Collective Behaviour

In order to perform the phenomenal aerial displays that we observe, the starlings need to act in a highly coordinated way. This behaviour is of considerable interest not only to biologists but to physicists, engineers and mathematicians.

In the early 20th century, British ornithologist Edmund Selous sought to explain the remarkable synchrony and coherence of flocking birds in his book, Thought-transference (or what!) in birds. He concluded that a telepathic connectivity of minds must underlie this behaviour. [1]

They must think collectively, all at the same time, or at least in streaks or patches - a square yard or so of an idea, a flash out of so many brains

Although we can now dismiss Selous's telepathy hypothesis, over a century on, we still have much to learn about how social interactions affect the way animals within highly coordinated groups acquire and process information.

The field of collective animal behaviour boasts a wealth of models. I.e. How we think they might interact. But to test these models they must be compared to empirical evidence. Unfortunately, 3D data on even moderately large groups of animals is hard to obtain. Until recently, data has been scarce, and limited to small group sizes. We have relied on data such as studies conducted on fish in laboratory tanks going back to the 1960’s.

Papers published by an Italian group, including Nobel prize winning physicist Giorgio Parisi, from 2007 onwards have gone a long way to provide valuable data. They used a technique called stereophotography to reconstruct the 3D coordinates of each individual bird in flocks of up to 8000 birds. Stereophotography works in the same way that our eyes do to work out how far away an object is. By using two cameras (or in our case, both eyes) with different viewpoints we can triangulate an object’s position.

This was still a very difficult problem to solve due to many constraints both mathematically and practically. The major difficulty comes from figuring out which starling is which in your two images.

By studying these 3D reconstructions, the scientists could distinguish between 2 classes of aerial display [2]:

  1. Some particularly large flocks flew very high and performed the most dramatic displays, perhaps signalling the location of the roost.
  2. Smaller, lower altitude flocks flew in a random walk pattern above the roost and it was these groups that the scientists studied further.

Amongst the researchers' findings were the following observations:

  1. Flocks are thin and horizontal in shape, like a sheet of paper and they fly almost parallel to the ground. This makes sense to conserve energy, but to anyone who has craned their necks to watch what seems to be huge volumes of birds this may seem odd. The scientists explain that with our heads tilted towards the sky, we lose perception of the flock’s orientation and relatively thin aspect.
  2. The birds turn in equal radius paths rather than parallel paths. That is, they do not stay in formation. This could be a fascinating mechanism for distributing risk throughout the flock, as not all positions in the group are equivalent. Birds at the boundary are more likely to be attacked by a predator and if the cost/benefit of group membership were negative for too long, the group would break up. In fact, we see that flocks are incredibly robust and it seems as though by twisting and turning individuals exchange positions in the group and share the risk.
  3. The flocks were denser at the border than in the centre. Intuitively, the middle of the flock should be safer from predation so this result was surprising. The researchers suggested that the high density boundary “wall” may confuse any predators which tend to hunt by locking onto a single target, although how the starlings achieve this is a mystery.
  4. Given the need to avoid collisions, it was not a surprise to find that there is an exclusion zone around each bird, the diameter of which is comparable to the bird’s wingspan.

So, how do they move in such synchrony? Thanks to modern understanding of complex systems, we know that order such as this can have radically different origins. At one extreme, it can be the result of a top-down leadership structure. At the other end of the spectrum, order can arise organically from local behaviour rules.

How the flock reacts to predatory pressure provides crucial clues about its organisational structure. In a group where all individuals follow a leader, everyone will move in synchrony, but information transfer will be poor. Unless the falcon is detected by the leader, the group will not respond to it.

Starling flocks are what is known as “self organised” systems. When one bird detects a falcon attack it will turn to avoid it. This turn is noticed by its neighbours who also turn. They influence their neighbours, and so on. By assuming some simple rules we can describe the behaviour we see in flocks:

These rules are:

  1. Short range repulsion (collision avoidance).
  2. Attraction to the birds around you. This keeps the flock cohesive.
  3. Fly in the same direction and with the same speed as the birds around you.

By comparing these rules to the data, the scientists found that each bird interacts with its closest 6 or 7 neighbours. This is far fewer than the number of birds it can see. The starlings have optimised how much information to take in. If the bird takes information from too many neighbours the data is too noisy. Its decision making is ill-informed. If the information is taken from too few neighbours, the information is too short-ranged. [3]

Interestingly they found that this 6 or 7 number persists, regardless of the density of the flock. In fact, even sparse flocks keep their cohesion and because their visual range is so far any stragglers can quickly rejoin the flock.

Although each bird only interacts with 6 or 7 of those around it, we see flocks of hundreds, even thousands, moving in unison. How does this happen?

The starlings pass information from neighbour to neighbour to neighbour in a wave that crosses the entire flock so efficiently that it is almost instant. Birds on either side of the flock are effectively correlated with each other even though they do not directly interact. The group can respond as one, returning us to the concept of the “collective mind”. How the starlings achieve such long distance correlation is still a mystery to us. [4]



[1] Collective minds - I Couzin, NATURE, 2007

[2] Empirical investigation of starling flocks: a benchmark study in collective animal behaviour - M. Ballerini, G Parisi et al 2008, ANIMAL BEHAVIOUR (2008)

[3] Interaction ruling animal collective behavior depends on topological rather than metric distance: Evidence from a field study - M. Ballerini, G Parisi et al, PNAS (2008)

[4] Scale-free correlations in starling flocks - A. Cavagna, G. Parisi et al, PNAS (2010)