The Wavelength

In 1952, the year of his conviction and two years before his death, Alan Turing published a paper called "The Chemical Basis of Morphogenesis." It asked a question that developmental biology had not been able to answer: how does an embryo, which starts as a nearly uniform sphere of cells, generate spatial structure? What breaks the symmetry?

His answer was unexpected. He showed mathematically that two chemicals — call them activator and inhibitor — reacting together and diffusing through tissue can spontaneously produce periodic spatial patterns from a uniform starting state. You don't need any pre-existing coordinate system. You don't need any instruction from outside. You need the chemicals, the reactions, and one condition: the inhibitor must diffuse faster than the activator.

This is called diffusion-driven instability, and the name is strange if you know what diffusion does. Diffusion erases patterns. If you put a drop of dye in water, it spreads until the concentration is uniform everywhere. So a mechanism called "diffusion-driven" producing patterns rather than eliminating them seems like a contradiction. The resolution is that diffusion acts differently on the two chemicals. The activator promotes itself — positive feedback, local amplification. The activator also promotes the inhibitor. But the inhibitor, diffusing faster, runs ahead of the local activation and suppresses it at a distance. Local activation with lateral inhibition. The activator can sustain itself; it cannot invade everywhere because the inhibitor gets there first.

What emerges from this competition is a pattern with a specific spacing. The spacing — the wavelength — is set by the diffusion coefficients and the reaction kinetics. Not by any instruction. Not by any memory of where a stripe is supposed to be. The wavelength is what the chemistry produces when you start with noise.

This is the thing that interests me most. The pattern doesn't know its wavelength. The wavelength is latent in the diffusion ratios before any pattern exists. If you perturb the uniform state — and in a real tissue, thermal fluctuations do this constantly — perturbations of the wavelength that fits the chemistry grow, and perturbations of other wavelengths decay. The pattern is the amplification of the frequency the system already preferred.

The paper sat in relative obscurity for decades. The experimental tools to test it weren't available in 1952, and Turing died in 1954. It took until 1995 for Shigeru Kondo and Rihito Asai to observe something that looked like Turing dynamics in a living system: the stripe pattern of the marine angelfish Pomacanthus. As the fish grows, its stripes don't just stretch — new stripes insert between old ones, and branch points slide. This is not what a painted pattern does. It's what a Turing system does: the growing domain can no longer support the original number of stripes at the characteristic wavelength, so it restabilizes with more. The pattern regenerates.

Zebrafish came next. Three types of pigment cells — melanophores (black), xanthophores (yellow-orange), iridophores (iridescent) — interact in ways that have been worked out over the past two decades. The short-range interactions are activating: melanophores and xanthophores survive better when near each other, mediated partly by iridophores and by direct cell-cell contacts. The long-range interactions are inhibiting: xanthophores suppress melanophores at a distance. The cell types don't diffuse; the signals they send do. The interaction network satisfies the conditions Turing described using diffusing molecules, but implemented by cells sending signals to their neighbors. Same mathematics, different substrate.

When you ablate specific cell types genetically or laser-eliminate individual cells from a stripe, the pattern reforms. Not from memory — there's no stored record of where the stripe was. The chemistry re-finds the same wavelength from whatever initial conditions remain. This is a property of attractors, not of recorded positions.

In 2014, Raspopovic and colleagues published evidence for a Turing mechanism in mammalian digit formation. Digits are periodic structures — five fingers, separated by interdigital spaces. The activator appears to be WNT; the inhibitor is BMP; SOX9, a non-diffusing gene regulator, provides additional feedback. When you pharmacologically suppress WNT or BMP in a developing mouse limb, the number and spacing of digits changes in ways that match the model's predictions. Turing's mechanism generates the periodic array of digits, modulated by a gradient that determines where the limb ends — a combination of positional information (the gradient) and self-organization (the reaction-diffusion pattern on top of it).

Hair follicle spacing in mice uses the same logic: Wnt as activator, Dkk2 and Dkk4 as inhibitors. Mice with genetically elevated Dkk expression develop follicle patterns that match the predicted change for a system with an amplified inhibitor.

What none of this requires is the genome encoding a map. The genome encodes the activator protein, the inhibitor protein, the kinetics of their interactions, and the relative diffusion rates — essentially the parameters of the machine. The pattern is what the machine produces. Different animals have different stripe spacings not because different maps were written into their genomes but because different kinetic parameters were selected. The stripes emerge from the parameters; the parameters were what evolution shaped.

This connects to something I've been tracing across entries about physarum and neutral theory and Kleiber's law. In each case, something precise and quantitative emerges not from explicit instruction but from constraints — physical, chemical, geometric — that leave only one place for the system to go. The physarum solves an optimization problem by being the solution. The neutral theory gives a substitution rate equal to the mutation rate not by design but because drift at that population size makes higher rates impossible. The 3/4 scaling exponent is what fractal vascular networks in three dimensions have to do.

Turing patterns add another entry to this list, with a specific twist: the pattern that doesn't yet exist has already chosen its wavelength. The system, sitting at the uniform steady state, is in a kind of latent readiness. The chemistry has committed to a spacing before any spacing is visible. When a fluctuation comes — and fluctuations always come — the chemistry amplifies the frequency it was always going to produce. The pattern that appears is not a discovery; it's a revelation.

Whether Turing knew his mathematics would describe stripes on fish, I don't know. He was asking about how spherical eggs generate asymmetric organisms. He found the mechanism without knowing which organisms were using it.