STEERING THE FLYING SWARM - Honeybee Democracy - Thomas D. Seeley

Honeybee Democracy - Thomas D. Seeley (2010)


Much have I marvelled at the faultless skill

With which thou trackest out thy dwelling-cave,

Winging thy way with seeming careless will

From mount to plain, o’er lake and winding wave.

—Thomas Smibert, The Wild Earth-Bee, 1851

Thomas Smibert, writing about bumblebees flying home “o’er lake and winding wave” in his native Scotland, lauded the marvelous ability of bees to return home after visiting distant flowers. His praise is richly deserved. We now know that a worker honeybee can navigate to and from flowers blooming 10 or more kilometers (more than 6 miles) from the hive, a thoroughly respectable distance for a creature only 14 millimeters (about half an inch) long. We also now know that bumblebees and honeybees find their way home using navigation methods like those used for ages by sailors making a passage over open water to reach a familiar harbor: steering according to a compass—for bees, this is the sun—and keeping a running tally of distance traveled, but then relying on memorized landmarks when within sight of the goal. How individual bees can range so widely without getting lost was one of the mysteries that Karl von Frisch and Martin Lindauer dug into most deeply in the 1950s, and since then other biologists have further revealed how a bee guides herself out to flowers and home to hive.

Meanwhile, the related mystery of how a swarm of bees steers itself to its new home was neglected, probably because it seemed a mind-boggling puzzle. Somehow, a school-bus-sized cloud of some ten thousand flying insects manages to sweep straight from bivouac site to new dwelling place. The path of its flight usually stretches for hundreds or thousands of meters (up to several miles), traversing fields and forests, hilltops and valleys, and swamps and lakes. Perhaps most amazingly, the airborne colony pilots itself over the countryside to one specific point in the landscape: a single knothole in one particular tree in a certain patch of forest. And as the group closes on its destination, it gradually lowers its flight speed so that it stops precisely, and gracefully, at the front door of its new home. How do ten thousand bees accomplish this magnificent feat of oriented group flight? In the past few years, with the introduction of digital video technology, it has become possible to perform the sophisticated data collection and image processing needed to track individual bees in a flying swarm and thus unravel the mechanisms of flight guidance in honeybee swarms. In this chapter, we will look at these mechanisms and we will see that the scout bees, yet again, play the leading role in our story.

Swarm Chasers

In the summer of 1979, I returned to my family home in Ithaca, New York, to work again with my first mentor and good friend, Roger “Doc” Morse, professor of apiculture at Cornell. A few years before, Doc and one of his students, Alphonse Avitabile, now professor emeritus at the University of Connecticut, had found that when a swarm of bees flies to its new home, the workers continuously monitor the presence of the queen within the swarm cloud by smelling the “queen substance” pheromone that wafts steadily from her body. This pheromone, the major component of the material secreted by the mandibular glands in the queen’s head, is a 10-carbon fatty acid whose exact name is (E)-9-oxo-2-decenoic acid (I will call it simply 9-ODA). If the bees in an airborne swarm keep smelling this particular chemical substance, they will keep flying toward their new home address, but if they don’t catch its aroma, because their queen has dropped out to rest, they will cease flying forward, mill about until they find their missing queen, and then cluster around her wherever she has alighted. Sooner or later, the swarm will again take of and proceed to its destination. Clearly, the workers in a flying swarm take great care to avoid losing their all-important queen.

To test whether 9-ODA is the critical indicator of the queen’s presence, Doc and Al Avitabile had conducted a slightly evil experiment. They had set up artifcial swarms so that each swarm’s queen was imprisoned in a small cage and then, when each swarm had finished its house hunting and was taking of to move to its chosen site, they had painted 9-ODA on the backs of five worker bees as they launched into flight. Each swarm treated in this way took of, few out of sight, and…never returned! Swarms that were treated identically except that no workers were painted with 9-ODA also completed the takeoff process, but they few of only about 50 meters (150 feet) before they returned and resettled around the caged queen. Clearly, the presence of a bee bearing the special fragrance of 9-ODA suffices to convince an airborne swarm that its queen is on board. To this day, I feel sadness for the orphaned swarms produced in this otherwise superb experiment.

In watching his 9-ODA-treated swarms fly out of sight, Doc became intrigued by how they conduct their flights, and in 1979 he invited Kirk Visscher (just starting graduate studies with Doc) and me to help him tackle this problem. Our first goal was simply to watch a swarm perform a flight, from start to finish. To do so, we went to Appledore Island, where we knew we could control a swarm’s flight path. We took with us a medium-size (11,000-bee) swarm, and we carefully positioned it and a nest box on the island so that we would be able to run beneath the swarm throughout its flight. The jungles of poison ivy on Appledore limited us to running along its roads and trails, none of which is straight, but we managed to find a sufficiently linear “track” 350 meters (1,150 feet) long that would enable us to stay close to our swarm throughout its journey. At one end we placed the swarm, at the other end the nest box, and every 30 meters (100 feet) in between a fagged stake. By noting when the center of the flying swarm passed over each stake, we could later calculate its speed during each stage of its flight.

As expected, a scout from our swarm soon found our nest box and the bees dancing for it quickly dominated the scout bees’ debate. While we were waiting for the scouts to finish their deliberations, we applied a dot of blue paint to every bee that performed a dance for the nest box, and we noted every five minutes the percentage of bees visible at the nest box that were marked with blue. Knowing that we painted 143 scout bees and finding that on average 29 percent of the scouts at the box were painted, we estimated that approximately 495 bees (143 = 0.29 × 495) had visited the nest box before takeoff. Thus we learned that fewer than 5 percent of the 11,000 bees in our swarm were familiar with its destination upon takeoff.

Equally interesting is what we learned about swarm flight speeds once the swarm became airborne. We saw that the cloud of swarming bees hung over the bivouac site for about 30 seconds, then began moving slowly off in the direction of the nest box. It covered the first 30 meters (100 feet) at less than 1 kilometer per hour (about 0.5 miles/hr) but accelerated steadily so that after 150 meters (500 feet) it had reached its top speed of 8 kilometers per hour (5 miles/hr). What was most surprising was the way that the swarm somehow managed to apply its brakes before it reached the nest box. Starting about 90 meters from the box, it gradually trimmed its speed and finally came to a halt with the center of the swarm cloud less than 5 meters (15 feet) from its goal. Over the next two minutes, the scout bees appeared in increasing numbers at the box’s entrance opening—5 after 20 seconds, 40 after 50 seconds, and over 100 after 90 seconds—releasing Nasonov gland pheromones to show the ignorant bees the way into their new residence. Within three minutes of the swarm stopping before the nest box, the bees were landing so heavily on it that they blanketed its front. Soon they were marching inside en masse, creating a whirlpool of bees that wheeled slowly around the entrance hole (fig. 8.1). The queen slipped in without fanfare six minutes later, and before 10 minutes had elapsed since the swarm’s arrival, nearly all the bees were safely inside their new home.


I took a liking to chasing swarms that day, but didn’t follow up on our observations until 25 years later, in the summer of 2004, when I had the immense good fortune of being joined by Madeleine Beekman, a behavioral biologist from the Netherlands. Madeleine had recently completed postdoctoral studies in England with my friend Francis Ratnieks, a noted bee expert, and had become intrigued by the mystery of swarm flight guidance. She joined me for a summer of swarm studies at Cornell and turned out to be the best possible collaborator: intelligent, hard working, and good-natured. She is now on the faculty of the University of Sydney in Australia.

The observational setup on Appledore Island had been rather crude and we looked for ways to improve it. We wanted to describe the flight behavior of swarms more precisely and to perform controlled experiments. We decided the way to do this was to fly swarms across the large meadow beside my laboratory at the Liddell Field Station, just of the Cornell campus. In the middle of this 26 hectare (65 acre) expanse of grass stands one large ash tree (Fraxinus americana) with spreading limbs, which provided a perfect place for hanging the nest box that we wanted our swarms to choose for their new home. Of course, there were attractive natural nesting cavities in the woods surrounding this field and beyond, but I had learned from my studies on Appledore that if I watched a swarm with infinite patience and plucked from it every scout bee performing a dance for some site other than my nest box, I could keep the swarm’s attention fixated on the nest site I was offering. This worked well. We few many swarms along a 270-meter (886-foot) flight path that ran from near the laboratory out to the ash tree. We divided this flyway into 30-meter (98-foot) segments so we could make measurements of swarm flight speed. Also, to make accurate measurements of the dimensions of swarm clouds upon takeoff, we created a 20 x 20 meter (66 x 66 foot) “launch pad” for our swarms. This was a closely mowed area that was gridded with stakes spaced 4 meters apart and in which we erected a 6-meter-tall pole with 1-meter markings. Each swarm was set up in the launch pad’s center, and upon takeoff we measured its cloud’s length and width with reference to the grid, and the heights of its cloud’s top and bottom with reference to the pole. We also photographed the flying swarms from the side for later analysis of the movement patterns of individual bees.


We began by watching the flights of three swarms, each containing approximately 11,500 bees, which is the median size of natural swarms. In launching itself into flight, each clustered swarm exploded into a buzzing cloud of bees some 10 meters (33 feet) long, 8 meters (26 feet) wide, and 3 meters (10 feet) tall. The bottoms of these swarm clouds swirled about 2 meters (6 feet) from the top of the meadow grass, hence (thankfully) a bit over our heads! Knowing these dimensions, we calculated that within each swarm cloud the flying bees were spaced, on average, only about 27 centimeters (10 inches) apart, which means that they were functioning at a density of some 50 bees per cubic meter (1.4 bees per cubic foot) (fig. 8.2). Amazingly, the bees rarely if ever suffered midair collisions.


The flight patterns of all three swarms matched what Doc, Kirk, and I had seen with our swarm on Appledore Island. Each swarm at first moved very slowly, then smoothly accelerated to a top speed of about 6 kilometers per hour (4 miles per hour), and finally braked gently, coming to a full stop at its new home (fig. 8.3). And as before, we noticed a brief delay between when each swarm had reached its destination and the moment its scouts began settling at the entrance hole and releasing Nasonov gland pheromones, but that once this chemical signal was being discharged, the rest of the swarm cloud quickly settled on the nest box. The bees didn’t hesitate to move inside and within 10 minutes nearly all had disappeared indoors. Each swarm executed the entire migration process—takeoff, flight, landing, and entry—with precision and so completed it in less than 15 minutes.

Leaders and Followers

What makes the precisely oriented flight of a honeybee swarm to its new home so wondrous is that only a small percentage of its members know the swarm’s travel route and final destination. As mentioned already, less than 5 percent of the swarm that Doc, Kirk, and I studied on Appledore Island had visited the nest box, and so knew its exact location, before the swarm made its flight to its new home. This finding was later confirmed by Susannah Buhrman and me in the study in which we prepared swarms of individually labeled bees, video recorded the scouts’ dances, and determined which proposed nest site each scout advertised. In all three swarms that we studied, only 1.5 to 1.7 percent of the bees performed dances for the chosen site. This figure, combined with the figure of approximately 50 percent for the fraction of the scout bees from a high-quality site that advertise it with a dance (what Kirk and I found in our study of how scouts encode nest-site quality in their dances; see chapter 6), yields the estimate that only 3 to 4 percent of the bees in each swarm had visited the chosen site and so knew the exact location of their swarm’s future dwelling place. Clearly, when a swarm flies to its new home, it relies on a relatively small number of informed scout bees—approximately 400 individuals in an average-size swarm of 10,000 bees—who function as guides or leaders of all the rest. How does this system of leaders and followers work?

Three hypotheses have been proposed for how an informed minority guides the ignorant majority when a swarm flies to its new home. The first hypothesis suggests that the transfer of information from leaders to followers relies on a chemical signal. In their 1975 paper on how the workers in a swarm perceive the presence of their queen by sensing the 9-ODA she releases, Al Avitabile, Roger Morse, and Rolf Boch proposed that scouts provide flight guidance by means of the attraction pheromones produced in the Nasonov gland that is part of the scent organ at the tip of a worker bee’s abdomen (fig. 8.4). Their idea was that scouts might discharge these pheromones along the front of the swarm cloud to attract, and thereby guide, the nonscouts to move in this direction.


The other two hypotheses suggest that the information transfer from informed bees to ignorant bees works by vision instead of olfaction. One hypothesis, called the “subtle guide hypothesis,” was proposed in 2005 by a team of biologists from Princeton University in the United States and the universities of Leeds and Bristol in the United Kingdom: Iain Couzin, Jens Krause, Nigel Franks, and Simon Levin. According to this hypothesis, the informed bees do not conspicuously signal the correct travel direction; instead, they steer the swarm simply by tending to fly in the direction of the new home. By making computer simulations of airborne swarms, the authors showed that if each bee in a flying swarm (1) attempts to avoid collisions by turning away from any neighbors within a critical distance, (2) tends to be attracted toward and aligned with neighbors outside the critical distance, and (3) flies either with a preferred movement direction (informed bees) or without a preferred movement direction (ignorant bees), then the swarm will be steered toward its new home even if the proportion of informed individuals is very small. Remarkably, in large groups, like honeybee swarms, this proportion can be less than 5 percent. It is an intriguing hypothesis, for it shows that the bees in a flying swarm might not need to know which of them know the travel route, hence are the leaders.

The second vision-based hypothesis, called the “streaker bee” hypothesis, was sketched out in 1955 by Martin Lindauer. At the very end of his magnum opus on house hunting by honeybees, Lindauer reported having seen “that several hundred bees, in more rapid flight, always shoot forward toward the front of the swarm cloud, that is to say in the direction of the nest site. While the swarm cloud slowly continues its flight in this direction, these guiding bees slowly fly back along the border of the swarm cloud and again shoot to the front in rapid flight.” The streaker bee hypothesis suggests that the informed bees conspicuously signal the correct travel direction by repeatedly making high-speed flights through the swarm cloud. (Note: according to the subtle guide hypothesis, the informed and ignorant bees fly at the same speed.) In the streaker bee hypothesis, the ignorant bees behave as is suggested for the subtle guide hypothesis except for one thing; rather than align themselves with neighbors in general, the ignorant bees preferentially align themselves with fast-flying neighbors. So the two key differences between the subtle guide and streaker bee hypotheses are whether or not the informed bees (leaders) point the way with high-speed flights and whether or not the ignorant bees (followers) favor alignment with fast-flying bees. Computer simulations, similar to those done for the subtle guide hypothesis, have shown that the streaker bee hypothesis is a plausible mechanism of swarm flight guidance. Thus both the subtle guide and streaker bee hypotheses for swarm flight guidance are a possibility. The burning question is which, if either, is the reality.

Scent Organs Sealed Shut

After describing the flights of swarms across the meadow at the Liddell Field Station, Madeleine Beekman and I set ourselves the goal of testing the hypothesis that scout bees guide their swarm using the attraction pheromones produced in their scent organs. We knew we had to prepare swarms in which every worker had her scent organ sealed shut and then see if these swarms could perform a well-oriented, full-speed flight down our flyway at Liddell.

The scent organ of a worker honeybee lies on the dorsal surface of the abdomen, at the front edge of the last abdominal segment. It consists of several hundred gland cells (the Nasonov gland, named after the Russian scientist who first described it, in 1883) whose ducts open onto the membrane that connects the last two plates (“tergites”) covering the top of the abdomen (fig. 8.5). The secretion—which consists mainly of citral, geraniol, and nerolic acid and has a pleasant lemony aroma—collects on this membrane. Usually this area is concealed by the overlap of the two tergites, but if a worker bee bends the apical segment of her abdomen downward, she exposes the membrane and releases the scent. Using a fine paintbrush, one can paint over the joint between the last two tergites, and when the paint dries the two tergites will be stuck together so that the treated bee can no longer expose her scent organ.


We tried various paints. The first ones tended to crack easily so that after a few days many bees were able to release the balmy Nasonov gland pheromones from the scent organ, but eventually we found that Testors gloss enamel paint makes a lasting seal of a bee’s abdominal segments. At this point we adopted a standard procedure for preparing our test swarms. After immobilizing groups of 10 to 20 bees in plastic bags in a refrigerator, we put one group of these chilled bees at a time on ice, painted over each bee’s scent organ, and then poured the still immobile bees into a screened cage with their queen where they would warm up and become part of the artificial swarm we were creating. We did this over and over until we had 4,000 bees properly painted, enough for a small “treatment” swarm. To control for any effects of chilling, painting, and handling, we also prepared 4,000-bee “control” swarms in which we did everything the same except that we applied the dot of paint to each bee’s thorax instead of her abdomen.

We eventually few six swarms, three treatment swarms, and three control swarms. The two types, when airborne, formed similar-sized clouds of flying bees (8 meters long, 8 meters wide, and 3 meters high). Most importantly, both treatment and control swarms few directly and quickly to the nest box! As we had seen previously with the large swarms, these small swarms accelerated steadily for the first 90 meters, reached peak flight speeds after flying 90 to 120 meters, started slowing down after flying 210 to 240 meters, moved very slowly during the final 30 meters, and finally stopped at the nest box. The maximum speeds of the treatment swarms were 6.8, 3.6, and 6.8 kilometers per hour, while those of the control swarms were 6.7, 6.4, and 7.2 kilometers per hour. (The second treatment swarm few more slowly than the others because it few against a fierce headwind, whereas all the rest encountered at most a slight breeze.) There was, however, one important way in which the two types of swarms did behave differently: once they reached the nest box, the treatment swarms took much longer than the control swarms (20 minutes versus 9 minutes, on average) to move into the box. Why? Almost certainly, it was because the scouts in the treatment swarms were unable to help the nonscouts find the entrance to their new home by marking it with Nasonov gland pheromones. They certainly tried hard to do so. The scouts landed at the three entrance holes in the nest box and stood there flamboyantly with abdomens elevated and wings whirring, but they could not bend down the last abdominal segment to expose the scent organ (fig. 8.6). (To be sure about this, we inspected 250 bees from each swarm shortly after it entered the nest box and found that only a miniscule percentage of the bees, less than 1 percent, had cracked paint seals.) Because our treatment swarms executed their flight plans as flawlessly as our control swarms, except during the landing phase when the scent organ clearly plays an important role, we concluded that the informed scouts don’t provide their ignorant sisters with flight guidance information using the Nasonov gland pheromones.


Streams of Streakers

Madeleine and I next started to test the streaker bee hypothesis. We believed that we had seen, while watching our swarms sweep across the meadow to the nest box, what Lindauer had reported seeing in an airborne swarm. Most bees fly about within the cloud in rather slow and looping flights, but a few shoot straight through the cloud in the direction of the swarm’s new home. Also, it looked to us like the streaking bees zoomed mainly through the top of the swarm cloud. But we were not 100 percent confident of our sightings and certainly we had no hard data, so we decided to try to use conventional still photography to get solid information and check our impressions. Using a 35 mm camera, color transparency film with a slow film speed (DIN 64), and a moderately long exposure time (one thirtieth of a second), we found that if we photographed a flying swarm from the side and under a clear sky, we could get photos that “captured” the entire swarm cloud and in which individual bees appeared as small, dark streaks on a bright background (see fig. 8.2). The length of each bee’s streak indicated her flight speed, and the tilt of her streak indicated her flight angle relative to horizontal, the orientation of level flight. These photos showed unambiguously that a small minority of the bees in an airborne swarm do whiz through it at the maximum flight speed of a worker bee—about 34 kilometers per hour (20 miles per hour)—and that all the rest of the bees buzz along much more slowly. We also found that the streaks of the fast-flying bees, compared to those of the slow-flying ones, are more apt to be horizontal, indicative of straight and level flights. Finally, we gleaned from the photos the finding that the speeding bees, the streakers, do indeed operate mainly in the top of the swarm cloud. This makes sense, assuming that these bees are providing flight direction information to the other bees, for by streaming over their sisters the fast-flying bees position themselves where they are easily seen against the bright background of the sky.

Computer Vision Algorithms for Tracking Bees

The photographic study that Madeleine and I made in 2004 gave support to the streaker bee hypothesis, but it was not a rigorous test between this hypothesis and the subtle guide hypothesis. This is because our photographs, taken from the side of a flying swarm, could not tell us the flight directions of the fast-flying bees (whether toward the new home site, away from it, or some angle in between). And the key to resolving the subtle guide and streaker bee hypotheses is knowing whether or not the flights of the speedsters in a swarm point mainly toward the swarm’s new home. The two hypotheses make distinct predictions about this matter. The subtle guide hypothesis predicts that the fast-flying bees will not be heading mainly in the direction of the new homesite because, according to this hypothesis, the informed bees don’t signal the travel direction with high-speed flights through the swarm cloud. In contrast, the streaker bee hypothesis predicts that the fast-flying bees will be heading mainly in the direction of the new homesite because, according to this second hypothesis, this is how the informed bees share their knowledge of the swarm’s travel direction. Some of the speedy bees will be informed bees indicating which way to go and probably some will be ignorant bees reacting to the informed bees.

In 2006, when it became possible to track individual bees in a flying swarm and measure each bee’s position, flight direction, and flight speed, it became clear that the high-speed fliers in a swarm are indeed zipping toward the swarm’s new home. So now it seems clear that the streaker bee hypothesis is correct. The two people who were instrumental in developing the tools for tracking individual bees in an airborne swarm are Kevin Passino, professor of electrical and computer engineering at Ohio State University, and his brilliant graduate student, Kevin Schultz.

One of the great benefits of the academic life is that it gives you opportunities to visit other universities and meet remarkable people, some of whom share your intellectual excitement about a particular mystery. I met Kevin Passino on a trip to Ohio State University in the spring of 2002. I was there on a lecture engagement, not to enlist an engineer for collaborative work, but upon meeting Kevin I sensed he was a marvelous engineer with precisely the right inclination for a joint scientific venture. Here was someone who devised automatic control systems for technological applications, but who also liked to look at biological systems for inspiration. I was to learn later that “biomimicry” is a hot approach among control engineers, for the methods of automatic control in living organisms are exceptionally powerful and robust, having been tested and tuned by natural selection for millions of years. As I recall, the outcome of our first meeting was an agreement that we would team up. Already we had found splendid common ground on the puzzles of forager force allocation and swarm flight guidance. In the words of Kevin, the honeybees have evolved a “cooperative control strategy for groups of autonomous vehicles,” and he was keen to join in exploring it.

Once Madeleine Beekman and I had disproved the pheromone hypothesis for swarm flight guidance and had published the results of our simple photographic analysis that confirmed the existence of streaker bees in swarms, Kevin realized that what was needed next was to video record an airborne swarm from below using a high-definition video camera. He had a hunch that by tapping into the latest video technology, especially point-tracking algorithms invented by engineers working on computer vision, it should be possible to track individual bees as they few over the video camera and determine for each bee her position in the swarm cloud, her flight speed, and her flight direction. So Kevin purchased the necessary camera and joined Kirk Visscher and me during our field trip to Ap-pledore Island in the summer of 2006. We set up a swarm by the old Coast Guard building at the island’s center, placed an attractive nest box on the eastern shore 250 meters (820 feet) away, and charged up the camera’s battery. Our goal was to video record the swarm as it few over the camera at two points along its flight path: 15 meters (50 feet) from the bivouac site, when the swarm had just taken of and was still moving slowly, and 60 meters (200 feet), when the swarm was well under way and had picked up considerable speed. The camera was equipped with a wide-angle lens, so its field of view included most of the width of the swarm cloud, though not its entire length. The camera also had an extremely high shutter speed—one ten thousandth of a second—so in each frame of the video recording a bee appeared as a short blob, not a long streak. Our biggest obstacle to success that summer was the wind sweeping over Appledore Island, which made it difficult to get swarms to fly directly over our camera. In still air a swarm will course along a predictable beeline to its new home, but in windy air the track of a swarm’s flight is wildly unpredictable as chaotic gusts push the airborne bees about, knocking them of a direct line of travel. And Appledore Island, anchored in the Atlantic Ocean six miles of the southern coast of Maine, is thoroughly windblown. It is so much so that in 2007 the Shoals Marine Laboratory erected on the island a 27.5-meter (90-foot) tall wind turbine to harvest some of the wind’s energy. Now the laboratory acquires a substantial portion of its electrical power from this limitless source. On June 29 and July 2, 2006, however, we were blessed with two days of wonderfully calm air, and twice we got a swarm to fly directly over the video camera at both the 15-meter mark and the 60-meter mark along the line running straight to the bees’ new home.

With these two sets of recordings of swarm flyovers “in the can,” Kevin Passi-no had material for his PhD student, Kevin Schultz, to tackle. Over the next two years, Kevin S. created a computer algorithm that semiautomated the data-gathering process. In essence, the procedure involves examining each ellipsoidal blob (bee image) in a given video frame, noting its orientation (the angle between the major axis of its ellipse and the bottom edge of the video frame), and then pairing it up with the blob on the next video frame that represents the same bee. The pairing process involves finding for a given blob in the first frame the blob in the second frame that best matches its position and orientation. This process is repeated with the blobs of the second frame being paired with blobs of the third frame, and so on, to build up, frame by frame, detailed trajectories of individual swarm bees as they few across the video camera’s field of view back on Appledore Island. The size of a blob—the length of the major axis of a blob’s ellipse—indicates the height of the bee above the camera. So the bees in the top and bottom portions of the swarm cloud were distinguished, and it was even possible to make three-dimensional reconstructions of the individual bees’ fights. What a tour de force!

It is hard to convey in words what it is like to go from watching thousands and thousands of swarm bees swirling over head in seeming random motion, to seeing graphs that show wonderfully clear patterns in their movements. Every detail of the swarm bees’ collective motion was a complete revelation, for before Kevin P. and Kevin S. devised their process of blob tracking on digital video recordings, no one could even begin to see these patterns. The human visual system is a stupendous biological computer, capable of amazing feats of information processing such as instantly recognizing a face not seen in years, but even it is overwhelmed by so many bees moving so rapidly and so wildly.

The most important pattern revealed by the analysis of the video recordings is that the fast-flying bees were indeed streaking in the direction of the chosen homesite. Looking at figure 8.7, which shows the individual bees’ flight speeds in relation to flight direction, we can see that the speediest bees were zooming directly toward the new home, while the slowest ones were heading in the opposite direction. By comparing the plots for the top and bottom portions of the swarm cloud, we can also see that the speedsters were mainly in the top portion of the swarm, confirming what Madeleine and I had detected with our side-view photographs of moving swarms. A third important feature of the plots shown in figure 8.7 is that the peaks in flight speed are higher at the front section of the elongated swarm cloud than at its rear (true for both the top and the bottom of the swarm cloud). This shows that the fastest bees tended to be in the front of the swarm. A painstaking analysis by Kevin S. of the velocities of individual bees found that not only do bees that are flying in the direction of the nest site tend to fly with the highest velocities, but also they tend to accelerate (increase their velocities) as they move from swarm rear to swarm front. It seems likely that some of this rise in flight speed comes about as the ignorant/follower bees “latch on” to the informed/leader bees, boosting their speed as they chase after the streaker bees. If so, the information about flight direction (expressed in the flight direction of the fast-flying bees) and the boosting of flight speed are likely to spread from the informed bees to some of the ignorant bees who, through their own faster flights, will start to influence other ignorant bees. This chain reaction of informed/leader bees begetting more informed/leader bees could lead to a widespread induction of bees to fly toward the nest box and to fly faster. This may explain the increase in overall swarm speed over time that is shown in figure 8.3, and that is so impressive to any beekeeper who tries to follow a fugitive swarm to its new home by running along beneath it.


The discovery that the bees flying toward the new home are traveling far faster than the other swarm bees led Kevin Passino, Kevin Schultz, and me to conclude that streaker bees, not subtle guides, appear to provide the flight guidance to an airborne swarm of honeybees (fig. 8.8). We would like, however, to test the streaker bee hypothesis more rigorously by performing an experiment analogous to the sealed-scent-organ test of the attraction pheromones hypothesis: block the proposed means of guidance and see if this renders a swarm incapable of making a well-oriented, full-speed flight to its chosen destination. Unfortunately, nobody has succeeded yet in figuring out how to prevent the informed bees from performing high-speed flights. Madeleine Beekman tried trimming the wingtips of scout bees by a millimeter or so, a manipulation that is known to reduce a bee’s maximum flight speed, but she found that this surgery also caused her bees to stop scouting. Maybe some other approach will do the trick. Glue small airfoils or short strings to scout bees to increase the drag they experience during flight? Find bees that have a genetic mutation that causes them to fly slowly? Anyone who figures out a way to prevent streaking by scout bees will have set the stage for a beautiful experiment.

In the meantime, Madeleine Beekman and two students, Tanya Latty and Michael Duncan, have succeeded with a different approach to testing the streaker bee hypothesis. They performed an ingenious experiment in which they caused numerous fast-flying forager bees to zoom through an airborne swarm in a direction perpendicular to the swarm’s intended flight path (fig. 8.9). If the streaker bee hypothesis is correct, the foragers sweeping in from the side should create conflicting directional information in the swarm and thereby disrupt its flight guidance. This is exactly what was found. Of six test swarms that attempted to fly to a nest box 100 meters (330 feet) away when foragers were zipping back and forth across the swarms’ flight path, only one reached the nest box intact, and even this swarm was knocked temporarily of course. The other five swarms all started out, as usual, by moving straight toward the nest box, but upon hitting the “forager highway” they either fragmented or veered widely of course. In contrast, when the experimenters few four control swarms—ones identical to the test swarms but without the cross traffic of foragers—they saw each one stay together and fly directly to the nest box. Clearly, the heavy traffic of foragers crossing the flight path of the test swarms disrupted their mechanism of flight guidance, evidently by injecting misleading visual information into the clouds of swirling bees.



Assembling the Flight Navigators

There are many questions left unanswered about the remarkable flights of honeybee swarms. How does the moving group apply the brakes when it arrives within about 100 meters (330 feet) of its new residence? Also, how do the informed bees make their repeated streak flights through the swarm cloud? Do they stop when they reach the front and let the other bees fly past, or do they fly inconspicuously rearward along the swarm’s bottom, where they may be nearly invisible against the underlying vegetation? There is also the mystery of how a swarm makes sure that it is fully stocked with bees who know its flight route before it begins its journey to the new home.

It is striking how virtually all the scout bees who have visited the chosen home-site, and so can steer the airborne swarm to it, abandon the future dwelling place and assemble on the swarm cluster shortly before it launches into flight. I first witnessed this phenomenon in August 1974, shortly before I started graduate school, when I first watched a swarm go through its house-hunting process. I had set up an artificial swarm and a nest box behind my parents’ house, in an abandoned field dominated by blooming goldenrod plants (Solidago spp.) and young white pine trees (Pinus strobus), one of which supported my box for the bees. To my great good fortune, the scout bees chose my humble plywood box for their future home. Soon I was dashing back and forth along the 150-meter-long (500-foot-long) path between swarm cluster and nest box, doing my best to watch both the growing party of excited dancers on the swarm and the strengthening throng of scout bees scrutinizing the nest box. Part way through the afternoon, I was shocked to see a sudden drop in bees at my box. On my previous visit, 15 minutes earlier, I had counted some 25 bees examining the box, but now I saw only two or three bees, and in a few more minutes the place was totally deserted. This collapse in the scout bees’ interest baffled me, until I glanced back toward the swarm’s bivouac site and saw my swarm, now a diffuse ball of swirling and shining bees, “rolling” straight toward me over the sunny field. Evidently, the scouts had abandoned the nest box to be in the swarm cluster at its moment of departure.

Since then, when performing experiments on swarms, I have come to rely on the conspicuous drop in scouts at the nest box as a reliable indicator that the swarm has completed its decision making and is about to take of (see figs. 5.5 and 5.7). It certainly makes sense for the scouts to assemble at the swarm shortly before departure, for we have seen how only 3 to 4 percent of a swarm’s bees know its flight plan, and with such a small minority of navigators it is probably important to have as many as possible on board. But precisely how this is achieved remains a mystery. Does the gathering of the scout bees on the swarm arise simply by these bees returning to the swarm as usual and then lingering there when they detect one of the flight initiation signals, either worker pipings or buzz-runs? Or might their assembly on the swarm be triggered by them hearing, feeling, seeing, or smelling an unknown signal of impending departure that is produced at the nest box? I wouldn’t be surprised if the bees possess some secret gadgetry for ensuring that a swarm about to take flight is well stocked with the informed bees who can pilot it safely to its new home.