A brainless yellow blob spread across a wet plate in a Japanese lab and, meaning nothing by it, sketched a workable first draft of the Tokyo rail network.

The organism was Physarum polycephalum, a slime mould that lives on rotting logs and looks like something spilled. No neurons, no brain, no headquarters issuing orders. It is a single cell, though an odd one, holding thousands of nuclei inside one continuous body that can grow to the size of a dinner plate.

A team led by Atsushi Tero and Toshiyuki Nakagaki placed oat flakes at 36 points on a sheet of agar, each flake standing in for a city in the greater Tokyo area. The mould was set down at the spot representing Tokyo itself and left to feed. It explored everything, then pruned. What was left linked all 36 food sources, and when the group measured that lacework against the real railways, the paper published in Science reported comparable numbers for total tube length, for average travel distance between points, and for how well the whole thing coped with a random break.

How a cell without a plan makes a decision

The rule is almost embarrassingly simple.

Fluid sloshes back and forth inside the mould’s tubes. Where flow runs heavy, the tube widens. More fluid follows the wider path, so it widens again. Where flow runs thin, the tube narrows and eventually disappears. Feedback handles the rest. No tube knows where it sits in the wider structure, and nothing in the organism weighs one option against another. Engineers find that part unsettling. Tero’s group turned the behaviour into a set of equations and showed the model could produce similar networks to order.

Where did the mould get its map of Japan?

From a lamp. Slime mould shrinks from bright light, so illumination worked as fencing: shaded patches marked low-lying, liveable ground, while the ocean and the inland lakes were lit hard enough to keep the blob out. Mountains got the same treatment.

Geography, then, was handed over rather than discovered. What the mould supplied was the routing, and routing is the expensive part. Anyone who has watched a rail corridor get argued over for a decade knows the terrain was never the difficult bit.

Before the trains, there was a maze

In 2000, Nakagaki’s lab chopped a mould into fragments, spread them through a small plastic labyrinth and set oats at the entrance and the exit. The pieces fused, filled every corridor, then withdrew from the dead ends until one thick tube ran along the shortest route. Nakagaki, Hiroyasu Yamada and Ágota Tóth published that finding in Nature, filed under the heading “Intelligence”.

Learning arrived later, from a different lab. Romain Boisseau, David Vogel and Audrey Dussutour, working in Toulouse, built bridges to a food reward and soaked some of them in caffeine or quinine, both bitter and both harmless. Moulds crossed the untreated bridges around three times faster to begin with, then, across six days, stopped minding the bitterness. Withhold the stuff for two days and the caution came back. Writing in Proceedings of the Royal Society B, the Toulouse group treated this pattern as habituation, the plainest form of learning.

One paper is one paper. Whether something without a nervous system can be said to learn at all remains genuinely contested, and a handful of studies is not a consensus. Vogel and Dussutour did follow up, reporting in the same journal that a habituated mould can pass the trait to a naive one when the two merge into one body.

Astronomers borrowed the idea

“It’s somewhat coincidental that it works, but not entirely,” Joe Burchett said, describing to UC Santa Cruz what happened when slime mould logic got loose on the biggest structures in existence.

Burchett, an astronomer, and Oskar Elek, a computer scientist, took a virtual Physarum model, extended it into three dimensions and fed it the positions of roughly 37,000 galaxies from the Sloan Digital Sky Survey. Out came the cosmic web, those vast filaments of gas and dark matter along which galaxies are strung like beads. Run against a dark matter simulation, the algorithm reproduced the filament pattern closely, and the pair then used the map to estimate how much gas fills the space between galaxies. Their results appeared in the Astrophysical Journal Letters.

Different physics entirely, obviously. Gravity is not hunger. Both processes just happen to end up wiring scattered points together on the cheap.

A prize, and a caution

In 2010 the Tokyo team collected an Ig Nobel Prize for transportation planning, the annual award for research that gets a laugh before it gets a second thought.

Laughing is the easy part.

Thinking is where it stings, because the mould’s great advantage is having no opinions. It cannot lobby for a station near its own house, cannot defend last year’s ridership forecast, cannot mistake a sunk cost for a reason. Give it flow and time and it fattens whatever carries traffic and lets the rest wither. Planners around Tokyo arrived at something similar over decades of committees and compromise. The mould managed it in a matter of days, mostly because nobody on the committee objected.