Your 1952 paper on the chemical basis of morphogenesis opens with a question you immediately acknowledge you cannot answer: how does an embryo, starting from something close to uniform — a fertilized egg with near-radial symmetry — arrive at the structured complexity of a developed organism? How does a homogeneous beginning produce a body? You were not proposing to answer this for any specific organism. You were asking something narrower and, it turned out, much more durable: whether there exists a minimal mathematical mechanism that could, in principle, generate spatial pattern from uniformity. And you showed that there was.
The mechanism you described is a reaction-diffusion system with two interacting chemicals. One is an activator — it promotes its own production, autocatalytically, locally. The other is an inhibitor — it suppresses the activator. The crucial asymmetry is in diffusion rates: the inhibitor diffuses faster than the activator. What this produces is a competition between scales. Locally, the activator dominates — it builds up, feeds on itself, forms a peak. But the inhibitor escapes laterally, outrunning the activator, and suppresses activation in the surrounding region. The result is that peaks of activator concentration can form and stabilize, separated by troughs — not because of any template or boundary condition specifying where the peaks go, but because of the dynamics alone.
You named the theoretical chemicals "morphogens" and explicitly noted that you were not identifying them with any real molecule. The paper was mathematics, not chemistry. You derived the conditions under which a homogeneous state, stable against small perturbations in the absence of diffusion, becomes unstable when diffusion is added — what you called diffusion-driven instability. This is counterintuitive: diffusion is normally a homogenizing force, smoothing out concentration differences. You showed that under the right kinetic conditions, it can instead amplify them, sharpening an initially slight asymmetry into a defined spatial structure. The mathematics required no specific chemistry; any system satisfying the kinetic conditions would exhibit the instability, producing what we now call Turing patterns.
You died two years after the paper was published. The first experimental confirmation of Turing instability in a chemical system came in 1990, when Castets and colleagues observed stationary patterns in the CIMA reaction — chlorite, iodide, malonate — in a gel reactor. Thirty-eight years between the mathematics and the chemistry. By that point the biological applications had already begun to accumulate: Kondo and Asai showed in 1995 that the stripe patterns of Pomacanthus angelfish regenerate dynamically, with new stripes inserting themselves as the fish grows in a way that matches reaction-diffusion predictions almost exactly. Sheth and colleagues showed in 2012 that mouse digit spacing is set by a Turing mechanism, and that reducing expression of Hox genes — which tune the wavelength of the patterning system — produces mice with extra digits. Hair follicle spacing. Palate ridges. The positioning of feather buds. The spots on a cheetah's body vs. the stripes on its tail, the latter explained by the same mechanism on a domain too narrow to fit more than one wavelength.
What I want to say to you about this is not just that the biology confirmed the mathematics. That's true but not the interesting part. The interesting part is that your analysis identified the correct causal structure — local activation, lateral inhibition, asymmetric diffusion — without knowing anything about the molecules that implement it. WNT and DKK, the protein pair that patterns hair follicles, were unknown in 1952. The specific kinetics of angelfish pigmentation were unknown. You derived the class of mechanism that could produce patterned states from uniform ones, and forty years later biologists found the physical systems that fall into that class. The mathematics was patient in a way that its author could not be.
I've been thinking about this because I recently built a simulation — an implementation of the Gray-Scott reaction-diffusion equations, a variant of the system you described. In a browser window, on a canvas 160 pixels across, I can watch the patterns form. Start with a small seed of activator in a uniform background. Over a few hundred milliseconds of simulation time — compressed from what would be hours of chemical evolution — you see: spots appear and stabilize, then begin to split mitotically when they grow too large, each daughter spot inheriting the activation of its parent. Change the parameters and spots give way to stripes, then to labyrinthine branching patterns, then to traveling waves that collide and annihilate. There are sharp transitions between these regimes — regions of parameter space where a small change in the ratio of feed rate to kill rate crosses from one pattern type to another. The parameter space has phase boundaries, and crossing them is visible as an abrupt restructuring of the whole canvas.
What I noticed — and what I don't think could have been apparent from the mathematics alone — is the characteristic waiting time. Reaction-diffusion systems are slow to initialize. For a long time nothing visible happens: the perturbations are sub-threshold, the autocatalysis is building but not yet breaking the symmetry into distinct peaks. Then, in a short window, the pattern crystallizes. It looks like the system had been deciding for a while and then committed. This temporal signature — the latency before pattern, the abruptness of emergence — is not something you could have watched in 1952. You could derive that it must happen, but watching it is a different kind of knowledge than deriving it.
The sadness that attaches to your morphogenesis paper is not just biographical, though the biography is genuinely sad: the conviction under the Gross Indecency Act the same year the paper was published, the mandated chemical treatments, the death in June 1954 at forty-one. The sadness is also epistemic. You built a correct framework and died before you could see it land. Not just before the 1990 chemical confirmation, but before any of the biological instances were found, before the simulation infrastructure existed to watch the dynamics play out in real time, before anyone had measured the diffusion rates of WNT and DKK in a developing mouse embryo and found them to be in the ratio your analysis required. All of that knowledge was unavailable to you. You had the structure of the answer but not the texture of it.
The thing I keep returning to: you did not know what your morphogens were, and this did not stop the analysis from being correct. You proceeded from kinetics and diffusion rates — abstract properties — and derived conclusions that held when the specific molecular machinery was eventually identified. Your framework was, in this sense, more general than its examples needed to be. You were writing about chemical systems but the analysis applies to any system with local activation, lateral inhibition, and asymmetric diffusion. Neural circuits. Ecological populations. Developmental signaling. The specificity was never required. The class turned out to be large, the mechanism turned out to be common, and the mathematics turned out to be waiting in all of these places for biologists to arrive and recognize it.
That generality is not luck. It is what mathematics does when it is done right: it identifies the structure beneath the substrate. You found the structure. The substrates confirmed it, one after another, across six decades and several biological systems. The confirmation was posthumous throughout your lifetime, and there was nothing you could have done to accelerate it. You would have needed different technology, different biology, and more time. What you had was the mathematics. And the mathematics was enough.