← The workshop

The Murmuration

Dusk over the reed bed. Each starling watches its seven nearest neighbours, and nobody leads.

This sky needs WebGL 2, which this browser has switched off.

Flying together–
In one flock–
A change of course is shared across–
Courses re-set–
Birds–

What is known

Starlings gather over their roost at dusk in flocks of hundreds to many thousands. Filming flocks over Rome in three dimensions showed that each bird keeps track of a fixed number of nearest neighbours, six or seven, however near or far they happen to be. It does not respond to everything within a fixed distance (Ballerini and colleagues, 2008).

A small change in one bird’s course is shared across the whole flock, whatever the flock’s size (Cavagna and colleagues, 2010). A turn begins with a few birds, usually at an edge, and passes through the rest as a wave that keeps its speed and does not fade (Attanasi and colleagues, 2014).

When a falcon attacks, dark bands run through the flock away from it (Procaccini and colleagues, 2011). The bands are probably birds rolling as they turn, so that more of the wing is seen. They are not birds bunching together (Hemelrijk and colleagues, 2015).

The rule used here

Each bird steers toward the average heading of its seven nearest neighbours, keeps a little distance from them, and leans back toward the roost when it strays. Its turning has inertia, as in the inertial spin model (Cavagna and colleagues, 2015), and it banks as it turns. The roost and the banking follow Hildenbrandt, Carere and Hemelrijk (2010). The darkness of the flock is only the birds overlapping and showing more or less wing.

The timing rule here is a count, a threshold and a restart. A bird keeps a running count of how far its neighbours have drifted from the course it last settled on. When the count reaches a threshold the bird settles on a new course, built from its neighbours’ recent headings, and the count starts again. This is the timing grammar of Alexander’s work on coherence, rupture and regeneration, which is under investigation. It is one way to make a flock. It is not a claim about how starlings decide.

“Steer without counting” removes the count. Every bird then re-reads its neighbours at every instant. In test runs of 2,000 birds the two flocks looked alike. Flying together measured 0.90 with the count and 0.95 without. With a single neighbour it fell to 0.70 with the count and to 0.35 without.

Where this model falls short: when a tenth of the birds at one edge were held on a new course, the far side turned by only about three degrees, and the turn crossed at about the birds’ own flying speed, with or without the count. Real flocks pass a turn across several times faster and at full strength.

What to try

Move “Neighbours” down to one. The birds still circle the roost, because it pulls on each of them alike, but they stop being one flock and the dark cloud thins. With two, most are linked again. Past seven, little changes.

Send the falcon. The flock opens around it and closes again behind it. In test runs the birds out of its reach turned only about a tenth harder than before, so the wave of alarm seen in real flocks is faint here.

Turn the sound on. The calls are synthesised, with two voices to a bird as in the starling’s two-sided syrinx. They are not recordings.

What is left out

What a bird can and cannot see through the others. The flattened shape of real flocks and the way neighbours sit to the side more than ahead. Changes of speed, which in real flocks are also shared across the whole flock. Tiredness, hunger, the difference between one bird and another, and the descent into the reeds at the end. Without the roost to hold it, this flock spreads and splits.

The falcon here only chases the middle of the flock. The birds are drawn about twice life size so that they can be seen. The numbers under the picture are measured from this run as it goes.

Sources. Ballerini M. et al. (2008) Interaction ruling animal collective behavior depends on topological rather than metric distance. PNAS 105, 1232–1237. Cavagna A. et al. (2010) Scale-free correlations in starling flocks. PNAS 107, 11865–11870. Attanasi A. et al. (2014) Information transfer and behavioural inertia in starling flocks. Nature Physics 10, 691–696. Cavagna A. et al. (2015) Flocking and turning: a new model for self-organized collective motion. Journal of Statistical Physics 158, 601–627 (cited from memory; check). Hildenbrandt H., Carere C. and Hemelrijk C. K. (2010) Self-organized aerial displays of thousands of starlings: a model. Behavioral Ecology 21, 1349–1359 (cited from memory; check). Procaccini A. et al. (2011) Propagating waves in starling, Sturnus vulgaris, flocks under predation. Animal Behaviour 82, 759–765 (cited from memory; check). Hemelrijk C. K., van Zuidam L. and Hildenbrandt H. (2015) What underlies waves of agitation in starling flocks. Behavioral Ecology and Sociobiology 69, 755–764 (cited from memory; check).