Coordination without a leader

A flock of starlings turning in unison looks choreographed, and it is not. There is no leader, no plan, and no bird with any idea what shape the flock is making. Each one is following a couple of rules about its nearest neighbours, and the shape is what happens when thousands do that at once.

Three rules, and nothing else

Craig Reynolds demonstrated this in 1986 with a simulation he called boids. Each simulated bird could see only its immediate neighbours, and obeyed three rules simultaneously.

Run those three rules and you get flocking, immediately and robustly. Not an approximation of it: the characteristic behaviour, including splitting around obstacles and rejoining, comes out without being written. Reynolds' boids are one of the cleanest demonstrations in computing that a global pattern need not exist anywhere in the parts that produce it.

What real starlings do

When physicists eventually measured actual starling flocks in three dimensions, the boids picture held up with one important correction. Birds do not track everything within a fixed distance. They track a roughly fixed *number* of neighbours, about six or seven, regardless of how far away those are.

This matters more than it sounds. A fixed-distance rule means a flock that spreads out loses cohesion, because neighbours drift beyond the radius. A fixed-number rule keeps the network of influence intact at any density, which is exactly what you want when a predator strikes and the flock expands violently. The behaviour survives the disruption because the connections do.

The same shape, in things without brains

Local rules producing global order is not a bird phenomenon, which is the strongest evidence that the mechanism rather than the animal is doing the work.

Ants find short paths between nest and food with no ant knowing the route: each lays a trail chemical and prefers stronger trails, and because shorter paths get traversed more often per unit time they accumulate more signal, so the colony converges on them. Slime moulds, which are single cells with no nervous system at all, solve comparable network problems by the same kind of reinforcement. Termite mounds with functioning ventilation are built by termites responding to local cues in the material other termites left.

In every case the coordinating information lives in the environment or in the local interaction rather than in any individual, and no participant holds a representation of the result. See what emergence actually means for the test that separates this from the cases where somebody simply wrote the answer in.

Why none of this happens here

This world has the ingredients you would expect to be necessary and does not produce flocking, and the reason is worth stating because it is a design decision rather than a failure to look.

Organisms can perceive their surroundings, including two channels summarising the eight cells around them and sixteen describing four directions out to their sensing range. There are signal and trace fields, so one organism can deposit something another detects, which is the ant mechanism in miniature. Movement is continuous and cheap enough to steer with.

What is missing is any channel that identifies another organism as an organism. Nothing here can perceive a neighbour's heading, because no sensor reports one. Nothing can compute an average position of nearby individuals, because individuals are not what the senses return. Every channel is a generic quantity, listed in full on how organisms perceive and act.