The same number decides whether an epidemic spreads, a brain thinks, or a chemical soup makes itself

I went in expecting a physics story. A nuclear reactor either fizzles or runs away depending on a single number, and I assumed that was the end of it. Instead the same one-line rule kept surfacing where it had no business being: in measles outbreaks, in a living brain, in a chemical soup that makes itself. Here is the number, and here is where it holds and where it frays.

The "one more" rule is a chain reaction

Start with the example physics made famous. A nuclear chain reaction sustains itself only if each fission triggers, on average, at least one more fission. Below that, the reaction dies. Above it, it runs, and at the knife-edge it barely holds. The number that decides is one.

That arithmetic has a name older than the reactor that made it famous: a branching process. Francis Galton and Henry William Watson first wrote it down in the 1870s, to answer a Victorian puzzle about why aristocratic family surnames kept dying out. Picture anything that spreads step by step, where each active unit produces, on average, m new active units. If m is below one, the line of descent is doomed; trace any branch and it eventually hits zero. If m is above one, it has a real chance to survive and grow. The whole behaviour pivots on whether m sits below or above one. That single fact, written down a century and a half ago, turns out to be hiding inside a remarkable number of unrelated systems.

Epidemics: the number you already met in 2020 is R₀

The reproduction number R₀ is just the average number of people each case infects. An outbreak grows if R₀ is above one and fades if it is below. That is not a coincidence that resembles the chain reaction; it is the same branching process, with infected people in place of neutrons.

Measles makes it vivid. Its R₀ is usually quoted around 12 to 18, so each case in an unprotected population infects on the order of fifteen others. To stop it you do not need to reach everyone; you need to push the average each case infects back below one, and a little algebra says the share of the population that must be immune is 1 − 1/R₀. For measles that is about 93%. It is the whole reason measles demands such high vaccination coverage, and the reason it comes roaring back the moment coverage slips a few points. The same threshold, read from the control side.

When does a random network suddenly snap into one web?

Take a pile of dots and start drawing links between random pairs. While the average dot has fewer than one link, you get scattered little islands. The instant the average passes one link per dot, a single connected web snaps together and spans most of the dots. Erdős and Rényi proved this in 1960. Same threshold, same reason: building the network outward is a branching process, and it only takes off once each dot leads, on average, to more than one other.

How can a chemical soup start making itself?

For a soup of molecules to start producing itself, enough of them have to speed up each other's reactions. As the average number of reactions each molecule helps along crosses a narrow band of about one to two, the chance that a self-sustaining set exists jumps from almost zero to almost certain; Hordijk and Steel made this precise in 2004. And this is not only theory. When a 2015 study by Sousa and colleagues mapped the metabolism of E. coli, 1199 molecules and 1826 reactions, and gave it a set of simple feed molecules to draw on, they found a single self-producing set covering 1787 of those reactions, 98% of the whole. Living chemistry sits far above the threshold, and it leans on a few heavily used helpers to stay there. Remove one shared cofactor pool, the NAD and NADP carriers, which assist 298 reactions, and the self-producing set collapses by 1353 reactions, from 1787 down to 434. One family of helper molecules holds three-quarters of the network together. A set like this, given a food supply, is what a self-sustaining chemistry looks like: a web of reactions that, taken together, makes everything it needs to keep itself going. That is a long way from explaining how life first began, but it is the same threshold doing the work. Cross the catalysis level and the web switches on; fall below it and the same molecules are just a pond of unrelated reactions going nowhere.

A brain that lives right on the line

The systems so far cross the threshold. The brain seems to camp on it. When neuroscientists record the spontaneous bursts of activity that ripple through cortical tissue, called neuronal avalanches, they find that each active patch triggers, on average, close to one more. Beggs and Plenz, in 2003, measured a branching parameter of about 1.04, with sizeable scatter across preparations. The sizes of those bursts follow a precise fingerprint, a power law in which a burst of size s shows up with probability proportional to s to the power −3/2, much as theory predicts for a branching process sitting at criticality.

Why would a brain balance there on purpose? Because both ways off the line are failures. Tip the cortex below one, and a signal dies before it can cross the tissue, and a thought never propagates. Tip it above one, and activity runs away and recruits everything in its path, the dynamical signature of a seizure. But avoiding those two failures is only half the reason. A network poised exactly on the line is thought to be the one that computes best: it responds to the widest range of inputs and carries information the furthest, so sitting right at the threshold, rather than safely below it, may be what lets a cortex think at all. The branching number that decides whether an epidemic spreads is the same number the brain appears to hold near one so it can think without going silent or seizing.

Can a population really go extinct just from being small?

Yes, and it is the same rule seen from below. When animals depend on each other to find mates, raise young, or fend off predators, a shrinking population gives each individual fewer partners. Past a critical thinness, each one leaves behind fewer than one surviving descendant on average, and the population spirals to extinction even with food to spare. Ecologists call it the Allee effect.

It has a body count. In the African wild dog, a pack that drops below a critical size can no longer hunt or guard pups well enough to replace itself, so small packs slide toward zero rather than bouncing back. The most haunting case is also the most contested: the passenger pigeon, which went from billions of birds to none. The old story is that once the great flocks thinned below the size they needed to breed, the survivors simply could not recover. But newer genetic work argues that wild population swings, not thinness alone, set them up to fall. The unsettling point holds either way. A population can have all the habitat and food it wants and still cross a line below which being too few is itself the cause of death.

Where more connection backfires

Here is the part most "everything is connected" stories miss, and it is what keeps this from being a naive law. More connection is not always better. Past a point it is worse.

In 1972, Robert May proved that the more densely a large system is wired together, the more likely it is to be unstable, not less. The same logic later explained financial crises: Haldane and May showed in 2011 that dense banking networks are "robust-yet-fragile," shrugging off small shocks but transmitting big ones straight through the system. The cleanest demonstration comes from AI. A layer of a modern language model works by letting every word pull information from every other word, connection at its most extreme. What holds it up are the shortcut wires, the residual connections that carry each word's original signal straight past the mixing step. Strip those out, and Dong and colleagues showed in 2021 that the stack collapses: every word ends up looking like every other word, more completely the deeper the stack goes. The shortcuts are what save it.

So the living regime has two edges. Below one connection per part, the system falls apart into islands. With dense but undifferentiated wiring, it congeals into sameness. Life, and anything else that has to keep working, sits in the band between fragmentation and homogenization.

So is this one law of everything?

No, and the overreach is worth refusing directly. It is the same mechanism, but the thing that crosses the threshold differs by field: partners per animal, infections per case, catalysts per molecule, links per node. Calling them one number would be a trick. Calling them one mechanism, read on different materials, is accurate.

The honesty goes further. The clean equivalence holds best in idealized systems where the parts do not loop back on themselves. Real networks have clusters and short loops, and those shift the exact threshold and change the fine details. So the right claim is that these systems share a mathematical skeleton, sharp where they are simple and frayed where they are tangled. And the cross-field idea is not new: Scheffer and colleagues showed in 2009 that ecosystems, climate, and markets share warning signs before a tipping point, and percolation theory has unified threshold behaviour for decades. This extends that lens to chemistry and life, it does not found it.

Why one number is worth knowing

The threshold is not just a curiosity; it is the handle. Public health does not try to reach everyone, it tries to push R₀ under one, and the 1 − 1/R₀ rule sets the exact vaccination target for every disease, which is why measles needs about 93% coverage and a milder bug needs far less. A brain held near a branching parameter of one is thought to compute best for exactly that reason, while drifting off the line in either direction makes the tissue either fall silent or run away. In finance and ecology the same mathematics carries the warning Haldane and May made explicit: adding connections is not free, and past a point it flips a system from robust to fragile, which is how diversification can quietly manufacture the very contagion it was meant to prevent. Knowing the number tells you which way to push, and how hard.

What would prove it wrong

A claim worth anything can fail, so here is the test. If these really are one mechanism, then rescaled so the units cancel, the thresholds should all land at one. And each system should switch on the same way as it crosses the line, rising at the same rate just past it. For two of them, random networks and simple epidemics, it already checks out: both rise at exactly the same rate above threshold. Do the rest, from published data, and compare. If they coincide, the unification is real. If they scatter, it was a family resemblance, not a shared cause. That measurement, not any amount of eloquence, is what would settle it.

Until then, the picture that stands is a single pivot found again and again: an outbreak, a self-making soup, a random web, a thinning population, and a thinking brain, each one switching on or dying out at the same point, where every part triggers, on average, one more.