×
About Turing Patterns
How does a leopard get its spots, when every cell in its skin carries the same instructions? In 1952 Alan Turing gave an answer that sounds too simple to work: two chemicals that spread through the skin and react with each other can, on their own, break an even mix into spots or stripes. This page lets you watch that happen.
The history
1952. Turing's paper The Chemical Basis of Morphogenesis appeared in the Philosophical Transactions of the Royal Society , two years before his death. He coined the word morphogen for a chemical that shapes a growing body, and showed that spreading, which normally smooths everything out, can instead make an even mix unstable so that a pattern grows out of tiny random differences. He worked through his examples by hand and on the Manchester University computer.
1972. Alfred Gierer and Hans Meinhardt put the idea in the form most people now learn: a short-range activator that makes more of itself, and a long-range inhibitor that it also makes, which spreads faster and holds it back.
1980s. Peter Gray and Stephen Scott studied a simple chemical reaction in a stirred tank, in which one chemical turns another into more of itself. Spread out over a surface instead of stirred, their reaction is the Gray–Scott model this page runs.
1988. James Murray's article How the Leopard Gets Its Spots argued that the size and shape of the skin when the pattern is laid down decides what can form: a thin tail has room only for bands across it. So a spotted animal can have a ringed tail, but a striped animal should not have a spotted tail. The ANIMALS tab tries this.
1993. John Pearson ran the Gray–Scott model on a computer for a wide range of feed and kill rates, and sorted the patterns he found into a dozen or so types, which he labelled with Greek letters. THE MAP tab makes a small version of his map.
1995 onwards. Shigeru Kondo and Rihito Asai showed that the stripes of a marine angelfish shift and rejoin as the fish grows, as a Turing pattern should, rather than simply stretching. Later work on zebrafish found that the pigment cells themselves interact in a way that behaves like Turing's chemicals. In 2012 a study of the ridges on the roof of a mouse's mouth found pairs of signalling molecules acting as activator and inhibitor.
How it works
The dish is a grid of 256 × 256 cells; each holds an amount of chemical A and chemical B . Every step, each cell changes by:
Spreading. Each chemical moves towards the average of the cell's eight neighbours (sides counted four times as much as corners). A spreads twice as fast as B.
Reaction. A meeting two lots of B becomes B: the amount A × B × B turns from A into B each step. B makes more of itself, so it is the activator.
Feed and kill. Fresh A is fed in everywhere at rate f , topping A up towards 1, and B is removed at rate k (plus f , as the tank drains).
A patch of B grows by using up the A around it. Because A spreads quickly, the shortage reaches further than the B does, and that shortage is what stops a neighbouring patch from starting too close. Short-range making and long-range holding back: that is Turing's recipe, here with the shortage of A playing the inhibitor. The feed and kill rates decide whether the patches end up as spots, stripes, mazes or nothing at all.
On the dish, the darker colour is where B is high. The edges wrap round, so the pattern has no edge to line up with. Painting adds drops of B where you drag. The animal outlines have sealed edges instead: nothing spreads out of the shape.
Simplifications
The Gray–Scott model is a chemistry model, not biology. Real skin patterns come from pigment cells, signalling molecules and growth, and few have been pinned down molecule by molecule. The patterns are alike in kind, which was Turing's point.
The steps are a simple sum (each step adds the rates once), with A spreading at rate 1 and B at 0.5 per step, as is usual for this model on a computer. Bigger steps would blow up; smaller ones give the same pictures more slowly.
The map is a sheet of small separate dishes, each 32 cells across and started from two drops, so a pattern bigger than a tile can't show and some tiles depend on where their drops landed.
The seashell is the one-line version run over time, each row of the picture the line at one moment, clipped to a shell shape. Real shells lay down pigment along the growing lip in much this way, as Hans Meinhardt modelled in The Algorithmic Beauty of Sea Shells (1995), but with other reactions.
The dish opens on a Spots specimen grown earlier and stored with the page at half size; GROW starts again from drops so you can watch it form.
Keys: Space pause or carry on, G grow, C clear, 1 to 8 pick a specimen (on GROW).
FURTHER WATCHING AND READING