Honeybee Democracy - Thomas D. Seeley (2010)


And so doth this soft shivering passe

as a watch-worde from one to another,

untill it come to the inmost Bees:

wherby is caused a great hollownes in the pomgranat.

When you see them do thus,

then may you bid them farewel:

for presentlie they begin to unknit,

and to be gone.

—Charles Butler, The Feminine Monarchie, 1609

Anyone who has the immense good fortune of watching a honeybee colony cast a swarm will be treated to many astonishing displays of animal behavior. First there is the feverish rush of thousands of bees out of the hive and up into the sky. Minutes later, the cloud of swirling, swarming bees mysteriously condenses into a tight crowd hanging from a tree branch, where for several hours or several days nearly all the bees sit quietly, almost motionless. Only the swarm’s scouts remain active, flying to and from the swarm cluster and performing their eye-catching dances on its surface to advertise candidate nest sites. Next, after one of these sites becomes the unanimous choice of all the dancing bees, comes the most wondrous sight of all: suddenly, in about 60 seconds, the entire swarm cluster disintegrates and takes flight, filling the air with the roar of thousands of airborne bees (fig. 7.1). This flying mob immediately begins moving of in the direction of its chosen home and in another minute or two it will have vanished. As Charles Butler expressed it so nicely back in 1609, you may now “bid them farewel.”


In this chapter we will look at how an entire swarm of bees manages to leave its bivouac site together and at the right time. With few exceptions, a swarm’s tightly synchronized takeoff occurs only after its scouts has observed finished their job of choosing the new dwelling place. This means that as we review the mechanisms of social coordination during swarm departure, we will be seeing how a swarm maintains its coherence as it switches its mission from making a decision to implementing a decision. Probably it will be no surprise to learn that the scout bees are the rabble-rousers who initiate a swarm’s journey to its new home, thereby extending their leading role in our story into the chapters on how swarms take action. But what will be surprising are the nifty signals the scouts use to animate their drowsy swarm-mates and the way the scouts know when the time has come to initiate their swarm’s journey. Indeed, until recently, these were deep mysteries about the inner workings of honeybee swarms.

Preflight Warm-Up

In the spring of 1980, Bernd Heinrich, a gifted insect physiologist at the University of California at Berkeley (now at the University of Vermont), turned his attention to the mechanisms of temperature regulation in honeybee swarms. Over the previous 20 years Heinrich had pioneered the study of temperature control in insects, so he began his study of honeybee swarms with a great deal of background knowledge. He knew that two previous studies had reported that the temperature inside a swarm cluster, like that inside a beehive, can be kept by the bees at about 35°C (95°F), nearly the same as the core body temperature of a human. He also knew that an individual worker honeybee can produce heat by shiv-ering—isometrically contracting the two sets of flight muscles in her thorax—and that her flight muscles must be warmed to at least 35°C (95°F) to produce a sufficiently high wing-beat frequency (nearly 250 beats per second!) to generate the lift needed for bee flight. Furthermore, Heinrich knew that before leaving the parental nest, swarming honeybees stuff themselves with honey, so that a swarm starts out with a sizable but finite supply of fuel for warming itself, for powering the scout bees’ flights to and from the swarm, and for building the first beeswax combs in the swarm’s new home. What he did not know were the exact pattern of temperatures inside a swarm cluster, how the bees control these temperatures, and how they manage their energy supply. Being a hobby beekeeper and curious about bees, Heinrich worked with the police and fire departments of Walnut Creek, California, as well as the Ecohouse Swarm Hotline of Walnut Creek, to collect 14 natural swarms in the San Francisco Bay Area during May and June. Back at his laboratory on the UC-Berkeley campus, he studied these swarms using various scientific tools, including tiny electronic thermometers (thermocouple probes) and a special cylindrical chamber made of Plexiglas (respirometry vessel) in which he could place a swarm to measure its metabolic rate at various ambient temperatures.


Heinrich discovered many marvelous things about temperature regulation in honeybee swarms, all of which are key to understanding how a swarm prepares to fly to its new home. First, he found that a swarm does indeed precisely control the temperature of the cluster’s core so that it stays at 34–36°C (93–97°F) regardless of the ambient temperature. He also found that a swarm allows the temperature of the cluster’s mantle (outer layer) to vary with the ambient temperature, but that it keeps the mantle temperature above 17°C (63°F) even if the ambient temperature falls to freezing (0°C or 32°F). This means that the outermost bees, which are the coolest, keep themselves warm enough to stay active on the swarm. If they were to cool below 15°C (59°F) they would enter “chill torpor” and easily fall from the swarm. They would also be too cold to warm themselves back up by shivering.

When Heinrich looked at how the bees achieve their characteristic pattern of temperatures in a swarm, he found that they do so without expending much of their on-board energy supply, that is, the honey in their stomachs. At air temperatures above about 10°C (50°F), the resting metabolism of a swarm—the metabolism that occurs when the flight muscles of the swarm bees are not being ac-tivated—provides more than enough heat to keep a swarm’s core at 35°C and its mantle above 17°C. Indeed, at high ambient temperatures (above about 20°C or 76°F), the resting metabolism produces so much heat that both the mantle bees and the core bees spread themselves out, creating ventilation channels to release excess heat from the core. But when the ambient temperature falls below 17°C, and the mantle bees start to feel too cool, they crowd inward, causing the swarm cluster to shrink, its porosity to decrease, and its heat loss to diminish (figs observa 7.2 and 7.3). In this way the mantle bees skillfully trap inside the swarm cluster the metabolic heat generated by the thousands of resting, immobile bees, and they also keep themselves sufficiently warm. It is only when the air temperature falls below about 10°C (50°F) that the mantle bees must take the extra step of raising their metabolic rate by shivering.

Thus Heinrich discovered that the bees in a honeybee swarm have an effec-tive means of conserving their energy reserves. The mantle bees, those most exposed to low temperatures, minimize their need for active metabolism by doing two things when the air becomes cool: (1) letting their body temperatures drop to just above the chill-torpor temperature rather than working to maintain a higher body temperature, and (2) keeping their body temperatures above the chill-torpor temperature mainly by huddling rather than shivering. Of course, these energy conservation measures mean that most of the time the outermost bees in a swarm are too cold to fly, something that is easily demonstrated by skimming a spoonful of mantle bees from a swarm and shaking them into the air. The bees tumble to the ground rather than fly away. So before a swarm can take of to fly to its new home, the cool bees in the mantle must warm their flight muscles to the fight-ready temperature of 35°C. And not just in theory! When Heinrich made continuous recordings of the temperatures at various locations in a swarm cluster from when the bees settled to when they departed, he found that during the last hour or so before takeoff, the temperature in the mantle did indeed rise to match the 35°C of the core.


In June 2002, some 20 years after Bernd Heinrich published his insightful report on “The Mechanisms and Energetics of Honeybee Swarm Temperature Regulation,” I traveled to Germany to look more closely at the preflight warm-up of swarm bees. Shortly before, I had the great good fortune of receiving a Research Prize from the Alexander von Humboldt Foundation, which gave me the wherewithal to conduct research projects in Germany. I was hosted by my teacher and friend, Bert Hölldobler, who had become the director of the Institute for Behavioral Physiology and Sociobiology at the University of Würzburg. This institute includes a laboratory devoted to research on honeybees. It was started by Martin Lindauer when he was professor of zoology at Würzburg (1973–1987), and is now directed by another good friend, Jürgen Tautz. Jürgen is highly skilled at studying the sensory abilities of insects, and his laboratory is stocked with much state-of-the-art scientific equipment for probing the workings of nature. On this trip I was keen to collaborate with Jürgen to use one particularly powerful instrument: an infrared video camera. With it, one can measure the temperatures of many objects (such as bees) simultaneously and without disturbing them. Also in Jürgen’s laboratory were two superb graduate students, Marco Kleinhenz and Brigitte Bujok, both experts in using the video camera and the computer software that converts the camera’s images into accurate temperature readings. The goal of our four-person team was simple: explore how the mantle bees in a honeybee swarm warm their flight muscles prior to takeoff.

The plan of using video thermography to see how the outermost bees prepare for takeoff worked nicely. Over a two-week period, we recorded the temperatures of mantle bees within a 10 x 10 centimeter (4 x 4 inch) area on two swarms, starting when each swarm formed its cluster and continuing until it launched into flight. Both swarms showed the familiar set of events shortly before takeoff: the scouts became unanimous in their dancing and the nonscouts began to move excitedly. Both swarms also revealed something new in the images recorded with the infrared video camera (fig. 7.4): the thoraces of all the bees on the swarm’s surface began to glow with unusual warmth just moments before the swarm’s explosive takeoff.


The finding that most captured our attention was the way that the percentage of bees with a thoracic temperature of at least 35°C rose exponentially over the final half hour before takeoff. As is shown in figure 7.5, for the first 20 minutes the percentage of the surface-layer bees with thoraces warmed sufficiently for flight rose slowly and remained below 20 percent. Then, starting about 10 minutes before takeoff, the percentage of hot bees began to rise faster and faster. Soon 100 percent of the surface-layer bees had a thoracic temperature of at least 35°C, and at exactly this moment the swarm bees took wing. We are confident that at the start of a swarm’s takeoff all of the bees in the cluster, not just the outermost ones, are hot enough for rapid flight. After all, Heinrich’s work had shown that the bees in a swarm cluster’s core are warm enough for flight at all times. Also, the images from our infrared video camera showed that as our two swarms approached the moment of takeoff, the interior bees began to shine brightly before the surface bees did, looking like hot coals glowing beneath a layer of cool ashes. There is also the fact that immediately after both takeoffs had started—that is, when the outermost bees had taken flight—the inner bees began to take flight. Indeed, it is because there is so little delay between outer and inner bees taking flight that a swarm cluster needs only about 60 seconds to disintegrate.


What stimulated the mantle bees to warm up, and how was it that each swarm’s takeoff began just seconds after all of the surface-layer bees had warmed their flight muscles to at least 35°C? In other words, what stimulated the bees to prime themselves for flight, and what finally triggered them to take flight? We will now probe these two mysteries.

Piping Hot Bees

If you listen closely to a swarm, by carefully placing an ear beside the massed bees, you will hear pulses of a distinctive, high-pitched piping sound starting about an hour before the swarm flies off to its new home. Each sound pulse lasts about a second, and because its pitch sweeps upward, it resembles the rising en gine whine of a Formula One race car making a quick acceleration. At first one hears these shrill piping sounds only occasionally, because just one bee at a time is producing them, but over the last half hour before takeoff more and more bees start piping and the pulsing hum radiating from the swarm rises to a crescendo. When it does, the swarm cluster breaks up and all the bees take wing. Could this high-pitched piping be a signal from the scout bees to their quiet swarm-mates with the message, “Ladies, warm your flight muscles!”?

To begin to explore this possibility, I wanted to identify which bees in the swarm cluster were producing the high-pitched piping sounds. Actually, this was a long-standing goal. I had first heard these mysterious sounds way back when I began studying swarms as a graduate student in the 1970s, but I never could pinpoint which particular bees, among the thousands in a swarm, were their source. Finding the piping bees was especially difficult because the pulsing hum seemed to emanate from inside the swarm cluster, thus from bees out of sight. The piping bees also stymied Martin Lindauer in the 1950s, for he wrote, “Now a hundredfold high humming could be heard at the cluster, but I could not definitely find out whether this comes from the buzz-runners or from other bees.” (The “buzz-runners” mentioned by Lindauer will be discussed later in this chapter.)

The discovery of who does the piping came serendipitously in the summer of 1999. It started with a chance observation I made at my camp beside Ox Cove, in easternmost Maine, the wonderfully secluded site to which I had retreated to figure out how the dissent among dancing scout bees expires (see chapter 6). I can still remember as if it were yesterday witnessing for the first time a worker piping on a swarm. I had set up a swarm outside my cabin, labeled with paint dots the first few dancers (scouts) on the swarm, and was watching steadily my little band of brightly painted bees, recording their behaviors. On August 2, at 10:48 a.m., just five minutes before the swarm flew away, my attention was drawn to the scout bee Blue, who did something unexpected on the swarm’s surface: she ran excitedly over other bees for a few seconds, then paused for about a second, pressed her thorax against a stationary bee, and then ran on, repeating the sequence of run-pause-press six times before she burrowed into the cluster and disappeared (fig. 7.6). I noticed that each time my bee Blue paused and grabbed another bee, she drew her wings tightly together over her abdomen and then her wings seemed to vibrate slightly. Was Blue producing the piping sound? I could hear the sound, but with just my “naked” ears I could not be sure the sound came from her. So that afternoon, I drove to Morgan’s Garage in the nearby village of Pembroke and bought a 3-foot length of rubber vacuum hose about 6 millimeters (one-quarter of an inch) in diameter, a size that would fit snugly in my ear. This simple sound tube would enable me to localize the source of sounds coming from my swarms, for it would function like a primitive stethoscope, conducting to my ear only sounds produced near its open end. A few days later, when I watched a second swarm and used my rubber hose to listen in on another painted scout bee doing the run-pause-press maneuver, I was thrilled to hear the perky piping sound.


I was fascinated by the sights and sounds of the piping worker bees, and was keen to describe their signal in detail and to test the idea that they are alerting the quiescent members of the swarm to warm their flight muscles for takeoff. This would require a sophisticated sound analysis combined with careful observations and experiments. Fortunately, Jürgen Tautz was easily persuaded to join in the venture, and in August 2000, he joined me at Cornell, bringing with him from Germany the miniature microphones and digital audio and video equipment that we would need for the project. Soon we had a swarm set up in a quiet spot at my laboratory, with the swarm clustered on one side of a vertical board so that we could easily monitor everything that happened on the swarm’s surface. Inside the swarm we mounted two microphones and several temperature probes, and directly in front of the swarm we positioned a video camera that recorded both the bees’ sounds from the swarm’s interior and the bees’ actions on its surface. With numerous microphone and thermometer wires leading from it, a video camera continuously recording its activity, and two biologists hovering over it, our swarm looked rather like a patient in an intensive care unit.

Because I now had a search image for a piping bee—one dashing over the swarm’s surface but pausing frequently to seize a motionless swarm-mate—I was able to spot pipers at a glance when we started hearing their shrill piping sounds. From our video recordings, Jürgen and I quickly confirmed my previous observation that the piping bees are exceptionally excited scout bees. The bees made this crystal clear by switching between worker piping and waggle dancing while scrambling over the surface of the swarm (fig. 7.7). This mixing of signals became especially noticeable during the last half hour before takeoff, when the piping grew strongest. (How the scouts know when to start piping will be explained later in this chapter.) We saw then that after a scout bee had finished a bout of waggle dancing, she was very likely to start producing a string of piping signals.

From our audio recordings of pipers sounding off near one of the microphones, we learned that each pipe is a single pulse of sound that lasts about one second and is composed of a fundamental frequency of 200 to 250 hertz (cycles per second) plus many harmonics—multiples of the fundamental frequency—in the range of 400 to 2,000 hertz (fig. 7.8). It is these high-frequency harmonics that make each pipe sound so shrill. That the fundamental frequency of the piping sound matches the wing-beat frequency of a flying bee is strong evidence that a bee produces this sound by activating the flight muscles in her thorax to create strong vibrations in her body. Probably most of this vibration energy passes as a sharp blast into the bee that the piper has grabbed hold of and pressed against, but some passes into the surrounding air creating the sounds that humans can hear while eavesdropping on a swarm. Jürgen and I also learned from our sound recordings that the upward sweep in the pitch of each pipe is produced by a shift in the fundamental frequency from 200 to about 250 hertz and an increase in the amount of sound energy in the high-frequency harmonics. Probably a piper creates these changes by pulling her wings together, thereby stiffening her thorax and raising its resonant frequency.




At this point Jürgen and I wanted to test the hypothesis that the function of worker piping in swarms is to stimulate bees to prepare for takeoff. Our first step was a check that worker piping really occurs only in the last hour or so before takeoff, when the bees in a swarm are making their flight preparations. We did this by measuring simultaneously the level of piping in a swarm and the temperatures in the swarm’s core and mantle for many hours prior to takeoff. Figure 7.9 shows an example of the patterns in piping and warming that we found. Three hours before takeoff (11:30 a.m.), when the ambient temperature was 23°C (73°F) and the swarm’s core and mantle temperatures were 34° and 31°C (93° and 87°F), we heard no piping. Then, about 90 minutes before takeoff, we began to hear piping but only intermittently. Finally, during the half hour before takeoff, the sound of the piping bees was continuous and loud, for by then multiple bees were piping simultaneously. At the same time, the mantle temperature began rising, and just when the temperature throughout the clustered swarm reached 37°C (99°F) it launched into flight.

The finding that worker piping coincides perfectly with swarm warming—both phenomena build to a climax at takeoff—provided strong support for the hypothesis that this signal functions to stimulate bees to prepare for takeoff. However, because we had shown only a correlation between worker piping and swarm warming, and because a correlation does not demonstrate causation, we could not firmly conclude that a swarm’s warming is a response to the scouts’ piping. The possibility remained that the piping had not elicited the warming, and that there was some third factor that had stimulated both piping and warming. It has been suggested, for example, that what informs the cool and quiescent bees in a swarm that it is time to warm themselves for takeoff is another signal, called the shaking or vibration signal. To produce this signal, one bee seizes another with her two front legs and then violently shakes her own body in an unmistakable, up-and-down movement for a second or two (fig. 7.10). Like a person rousing a drowsy friend with a firm shake on the shoulder, the shaking bee incites her fellow swarm bee to greater activity. But because shaking signals are produced throughout the house-hunting process, not solely or even principally in the last hour before takeoff, it seems clear that the shaking signal is not the critical signal used by the scouts to prepare the swarm for flight. Evidently, the shaking signal serves instead to raise the general level of activity of other workers in the swarm so that they are more alert and responsive to waggle dances, worker pipings, and other stimuli. At night and during times of poor weather, all the bees in a swarm will become inactive, probably to conserve energy, so it makes sense that the scout bees have the shaking signal to stimulate the general reactivation of others—mainly other scouts, I suspect—when conditions again become favorable for house hunting.



To test more conclusively our hypothesis that piping is the flight-preparation signal, Jürgen and I needed to perform an experiment in which we would manipulate the piping signals in a swarm and look for an effect on its warming. In principle, the manipulation could consist of either artificially blasting swarm bees with the piping signal or artificially blocking them from receiving this signal. We took the latter approach. To prevent piping workers from contacting a group of bees in the swarm mantle, we mounted a 25 x 20 centimeter (10 x 8 inch) screen vertically over a swarm’s surface so that the outer layers of bees in the swarm cluster were on the outer side of the screen. On this side of the screen we had mounted two small cages, each of which was equipped with a temperature probe (fig. 7.11). Both cages soon became filled with mantle bees. We prevented the scouts from contacting—hence sending piping signals to—the bees inside one of the cages by closing it with a screen cover when we started hearing bees piping. Simultaneously, we treated the bees inside the other cage in exactly the same way, except that we “closed” their cage with a cover that had a large opening through which pipers could pass. Given our hypothesis that piping bees stimulate bees to warm up for takeoff, we predicted that the mantle bees in the closed cage would not warm themselves to a flight-ready temperature at the time of takeoff, whereas those in the open cage would do so. This is precisely what we found. The open-cage bees showed the usual pattern of a dramatic rise to approximately 35°C (95°F) in the final minutes before takeoff, but the closed-cage bees did not (fig. 7.11). For fun, at the end of each trial, after all the uncaged and open-cage swarm bees had departed, we removed the cover of the closed cage and prodded the bees inside, who were eerily calm. All tumbled to the ground, too cold to fly. Without a doubt, these caged bees had missed the scout bees’ persistent alerts to warm up for takeoff.

Boisterous Buzz-Runners

The investigation of the piping signal solved the mystery of how the scout bees prime their swarm-mates for flying to their new home, but it left unsolved the puzzle of what finally triggers the highly synchronized, virtually explosive, takeoff of the thousands of bees in a swarm. A strong possibility was an eye-catching behavior that Martin Lindauer, who first described it, named the Schwirrlauf. English-speaking bee biologists call this behavior the buzz-run. It is well named in both German and English because a buzz-running bee runs across the swarm cluster, turning this way and that, usually with her outspread wings whirring and buzzing noisily. Sometimes she is dashing over the backs of the immobile bees and other times she is bulldozing between them (fig. 7.12). Lindauer reported that buzz-runners are prominent on the swarm cluster in the final few minutes before takeoff starts, and he suggested that by barging and boring through the cluster, the buzz-runners disperse the interconnected bees and so initiate their concerted takeoff. This was an enchanting hypothesis, but it remained to be tested. And even if it proved correct, many questions would remain about the hyperactive buzz-runners. What is the interplay between worker piping and buzz-running as a swarm prepares for and then takes to flight? Which bees in a swarm perform buzz-runs? And how do buzz-runners know when to produce their rough signal?


In tackling these questions, I was joined by Clare Rittschof, a Cornell undergraduate student who turned out to be a born researcher. We began our investigation of The Case of the Buzz-Runner Bees in May 2005, as soon as Clare had finished the final exams in her spring semester classes. We started with a stakeout for buzz-runners on swarms to find out when they conduct their activities. To do this, we mounted a swarm of bees on one side of a vertical wooden board and video recorded the activities of the bees within a 10 x 15 centimeter (4 x 6 inch) region of the swarm’s surface. We started each round of surveillance when the swarm bees began to produce piping signals and ended it when the bees flew of to their chosen home. Clare would play back the recordings in slow motion, scanning them for bees running erratically over the swarm’s surface. My studies of worker piping led us to expect that some of the running bees would be pipers, and Lindauer’s report indicated that others would be buzz-runners. To know whether or not any given runner was a piper, we followed each one for several seconds with a small microphone (to pick up any piping sounds she might be producing) and added this audio information to the video recording. Buzz-runners were easily identified among the running bees by the conspicuous buzzing of their outspread wings.


Clare’s painstaking examinations of our surveillance records yielded two important findings. First, she saw that more and more runners went into action during the final hour preceding takeoff so that just before its takeoff a swarm teems with bees dashing over and through the cluster. Second, and more remarkably, she saw that all of the runners produced audible signals: pipes or buzzes, or both. At first, the running bees produced just piping signals. But little by little, they started to combine piping with buzz-running—ramming into other bees and revving their wings—and during the final five minutes before takeoff more than 80 percent of the running bees produced buzz-runs (fig. 7.13). This told us that the buzz-runners are the same bees as the pipers, who we already knew to be scout bees. Thus we learned that the scout bees give both the piping signal to prime the swarm for takeoff and the buzz-run signal to trigger (evidently) the takeoff.

What is the evidence that the buzz-run signal stimulates bees to take flight? First, there is the fact that the buzz-run is an ephemeral signal that is seen in just one context: when idle bees are being stimulated to take flight. So we see buzz-runners briefly just before a swarm pours forth from its hive (as described in chapter 2) and again just before it takes wing from its bivouac site. There is also the fact that buzz-run production rises to a crescendo moments before swarm takeoff, suggesting that the former causes the latter. And perhaps most telling, there is what Clare found when she reviewed the video recordings of many episodes of a buzz-runner blasting through a knot of lethargic bees. She found that they were more dispersed and more active after they had experienced the buzz-runner’s forceful persuasion than before.

One feature of a buzz-runner’s behavior that should be noted is that sometimes she will launch herself into flight, fly around the swarm cluster for a few seconds, and then land back on the cluster and resume her buzz-running. The phenomenon of buzz-runners taking flight is important because it points to the evolutionary origins of their lively signaling behavior. Almost certainly, the buzz-run signal is a ritualized form of a bee’s takeoff behavior, which consists of a bee spreading her wings, starting to buzz them, pushing clear of other bees if need be, and finally taking to the air.

“Ritualization” is the name biologists have given to the process whereby some incidental action of an animal becomes modified over evolutionary time into an intentional signal. Usually the incidental action is a by-product of an activity performed in one particular context, so the animal’s action is a reliable indicator of this context. The buzz-run illustrates this idea nicely: when a bee is about to take flight, she inevitably buzzes her wings, so wing buzzing by a bee is a reliable indicator to others that she is about to take flight. The next step in a signal’s evolution is for the receivers to detect it and use the information it provides to improve their decision making. If the receivers’ improved decision making also benefits the senders, then the senders will benefit by making the signal more conspicuous and so more easily detected by the receivers.

In the early stages of the evolution of the buzz-run signal, the quiescent bees in a swarm probably improved their decision making about when to take flight by responding to the wing buzzing of other bees taking off. The improved decision making by the quiet bees probably produced better-coordinated takeoffs, which also benefited the active bees, so natural selection favored modifications of the wing buzzing by the active bees to make it more conspicuous to the quiet bees. Given the present-day form of the buzz-run, it appears that these modifications include exaggerating the wing buzzing (starting it long before the moment the buzz-runner takes flight) and adding to it the actions of running and ramming. I think the buzz-run shows nicely how sometimes we can glimpse the evolutionary origins of the marvelous signals that bind bee to bee to bee in a swarm.

One final question regarding the buzz-run is why did honeybee swarms evolve this signaling system? In other words, why should the scout bees send everyone else in the swarm a signal of when to launch into flight? I suggest that this signaling system evolved because it is only the peripatetic scout bees that can sense when all the bees in the swarm cluster are ready for departure, and the buzz-run signal enables the scouts to share this critical information with their swarm-mates. As we have seen, for all the bees in a swarm to launch into flight together each bee must have her thorax warmed to at least 35°C (97°F). But how can all the bees in a swarm know when they’ve all become hot enough? One way would be to have some bees travel across the swarm cluster, with each one measuring the temperatures of her swarm-mates along the way, and then sounding a departure alarm when her canvassing tells her that the required warmth has been achieved by all. I suspect that this is how it works on swarms, for we now know that the scout bees move quickly throughout the swarm cluster, with each scout pausing every few seconds to press her thorax against another bee and produce the piping signal. Perhaps each time a scout presses against another bee she also senses her temperature. And we now know that it is the scout bees that strongly produce the buzz-run signal in the final few minutes before takeoff, when all the bees have the high body temperature required for departure.

If the hypothesis of scout bees as mobile temperature sensors, information integrators, and group activators proves correct, then the mechanisms mediating the initiation of takeoffs by honeybee swarms present us with an intriguing system of behavioral control within a large group. It is one in which a small minority of individuals actively poll the group to collect information about its global state and then, when the group reaches a critical state, these individuals produce a signal that triggers an appropriate action by the whole group. The governance of a honeybee swarm is proving ever more extraordinary.

Consensus or Quorum?

We know that a swarm starts to switch from making a decision about its future home to implementing this decision when its scout bees start to produce the piping signals that inform the nonscout bees that the time has come to warm up their flight muscles. So far, so good. But how do the scout bees know when to start producing their piping signals? Given the striking way that the dances on a swarm come to represent one site and then the swarm moves to this site, it is tempting to think that the scouts use the appearance of dancer consensus to know when to start piping, rather like Quakers discuss and wait to find common ground and then, recognizing they have reached a “sense of the Meeting,” know when to take action. By this hypothesis, a scout “votes” in favor of a site by dancing for it, the scouts act and interact (as we have seen in chapter 6) so that gradually their votes come into agreement in favor of a superior site, and somehow the voting pattern of the scouts is steadily monitored so that they know when they’ve reached an agreement and can start acting on their decision.

There were, however, two facts that cast doubt on this appealing hypothesis. First, neither Lindauer nor I nor anyone else had seen any sign of scout bees polling their fellow dancers, something that surely they would need to do to sense a consensus. Second, Lindauer had seen two out of the 19 swarms that he studied launch into flight without a consensus among the dancers, that is, when there were two strong coalitions of dancers advertising two distinct sites (e.g., his Balcony swarm, see fig. 4.4). Were these cases of takeoff with dancer disagreement bizarre anomalies that should be ignored, or were they valuable clues that should be heeded?

I chose to heed them, and did so in a series of collaborative studies with my good friend and former student Kirk Visscher, who shares my passion for figuring out how honeybee colonies work. I first met Kirk in the fall of 1976, when he enrolled in the Biology of Social Insects class I was teaching at Harvard. We clicked right away. Here was someone who was already extremely knowledgeable about honeybees from years of beekeeping with his father, someone with a powerful intellect, someone with a modest sense of self, a quick smile, and a love of biology. I was to learn later that Kirk is also a fabulous gadgeteer, expert statistician, and computer whiz—all things that I’m not. Today Kirk is on the faculty of the University of California at Riverside.

Even though Kirk and I now live on opposite sides of the North American continent, we teamed up because both of us had long wondered whether the scouts on a honeybee swarm know when to start their piping by sensing a quorum (sufficient number of scouts) at one of the nest sites rather than by sensing a consensus (agreement of dancing scouts) at the swarm cluster. By the quorum-sensing hypothesis, a scout “votes” for a site by spending time at it, the number of scouts rises faster at better sites, and somehow the bees at each site monitor their numbers there so that they know whether they’ve reached the threshold number (quorum) and can proceed to initiating the swarm’s move to this site. This hypothesis can explain the cases of takeoff with disagreement among dancers as instances where a quorum was reached at one site before the competition between dancers from different sites had eliminated the dancing for all but one site.

We tested these two hypotheses with experiments performed on Appledore Island. In our first experiment, we presented four swarms, one at a time, with two identical nest boxes that offered the bees two superb nest sites. Our goal was to foster strong debates on our swarms and then see if they would take off before their dancing bees had reached a consensus (as Lindauer had reported for two of his swarms). In each trial, we positioned the swarm at the island’s center, on a porch of the old Coast Guard building, and we placed both nest boxes near the rocky shore, each 250 meters (820 feet) from the swarm but in different directions, to the northeast and to the southeast. We also wanted to census the scout bees inside and outside each nest box, so each box was mounted against a window on the side of a lightproof hut (see fig. 3.11). The plan worked! We found that our swarms would discover both nest boxes at about the same time, would tend to develop a balanced debate over these two highly attractive sites, and would routinely take off when scout bees were still dancing strongly for both sites. Most telling was the spectacle that we witnessed on July 7, 2002. At 12:04 p.m., when both nest boxes were being advertised by dozens of bees performing vigorous dances, our swarm took of and then the large cloud of swarm bees split itself in two! Separate groups of airborne bees gathered on the north and south sides of the Coast Guard building, and at 12:09 each group began to move off slowly in the direction of “its” nest box. Both groups, however, traveled only about 40 meters (130 feet) toward their nest box before stopping, and at 12:15, when we noticed that the swarm’s queen was back on the porch of the Coast Guard building, both groups started to return there and resettle around her.

This first experiment showed us that consensus among dancers is not necessary for a swarm to initiate its move to a new home, hence we could reject the consensus-sensing hypothesis for how scouts know when to start piping. This experiment also provided some support for the quorum-sensing hypothesis, because we noticed that swarms consistently started preparing for flight—that is, their scouts started piping—once 20 to 30 or more bees were seen together at one of the nest boxes, usually with about 10 to 15 bees inside and 10 to 15 outside. This suggested that in the decision-making system of honeybee swarms, 20 to 30 bees present simultaneously at one of the potential nest sites is a quorum. It should be noted, however, that because scout bees spend much of their time at the swarm cluster, seeing some 25 bees at a prospective nest site at any one time means that the total number of bees making visits to and thereby expressing support for this site is about 50 to 100 bees.

In performing our second experiment on Appledore Island, in June and July 2003, we sought to make a direct test of the quorum-sensing hypothesis for how scouts know when to start producing their piping signals. Our plan was to test a critical prediction of this hypothesis: delaying the formation of a quorum of scout bees at a swarm’s chosen nest site, while leaving the rest of the decision-making process undisturbed, should delay the start of piping and thus the takeoff of the swarm. This was a critical prediction of the hypothesis because if we found that this prediction was wrong, then we would have dealt a deathblow to the hypothesis of quorum sensing.

Kirk and I devised a simple but effective way to delay quorum formation: we placed five desirable nest boxes close together at one location on the island (fig. 7.14). This caused the scouts visiting the site to be dispersed among five identical nest cavities rather than concentrated at one. We then saw how long it took a swarm, once it had discovered the site of the nest boxes, to start piping and eventually take of to fly to the site. We also performed with each swarm another control trial with just one nest box. The two trials for each swarm were performed using two different sites on the island, so each trial began in the same way, with one scout bee discovering an attractive nest box in a new site. In all four swarms that we tested, the scouts concentrated their attention on the single box in the one-nest-box trials and a crowd of bees built up there rapidly, but they distributed themselves evenly among the multiple boxes in the five-nest-box trials so the crowds of bees at these boxes built up more slowly. And in all four swarms, there was indeed a marked delay to start of piping and to start of takeoff in the five-nest-box treatment relative to the one-nest-box treatment. The times from discovery of the nest box(es) to start of piping and to start of takeoff were 162 and 196 minutes on average in the one-nest-box trials, but 416 and 442 minutes on average in the five-nest-box trials. It should be noted that the amount of waggle dancing back at the swarm did not differ between the two treatments. Also, the level of dance consensus was the same for both treatments; the bees were always unanimous in dancing only for the site of the nest box(es). It seems clear, therefore, that our five-nest-box treatment did not disturb anything in the decision-making process at the swarm cluster, and yet it delayed the start of piping and the start of takeoff. Thus this experiment yielded strong support for the quorum-sensing hypothesis.


Based on our two experiments conducted on Appledore Island in 2002 and 2003, Kirk and I drew the conclusion that a quorum of scouts at one of the proposed sites, not a consensus among dancers at the swarm, is the key stimulus for scouts to start piping and thereby initiate preparations for swarm takeoff. But how do we reconcile this conclusion with the fact that by the time the swarm takes of it must have a consensus among its scouts in order to fly as a unit to a single chosen site? One possible answer is that the preparations for takeoff, which generally take an hour or more, provide sufficient time for the positive feedback process of recruitment to the best site to produce the necessary unanimous agreement among the scouts. There may, however, be more to the story. It may be, for example, that the piping signals—which Kirk and I learned in 2006 are produced only by scouts returning to the swarm from the chosen site—inform the scouts from the losing sites (all those without a quorum) that the contest is over and that they should stop advertising these sites. This would certainly help the scouts reach full agreement about their future home, but whether the scouts from losing sites really respond to piping signals in this way remains unknown.

Exactly how the scouts sense a quorum also remains unknown. One possibility is that they use visual information. For humans, and perhaps also for bees, the constantly moving scout bees are easily detected visually outside the cavity and even inside it, at least around the entrance opening, which admits considerable light. Another possible means of sensing the number of bees at a site is by touch. It is a curious fact that as soon as a site acquires multiple scouts, they begin to make frequent contacts with one another. Many even start to perform what look like buzz-runs inside and outside the prospective nest site and so butt against other bees. It seems entirely possible that a bee could use the rate of contacts with scouts in general, or collisions with buzz-runners in particular, as an indicator of the number of fellow scouts at a site. Still a third possibility is the use of olfactory information. Scout bees standing in the entrance opening of a potential home often fan their wings and expose their scent organs—thereby releasing the lemon-scented blend of attraction pheromones, the message of which is “Come here!”—presumably to help other scouts find this special spot. It is possible that the level of these attraction pheromones rises with increasing numbers of bees at a site. Testing these various possibilities remains a subject for future study.

Why Quorum Sensing?

At first thought, it seems odd that the scout bees use quorum sensing rather than consensus sensing to know when to begin preparing their swarm for its flight to the new home. After all, a consensus among the dancers is needed for a swarm to execute successfully a move to its chosen homesite. Both Lindauer, with his Balcony and Moosach swarms, and Kirk and I with our Appledore swarms in 2002, saw what happens when a swarm takes off when its dancers are strongly divided between two sites: the cloud of airborne bees splits up, both halves stall in their moves, and finally they rejoin by resettling wherever their weary queen alights. Thus the bees have a big to-do but get nothing done.

Why don’t the scouts use consensus sensing and thereby avoid the risk of their swarm splitting after takeoff and going nowhere? One likely reason is that sensing a consensus among the dancing bees would be extremely difficult for the bees. Presumably, each scout would have to poll the advertisements of her fellow scouts, which would involve traveling over the swarm cluster, reading dances, and keeping a mental tally of these readings. Doing all these things would be especially difficult on larger swarms with more scouts and thus more dances to poll. Quorum sensing, however, need not become more difficult with increasing swarm size, because the quorum size could be fixed, hence independent of swarm size.

Another likely reason that scouts don’t use consensus sensing is that quorum sensing, unlike consensus sensing, enables a swarm to strike a good balance between speed and accuracy in its decision making. Consider first the matter of speed. Using a quorum as the trigger for the start of takeoff preparations means that these preparations can begin as soon as enough scouts have approved of one of the sites, even if many other scouts are still visiting and advertising other sites. In other words, there’s probably no need to wait for full agreement if the outcome can be sensed in advance. If the bees used a consensus as the trigger, then the start of a swarm’s takeoff preparations would be delayed by the extra time needed to reach a consensus. Consequently, the swarm bees would burn through more of the small store of energy (honey) they brought with them. This further depletion of a swarm’s energy reserve would be considerable if the delay in start of flight preparations were to force a swarm to postpone its departure to the following day—swarms rarely take of after 5:00 p.m.—and so spend another chilly night camping out.

Now consider the matter of accuracy. It seems that the quorum used by the bees is 20 to 30 bees present simultaneously at a site (half inside, half outside), which requires that some 75 scout bees are actively supporting this site since each one spends only part of her time at the site. Using a 20- to 30-bee quorum evidently helps ensure accurate decision making because it guarantees that scout bees will not begin producing piping signals until a sizable number of them have independently scrutinized a site and judged it worthy of their support. This makes it extremely unlikely that a swarm will choose a poor site when a better one is available, for a poor site will not attract a (large) quorum of scouts. To see why, imagine that a scout makes a mistake—judging a poor site to be a good one—and recruits strongly to a poor site. Her followers will correct her error when they examine it themselves, find that it fails their scrutiny, and refrain from advertising the site further. Thus the mistake of the erring scout is soon silenced, the number of scouts at the site quickly dwindles, and the swarm rejects this low-quality site. I suspect that quorum size is a parameter of the bees’ decision-making process that has been tuned over evolutionary time to provide an optimal balance between speed (favored by a small quorum) and accuracy (favored by a large quorum). We will examine this matter further in chapter 9.

The idea that a group can strike a better balance between speed and accuracy in its decision-making if it uses a quorum rather than a consensus to know when to start taking action is nicely illustrated by a story told to me by a neighbor friend who is a Quaker. Some years ago, the members of her Meeting wrestled with the question of whether or not to change the location of their meetinghouse. In meeting after meeting the subject was discussed, with the Friends always seeking a united wisdom, but every discussion ended with an adjournment for further consideration because an agreement could not be reached. Why? Because there was one Friend, an elderly lady, who felt strongly that the proposal was a mistake, withheld her consent, and so blocked a decision. If the Meeting had used a quorum, acting when a sufficient number or proportion of its members agreed, it would have reached its decision in a few weeks, but waiting for a consensus required four years. What finally enabled the Meeting to achieve a united judgment was the death of the one disapproving Friend. Some decisions do need to be made quickly, even if the choice is imperfect. The Quaker way of business, with its unceasing patience in finding full agreement, would be extremely risky for a homeless swarm of bees hanging in limbo under the open sky.