Explaining Bird Flocks
A dark flock of dunlins sprints straight over a marsh—until a merlin appears and they all veer at the same moment, flashing their bright white underparts and rearranging their group into an hourglass shape with shocking swiftness. A distant murmuration of starlings—and yes, that really is the marvelous term for a group of these often-maligned birds—10,000 or more, rolls “like a drunken fingerprint across the sky,” as the poet Richard Wilbur wrote, smudging the dusk horizon with the quickness of a pulsating jellyfish.
Since primeval times people have looked at masses of birds moving as one and wondered how they do it. The ancient Romans had their explanation: Gods, they believed, hinted at their intentions in the way birds flew. Scientists of the early 20th century, perhaps almost as credulous, groped for such mysterious and even mystical concepts as “natural telepathy” or a “group soul.” “It is transfused thought, thought transference—collective thinking practically. What else can it be?” mused one British naturalist, rather plaintively, in 1931.
Many birds flock, of course. But only a relative handful really fly together, creating what University of Rhode Island biologist Frank Heppner, in the 1970s, proposed calling “flight flocks”: namely, highly organized lines or clusters. Pelicans, geese, and other waterfowl form lines and Vs, presumably to take advantage of aerodynamic factors that save energy. But the most impressive flockers are arguably those that form large, irregularly shaped masses, such as starlings, shorebirds, and blackbirds. They often fly at speeds of 40 miles or more per hour, and in a dense group the space between them may be only a bit more than their body length. Yet they can make astonishingly sharp turns that appear, to the unaided eye, to be conducted entirely in unison. Imagine doing unrehearsed evasive maneuvers in concert with all the other fast-moving drivers around you on an expressway, and you get an idea of the difficulty involved.
No wonder observers have been left groping for an explanation. When Heppner, now semi-retired, began studying pigeon flocks more than 30 years ago, he suggested that they communicate through some sort of neurologically based “biological radio.”
“The fact that we weren’t hooted out of town is an indication of how desperate we were to explain this stuff,” he says now.
Today, though, technological innovations, from high-speed photography to computer simulations, have enabled biologists to view and analyze bird flocks as never before. So has a new wave of interest from other scientists, including mathematicians, physicists, even economists. As a result, researchers are closer than ever to really getting inside the mind of the flock.
“There’s a lot we don’t know now,” says Heppner, “but I think we’re actually going to know how and why birds fly in organized groups within five years.”
On one level it has long been obvious what’s going on when animals synchronize their movements—be they ducks, wildebeest, herring, or social insects. More eyes and ears mean increased opportunities to find food and improved chances of detecting a predator in time.
It’s when a predator lunges, though, that being in a crowd really pays off. Numerous studies have shown that individuals that travel in groups are almost always more vulnerable when they stray off by themselves. That’s due in no small part to the bewildering things that an assemblage can do. By turning rapidly or simply tilting a bit on their axis, dunlins are able to shift the appearance of their plumage from dark (their upperparts) to light (their underparts), creating a swift flashing effect that might startle or confuse predators. Studies have shown that merlins hunting shorebirds are in fact most successful when they’re pursuing individuals. Falcons do go after tightly packed crowds of dunlins and other shorebirds, but those hunts are most likely to succeed when the attack causes a solo bird to stray. Safety in numbers, in other words: Birds that stay together tend to survive together.
“Being single is always more risky,” says Claudio Carere, an Italian ornithologist who is involved in a collaborative study of flocking starlings in Rome.
The British evolutionary biologist William Hamilton, in 1971, coined the term “selfish herd” to describe this phenomenon. Each member of a flock, he wrote, acts out of simple self-interest. When a predator approaches a flock, all the individuals in the group move toward the safest place—namely, the middle of the group—in order to reduce the chances of being captured. Observations of juvenile shorebirds have hinted that it may take them a while to get the hang of this, because they learn to form cohesive congregations only over time. As they do, natural selection dictates that the birds least able to hang with the group are most likely to be caught by predators.
Self-interest by itself may explain many of the observed dynamics of flock motion, such as density. But it can’t explain how the birds get the information they need to move in synchrony and avoid a predator. There’s no way every member of the group can see a fast-flying falcon at the same time. How, then, can they possibly know what direction to move in to avoid it?
One clue came from studies of fish. Many schooling species maneuver as intricately as the most cohesive bird flocks—and they’re much easier to study, because they can be watched and photographed from above in open tanks. In the 1960s a Russian biologist, Dmitrii Radakov, tested schools and found that they can successfully avoid predators, as a whole, if each fish simply coordinates its movements with those of its neighbors. Even if only a handful of individuals know where a predator is coming from, he wrote, they can guide a huge school by initiating a turn that their neighbors emulate—and their neighbors’ neighbors, and so on. Unlike linear flocks of geese, which do have a clear leader, clusters are democratic. They function from the grassroots; any member can initiate a movement that others will follow.
Refining Radakov’s theory had to wait until the 1980s, when computer programmers began to create models that show how simulated animal groups can respond to the movements of individuals within them. It turns out that only three simple rules suffice to form tightly cohesive groups. Each animal needs to avoid colliding with its immediate neighbors, to be generally attracted to others of its kind, and to move in the same direction as the rest of the group. Plug those three characteristics into a computer model, and you can create “virtual swarms” of any sorts of creatures you like. They change density, alter their shape, and turn on a dime—just as real-world birds do. The makers of movies, from The Lion King to Finding Nemo, have used similar software to depict realistic-looking movements in large groups—whether stampeding wildebeest or drifting jellyfish.
The real world, though, doesn’t run like software. One problem with the basic model is that it doesn’t adequately explain how bird flocks can react as quickly as they do. That’s something Wayne Potts realized as a graduate student in the late 1970s. Now a biologist at the University of Utah, Potts ended up studying dunlins on Puget Sound. By making movies of their flocks and analyzing, frame by frame, how each individual bird moved, he was able to show that a turn ripples through a flock just as a cheerleading wave passes through sports fans at a stadium. He explained the finding with the name of his theory: the “chorus line hypothesis.” An individual dancer who waits for her immediate neighbor to move before initiating her kick will be too slow; similarly, a dunlin watches a number of birds around it, not just its nearest neighbors, for cues. This finding put to rest the old telepathy idea.
“The wave was propagating through the flock at least three times faster than could be explained if they were just watching their immediate neighbors,” says Potts. “But there was probably nothing extrasensory going on.”
Every year flocks of many thousands of starlings winter at large roosts in Rome. Smearing the dimming sky each afternoon, just before dusk, they fly in from the rural olive groves where they feed—faithful commuters in reverse, as Rachel Carson once wrote about birds’ predictable habits. Thousands coalesce and form dense spheres, ellipses, columns, and undulating lines, sequentially changing the shape of their flocks within moments. They exasperate many residents, who tire of the droppings they leave behind. Others love their elaborate displays.
“As they approach the roosts, the starlings are regularly attacked by falcons and display amazing flocking behaviors,” says Carere. “They compact and decompact, split and merge, form ‘terror waves’ ”—pulses that move away from an approaching falcon in a split second. “This is something that by sight is fantastic, like Indian smoke signals.”
In the coastal wetlands of southwestern Denmark, where some starling flocks in spring can number more than a million, locals term their late-afternoon displays “black sun” because they literally darken the sky. But the starlings in Rome are particularly convenient to study because one of their principal roosts is in a park between the city’s central railroad station and one of the branches of the Roman National Museum.
Researchers from a collaborative, pan-European project named StarFLAG logged a lot of hours on the roof of the museum’s historic Palazzo Massimo in two recent winters, aiming a pair of aligned cameras at flocks of many thousands of starlings performing aerobatic displays. Some researchers had previously used high-speed stereoscopic photography to analyze the structure of the whole, but they were able to do so only with relatively small groups. Once a flock exceeded 20 to 30 birds, its structure became impossible to tease apart. “You have to say who is who in the images from the different cameras, which look very different from one another,” says Andrea Cavagna, an Italian physicist working with StarFLAG. “This is very difficult to do by eye, and totally impossible for a thousand birds.”
By using software borrowed from the field of statistical mechanics, which explains properties of materials by examining their molecular structure, Cavagna and other physicists have now been able to match up to 2,600 starlings in different photographs with one another. That allows them to map the three-dimensional structure of flocks much more precisely than has ever been possible before. Onscreen, they can take what appears to the human eye as a solid, rounded mass of birds and learn whether it is in fact a ball or rather some other more complicated shape, such as a pancake, a column, or an open cup. They can view it from any angle, and watch it alter shape at 10 frames per second.
The result has been an infusion of quantifiable observation into a field long rife with speculation. By zooming in on the three-dimensional reconstructions, the researchers can begin to understand the spatial relationships individual starlings within it have with one another. They’ve found that however dense a flock appears from the outside, its members are not evenly distributed like points on a grid. Rather, each member has a good deal of space behind and in front. Like drivers on a freeway, starlings don’t appear to mind having neighbors nearby on their sides—or above and below, for that matter—as long as they have open space ahead.
That makes sense, since the presence of a clear path in the direction of travel minimizes the likelihood of collisions should the birds need to shift their course abruptly, as is likely when a falcon attacks. But what’s really nifty about this spatial asymmetry is that the researchers have been able to use it to calculate the number of neighbors to which each starling pays close attention—a quantified elaboration of Potts’s chorus line idea. By looking at correlations between the movements of neighboring starlings, they can show that each bird always pays attention to the same number of neighbors, whether they’re closer or farther away.
How many neighbors is that? Six or seven, says Cavagna, who points out that starlings in flocks can almost always see many more nearby birds—but the number may be closely tied to birds’ cognitive ability. Laboratory tests have shown that pigeons are readily able to discriminate between up to six different objects, but not more. That seems to be enough. Focusing on more than one or two neighbors enables a starling to maneuver quickly when needed. But by limiting to six or seven the number of neighbors it pays attention to, it may avoid cluttering its brain with less reliable, or simply overwhelming, information from birds farther away.
Whether watching those neighbors is all they do, though, is not yet known. Several StarFLAG collaborators at the University of Groningen, in the Netherlands, have been using these closely watched flocks to calibrate computer simulations more sophisticated than any others used before to analyze flock behavior. They’re trying to refine the models created by the physicists to more accurately reflect the real conditions starlings face, such as gravity and turbulent air. The researchers are also trying to understand how starlings in flight communicate; though everyone agrees that they use sight to navigate in close quarters, that may not be all they use.
“I think it’s acoustic and visual,” says Carere, “but the exact way it works no one knows.” He suggests that a starling may even use the tactile sense of onrushing air from close neighbors to help guide its direction. Clearly, there’s a lot still to be learned from these most mundane of birds.
Frank Heppner is confident that researchers will soon be able to explain many such mysteries, even as he continues to question some of the most basic assumptions about flocking behavior. He wonders, for example, why the Roman starlings so spectacularly maneuver above their roosting sites for many minutes before settling down. If they really wanted to avoid falcons, he asks, wouldn’t they disappear into the trees more quickly? “What they do is not predator avoidance,” he says. “It’s inviting predators.”
He speculates that there may be some fundamental math-based behavior going on—the kind of thing that physicists call an “emergent property,” in which the whole is much greater than the sum of its parts. Starlings may do what they do simply because their individual programming makes complex behaviors, like flocks, inevitable. Birders, of all people, ought to understand that, since they know how simple biological rules like a basic human interest in brightly colored, moving objects can lead to unpredictable and apparently irrational behaviors—such as jetting off to Brownsville to spot a golden-crowned warbler.
“It may be that these types of behaviors are like a mathematical by-product of the rules the birds follow,” Heppner says. “It is entirely possible that you get unpredictable behavior out of predictable rules.” Perhaps Rome’s starlings will yet shed some light on collective decision making by people.
Some scientists affiliated with the StarFLAG project are examining how voters affect one another’s choices, and whether decisions on where to locate new bank branches constitute a possible example of flocking behavior.
Such practical applications of understanding flock behaviors might be worth as much to some people as knowing the intentions of the gods. Yet they’re probably less valuable than an acknowledgement of how people have already affected flocks. Starlings did not winter in Rome in such numbers in years past, but climate change, combined with other factors, has made the city more comfortable for them. Flocks of many shorebirds are diminishing as their habitats and foods are altered. And it is due to us, of course, that no one can anymore enjoy the sight of one of the greatest of flocking species: the passenger pigeon.
The most quintessentially human behavior flocks reveal, though, may turn out to be the quest to both understand and enjoy them. People want to know how the world at large operates, but they also want to simply appreciate it. Those flashing dunlins, and those starlings whirling like swift black smoke, will remain a compelling sight no matter what the computer models postulate. At least in part, they’ll continue, as Richard Wilbur wrote, “refusing to be caught . . . in the nets and cages of my thought.”
This story originally ran in the March-April 2009 issue as "Flight Plan."