — X is the set of possible configurations. When regulation is working, the system is constrained to a narrow range of those configurations. It isn’t visiting random states; it’s organized. Entropy is low.

But Shannon gave us something even more useful than entropy: mutual information. This is the measure of how much knowing one thing tells you about another:

I(X; Y) = H(X) + H(Y) − H(X, Y)

When your cortisol dynamics and your inflammatory signaling are coherent — when they’re in that coordinated conversation — knowing the state of one tells you something reliable about the state of the other. Mutual information is high. The joint entropy of the system (the total uncertainty across both channels together) is lower than what you’d get if they were operating independently.

That reduction in joint entropy is coherence, measured.

And stress incoherence — the central concept of the ESF — is what happens when that mutual information declines. The channels stop talking to each other. Cortisol does its thing; inflammation does its thing; but the coupling between them weakens. Each channel’s individual entropy might not change much. But the joint entropy — the total uncertainty of the system as a whole — rises. The system starts visiting states it shouldn’t be in. Configurations that a coherent system would never produce start showing up.

This is not metaphorical disorder. This is disorder in the most precise, mathematical sense information theory can offer: the system’s entropy is increasing because its internal coordination is degrading.

The Envelope

The ESF calls the boundary within which coherence can be maintained the allostatic envelope. In Prigogine’s terms, this is the regime within which the dissipative structure holds — where the organism successfully exports enough entropy to keep its internal organization intact.

Inside the envelope, you can take hits. Cortisol spikes from a bad meeting and comes back down. You catch a cold and your immune system activates and resolves. You lose a night of sleep and your recovery systems compensate. The mutual information between channels dips and recovers. The flame flickers but holds its shape.

At the edges of the envelope, recovery starts to falter. Cortisol stays elevated a little longer. Inflammation doesn’t fully resolve. Sleep fragments. Each of these, measured alone, might look like a mild abnormality. But what’s actually happening is more fundamental: the couplings between channels are weakening. The mutual information is declining. The joint entropy is creeping upward.

And because this is a dissipative structure, the failure isn’t linear. You don’t get 10% worse, then 20%, then 30%. You hold, and you hold, and you hold — and then you hit the bifurcation point, and the system undergoes a phase transition. The organizational pattern that was your health reorganizes into something else. Clinically, this looks like sudden onset. A collapse that seems to come from nowhere. But the physics was moving toward it the whole time, visible in the slow erosion of inter-channel coherence that nobody was measuring because medicine doesn’t think in terms of mutual information.

What the Biggest Model Gets Right (And What It Misses)

There is one major existing framework that already thinks about biology in these terms: Karl Friston’s free energy principle.

The FEP proposes that living systems minimize variational free energy — essentially, the mismatch between what the system predicts and what it encounters. Under the FEP, your brain builds a generative model of the world and continuously updates it to reduce surprise. When the model is good — when predictions match reality — free energy is low. When the model fails — when the world delivers something unexpected — free energy spikes, and the system must either update its model or act on the world to make its predictions come true.

The ESF respects this framework. At a high level, the two are saying compatible things: biological regulation is about maintaining organized states against entropy. Coherence maps onto low free energy. Incoherence maps onto high free energy. The math rhymes.

But the ESF parts company with the FEP at four specific points — and these are not minor technical quibbles. They are structural differences that change what questions you ask and what interventions you design.

First, the body. The FEP, as typically developed, lives in the brain. The generative model is neural. Prediction error is cortical. The periphery — the immune system, the gut, the autonomic nervous system, the mitochondria — appears mainly as an effector: the brain predicts, the body executes. The ESF insists that this gets the directionality wrong, or at least incomplete. Regulatory coherence is a whole-organism property. Stress incoherence can originate in the periphery — in immune dysregulation, in mitochondrial dysfunction, in autonomic instability — and propagate upward into the brain. The periphery is not waiting for instructions. It is a primary regulatory player.

Second, developmental time. The FEP describes a system minimizing free energy right now. The ESF embeds regulation in history. Your allostatic envelope was shaped by your developmental experience — the signal environment of your early life calibrated which regulatory configurations you can access. Two people facing the same stressor today can have fundamentally different free energy landscapes because their developmental histories carved different envelopes. The FEP acknowledges this in principle, through hierarchical priors. But it rarely centers it.

Third, the geometry of the problem. The FEP assumes a system minimizing free energy, but it doesn’t formally address the possibility that different systems might be minimizing free energy across state spaces of different dimensionality. This is where the ESF’s bio-neurotype concept enters, and it matters enormously. We’ll come back to it.

Fourth, the collective. The FEP models individual organisms. The ESF proposes that entropy management is partly a cooperative problem — that different bio-neurotypes within social groups distribute the regulatory workload differently, and that this distribution is itself an evolved strategy. The unit of analysis is not just the organism. It is the organism in its cooperative context.

The Neuroperipheral Problem (Which Is Not a Problem)

Here is where all of this converges on neurodiversity.

The ESF proposes two broad bio-neurotypes: neurosocial and neuroperipheral. These describe the relative weighting of regulatory architecture — whether a nervous system leans more heavily on social-affiliative circuits or on peripheral sensory and autonomic circuits for its coherence maintenance.

Neuroperipheral systems — the pattern associated with autism and related profiles — process more raw signal from the body and the environment. More sensory data. More autonomic fluctuation. More interoceptive input. More peripheral noise that the brain has to organize into something coherent.

In Shannon’s terms, this means the neuroperipheral system’s regulatory state space has more dimensions. More channels contributing to the joint distribution. More degrees of freedom across which mutual information must be maintained.

And entropy scales with dimensionality.

This is not a subtle point. A system managing coherence across, say, seven strongly coupled regulatory channels faces a fundamentally different thermodynamic problem than a system managing coherence across twelve. The potential entropy — the size of the state space that must be constrained — is larger. The mutual information that must be maintained is more extensive. The energy cost of keeping all those channels in conversation is higher.

The ESF calls this difference in thermodynamic workload a coherence tax. And it has direct implications:

Bandwidth is narrower. Not because the system is deficient, but because it’s already managing more signal. There is less margin for additional perturbation before coherence begins to degrade. The allostatic envelope is structurally tighter.

The bifurcation boundary is closer. In Prigogine’s terms, the neuroperipheral system’s dissipative structure operates nearer to its phase transition threshold under baseline conditions. Less additional stress is needed to push it across. This is the thermodynamics behind what clinicians observe as “low frustration tolerance” or “sensory sensitivity” — names that imply deficit, for a phenomenon that is actually about the physics of maintaining coherence across a larger state space.

But coherence, when achieved, is extraordinary. A system that maintains low joint entropy across a high-dimensional state space is doing something more thermodynamically impressive, not less. The pattern recognition, the perceptual depth, the ability to detect signal in what others experience as noise — these are what coherence looks like when there is more signal to cohere. The music, when the high-entropy ensemble locks in, is richer precisely because there are more instruments playing.

What Changes When You See It This Way

The following table maps the shift. Left column: how stress, disorder, and neurodevelopmental variation are typically framed in research and clinical practice. Right column: what the entropy framework proposes instead.

Dimension	Conventional Framing	ESF Entropy Framework
What “disorder” means	Deviation from normative function; categorical pathology	Rising entropy across regulatory state space; declining mutual information between channels
Role of entropy	Not formally engaged	Central: disorder is thermodynamic entropy in regulatory systems
What gets measured	Single biomarker or single behavioral dimension	Joint state space across channels; coherence as mutual information
How disease happens	Linear: gene → brain → behavior; cause → effect	Nonlinear phase transition at bifurcation boundary
What neurodevelopmental variation is	Deficit or deviation from typical	Different state space dimensionality; different entropy landscape
Where regulation lives	Brain-centric; periphery as effector	Whole-organism; periphery as primary regulatory player
What development does	Accumulates risk factors (dose-response)	Calibrates the dissipative structure; shapes the allostatic envelope
What intervention targets	The aberrant signal (suppress cortisol, modulate serotonin)	Inter-channel coherence; entropy load reduction; envelope widening
What recovery means	Symptom reduction; return to normative range	Re-coherence: restoration of mutual information across channels
What neurodiversity is for	Social identity (social model) or clinical population (medical model)	Cooperative entropy management: distributed regulatory strategy across collectives
How collapse happens	Gradual worsening; cumulative damage	Phase transition at bifurcation boundary; threshold dynamics
What environment does	Delivers stressors (magnitude matters)	Shapes entropy load through information quality (signal coherence of inputs, not just dose)

The Word We’ve Been Using Wrong

Disorder.

Dis-order. The loss of order. The increase of entropy. The expansion of the state space a system occupies when the constraints that held it together weaken.

Medicine has been using this word for centuries without noticing that it already has a precise scientific meaning — and that the scientific meaning changes everything about how you respond to it.

You don’t fix entropy by correcting the system. You don’t fix it by finding the broken part and replacing it. You fix it by restoring the conditions under which the system can maintain its own coherence — by reducing the entropy load, by supporting recovery, by not demanding that a high-dimensional system perform as though the physics of its organization were simple.

You respect the thermodynamics. You widen the envelope. You let the flame re-establish itself.

And you stop calling it a disorder when what you mean is that the physics is hard.

This essay reflects the perspective of the Evolutionary Stress Framework (ESF), a conceptual lens developed by the Center for Adaptive Stress. The ESF applies complexity science to stress physiology and neurodiversity. Its claims are offered as theoretical propositions and hypotheses for investigation — not as established clinical conclusions or medical advice.Lori is an autistic independent scholar and founder of the Center for Adaptive Stress. She writes at evostress.blog and @peripheralminds.

The Evo-Stress Blog