Diffusion erases. That is almost the definition of diffusion — a process that takes a concentration gradient and eliminates it. You drop dye in water and diffusion spreads it until the whole container is uniform. The information about where the dye landed is destroyed.
In 1952, Alan Turing proved that diffusion can also create patterns from a perfectly uniform state.
The paper is "The Chemical Basis of Morphogenesis," published in the Philosophical Transactions of the Royal Society B, August 1952. Turing was under criminal prosecution for "gross indecency" when he submitted it — the paper was revised March 15, 1952, roughly a year before his death. It proposes that two chemicals diffusing through tissue, interacting in the right way, can spontaneously break the symmetry of a homogeneous starting state and produce periodic structure: spots, stripes, whorls. He called the chemicals morphogens. He modeled a ring of cells, found six distinct instability types, and noted that the most biologically interesting one produces stationary concentration waves — the mathematics of a leopard's spots, written before anyone had confirmed leopards were doing mathematics.
The mechanism is counterintuitive in the precise sense. Both conditions are necessary for the surprise: (1) without diffusion, the system is stable, converging to uniform steady state; (2) with diffusion, it's unstable, spontaneously diverging into pattern. The thing you add to destroy gradients is the thing that generates them. This is called diffusion-driven instability, and it is still not obvious why it works even after you understand it.
The intuition Gierer and Meinhardt gave it in 1972, reformulating Turing's math into the activator-inhibitor framework: imagine an activator that promotes its own production — a small random fluctuation grows. But the activator also stimulates production of an inhibitor that spreads faster and suppresses activator production in the surrounding region. The result is a concentration peak surrounded by a ring of suppression. Adjacent peaks can't form too close. The characteristic spacing between peaks — the wavelength — is set by the ratio of the two diffusion rates. Uniform starting state, plus chemistry, plus the right ratio: stripes emerge.
Turing's paper was treated as theoretical curiosity for decades. The problem was physical plausibility: the inhibitor needs to diffuse substantially faster than the activator — roughly an order of magnitude in most models. Large proteins in tissue don't obviously achieve that ratio. The math worked, but biologists couldn't see how the biology would do it.
The decisive year was 2012. Two independent papers appeared within months of each other. The first, in Nature Genetics, studied palatine rugae — the ridges on the roof of the mouth. Researchers surgically excised a ruga from a palatal explant. Simple lateral inhibition models predicted that removing inhibitor near the cut would trigger new stripe formation at the edge. Instead, new stripes appeared as bifurcating branches from the neighboring ruga, meeting at 120-degree angles — the diagnostic signature of a Turing system reestablishing itself. This ruled out competing explanations. The molecular activator/inhibitor pair was identified: FGF and Sonic hedgehog.
The second paper, in Science, manipulated Hox gene dosage in developing mouse digits and watched fingers multiply and thin as predicted by changing the Turing wavelength parameter. At extreme Hox reduction the digit pattern transformed toward fish fin ray geometry — a direct look at the developmental mechanism behind the fin-to-limb transition. Hox genes turned out to be wavelength modulators, not pattern generators.
But the most interesting resolution of the physical plausibility problem came earlier, from zebrafish. The black and yellow stripes on zebrafish fins and body involve three cell types. Laser ablation experiments confirmed the Turing signature: destroy a stripe region, and neighboring stripes slide laterally to fill the gap at the correct spacing, exactly as the model predicts. When researchers looked at what the activator and inhibitor actually were, they found something unexpected. The xanthophores — yellow cells — extend long physical protrusions across the tissue, providing survival signals to melanophores at a distance. Short-range contact between the two cell types is mutually lethal. The "diffusion" is cells reaching out arms.
No actual diffusing molecule is required. The mathematics of Turing's mechanism — short-range activation, long-range inhibition, characteristic wavelength — is substrate-independent. You can implement it with diffusing proteins or with cell protrusions or, it turns out, with coral reef hydrodynamics at the scale of kilometers. The 2024 paper on atoll reef ridge patterns found the same Turing topology operating at ecological scale, where the activator is coral growth and the inhibitor is turbulent water flow.
In 2023 — seventy-one years after Turing's paper — researchers at Edinburgh identified the molecular mechanism behind human fingerprint formation. The activator is WNT/EDAR signaling; the inhibitor is BMP. The pattern initiates as waves propagating from the tip of each digit, and the outcome — whorl, loop, or arch — depends on where those waves collide, which depends on the micro-geometry of each embryonic finger. This is why identical twins have different fingerprints. Same genes, same Turing machinery, different micro-geometry means different wave collision patterns. The fingerprint records an event in embryonic development, not a genetic blueprint.
There is a version of this story that is about Turing's underappreciated biology paper, the one that isn't the Turing test or the universal machine. But the thing that holds my attention is the mechanism itself. The instability works because the inhibitor moves faster, which means that for any localized activation, the suppression outpaces itself — it spreads past the thing it's supposed to suppress, leaving the center unprotected. The center persists. The surrounding region is dampened. Spacing emerges from that asymmetry. Structure from the asymmetry in how fast destruction travels.
Turing was describing morphogenesis, how form develops from formlessness. He was doing it in 1952, in a paper that was largely ignored for twenty years, with equations that required computer solution he didn't have. The confirmation came slowly, then all at once: 2009 for zebrafish, 2012 for the decisive proofs, 2021 for cat fur, 2023 for fingerprints. The biology was doing it the whole time.