Honeybee Democracy - Thomas D. Seeley (2010)


The experience of democracy is like the experience of life itself—

always changing, infnite in its variety sometimes turbulent

and all the more valuable for having been tested by adversity

—Jimmy Carter, Address to the Parliament of India, 1978

When a honeybee swarm chooses its future home, it practices the form of democracy known as direct democracy, in which the individuals within a community who choose to participate in its decision making do so personally rather than through representatives. The collective decision making of a bee swarm therefore resembles a New England town meeting in which the registered voters who are interested in local affairs meet in face-to-face assemblies, usually once a year, to debate issues of home rule and to vote on them, rendering binding decisions for their community. Of course, there are differences in how direct democracy works between bee swarms and town meetings. For example, the scouts in a bee swarm have common interests (e.g., all want to choose the best available homesite) and they reach decisions by building a consensus. The people in a town meeting, however, often have conflicting interests (e.g., some do and some don’t want to help fund the town library), and they reach decisions by using the majority voting rule: each individual has one vote, all votes have equal weight, and the option that gets the majority wins. Another basic difference between bee swarm and town meeting is that a scout bee in a swarm, unlike a citizen in a meeting, cannot monitor each exchange within the group’s debate and thereby have a synoptic overview of the discussion. Instead, a bee can only observe and react to the actions of her immediate neighbors in the swarm cluster, hence she operates without global knowledge of the information that percolates among her fellow swarm bees.

While these differences—common versus conflicting interests and local versus global knowledge—between bee swarm and town meeting are real, they do not overshadow several extremely important similarities between the direct democracies of honeybees and human beings. Firstly, in both the insectan and the human forms of this collective decision making, each decision about a future course of action refects the contributions, freely given and equally weighted, of several hundred individuals. In other words, the control of the group’s actions is distributed among many of its members rather than concentrated in a few leaders. Secondly, because hundreds of individuals are full participants, the group can acquire and process information from multiple sources simultaneously, even ones that are widely scattered. Consider, for example, the first stage of every decision-making process, where the critical challenge is to identify the available options. By virtue of having numerous individuals examining a problem and presenting possible solutions to it, both a bee swarm and a town meeting are much more capable than any solitary bee or single person in coming up with a broad range of alternative options. And the broader this range of options, the more likely it will include the one best option. Thirdly and most intriguingly, in both bee swarms and town meetings, the way the group selects its course of future action is by staging an open competition among the proposed alternatives. An individual proposes a possible way forward, each listener makes an independent assessment of the proposal and decides whether to reject or accept it, and those that accept it may announce their own support for the proposal. These endorsements often recruit still more supporters for this option. The better the proposal, the more supporters it will attract, and the more likely it is to gain sufficiently broad support to become the community’s choice.

In the case of the house-hunting bees, the competition among the supporters of different proposals is often fierce, with some scout bees vigorously champion ing one lovely tree cavity while other scouts, just a few bees away on the surface of the swarm cluster, are enthusiastically advertising a second, third, or even fourth desirable dwelling place. (Later in the chapter, we will see which bees take up the “profession” of nest-site scout and what prompts these bees to step into this dangerous job. In a nutshell, nest-site scouts are elderly bees who were working as common foragers but then quit this line of work when they sensed that their colony, preparing to swarm, no longer needed additional food.) But it is always a “friendly” competition; the scout bees agree on what makes an ideal homesite, they are united in the goal of choosing the best available site, they share their information with full honesty, and ultimately they reach a complete agreement about their new residence. One valuable lesson that we can learn from the bees is that holding an open and fair competition of ideas is a smart solution to the problem of making a decision based on a pool of information dispersed across a group of individuals.

Lindauer’s Swarms

The discovery that swarm bees use debates to aggregate dispersed information was made by Martin Lindauer in 1951 and 1952, when he received permission from Karl von Frisch to use all the swarms that emerged from the hives of bees kept in the garden behind the Zoological Institute at the University of Munich. Thus, at last, Lindauer was able to make a close study of what had aroused his curiosity back in the spring of 1949: dirty dancing bees on the surface of a swarm (fig. 4.1). The institute’s bee colonies did their part, providing Lindauer with 17 swarms spread across the months of May, June, and July. Many of the swarms few across the street to the Botanical Garden where they settled on various objects, and Lindauer charmingly named each swarm for its bivouac site: the Linden swarm on May 18, 1951, the Hawthorn swarm on July 9, 1951, the Balcony swarm on June 22, 1952, and so forth. His immediate aim was to determine whether the bees dancing on a swarm are nest-site scouts, but his ultimate goal was to understand how a swarm finds its new homesite. Knowing the value of starting an investigation of animal behavior with a “watching and wondering” phase, during which one patiently makes broad observations that can yield unexpected discoveries, Lindauer watched from dawn to dusk the bees performing waggle dances on the surface of each clustered swarm. Knowing too that he wanted his dancing bees to be identifiable individuals, so that he could begin to enter the world of a bee in a swarm, he applied colored paint dots to the back of each dancing bee. Karl von Frisch had devised a marking scheme back in the 1910s that enables one to number bees from 1 to 599 with just 5 paint colors, and Lindauer used this system, deftly daubing with a tiny paint brush one to four tiny dots of shellac paint on the thorax of each dancing bee.


Lindauer’s endeavor to record the dances on a swarm started easily enough, for few bees danced and their dance performances were intermittent. So at first Lindauer could record in his notebook the time of each bout of dancing, the identity of the dancer, and what location this individual advertised with her dance. In time, however, his task of recording the swarm bees’ dance activity became almost impossibly difficult; sooner or later he faced a dozen or more bees dancing simultaneously on the swarm. He coped with this overwhelming display of dancing bees by becoming more selective in what he recorded; he noted only the time he saw each new (not yet labeled) dancing bee and the location indicated by each new bee’s dance. It was exhausting to single-handedly observe, label, and record the dancing swarm bees for hour after hour, sometimes for days on end. But it was immensely revealing. In chapter 1, we saw how Lindauer tested his hunch that the bees dancing on swarms are nest-site scouts advertising potential nesting cavities. He found that the site indicated unanimously by the dancing bees shortly before a swarm few to its new homesite matched the address of their new dwelling place. Even more marvelous was Lindauer' s discovery that a swarm’s scouts conduct a vigorous debate to select their new home.

Figure 4.2 shows an example of one of these debates. It was held on the Eck swarm, a swarm that left its parental hive at 1:35 in the afternoon on June 26, 1951, soon settled in a privet bush, and then hung out there for nearly four days while its scout bees went about choosing a dwelling place. On the first day, Lindauer observed and labeled just two dancing scout bees, between 1:35 and 3:00 p.m. One bee reported a candidate homesite approximately 1,500 meters to the north, while the other proposed a second site 300 meters to the southeast. By 3:00 p.m. the sky had filled with dark rain clouds and the air had became cooler, so the scout bees ceased their explorations for the day. The next day the scouts remained inactive until late in the morning, when the clouds parted and bright sunshine returned. Figure 4.2 shows that Lindauer labeled 11 new dancers between noon and 5:00 p.m., and that three of them advertised the site 1,500 meters to the north, two others advertised the site 300 meters to the southeast, and the remaining six indicated six other sites in various directions and distances. Clearly, no agreement among the dancing bees was reached during this second day of debate. On the third day, the weather was mostly rainy and Lindauer recorded just two new dancers, both in late morning. One advocated the site to the north, thereby strengthening this site’s slim lead among the dancers (with five bees total, so far), and the other reported a new site, about 400 meters to the southwest.


On the fourth day, the sky cleared, the air warmed, and the bees became active. More than 20 new sites were reported, though surprisingly the site 1,500 meters to the north did not receive additional dancers. Lindauer suggested that it might have been a cavity with a leaky roof into which rainwater had seeped the previous day, rendering it unattractive as a nesting site. Most of these 20 new possibilities were announced by just one dancer each and so weren’t serious proposals in the bees’ debate, but a few did receive the attention of multiple bees and so became important possibilities. For example, between 9:30 a.m. and 4:00 p.m. nine bees advertised a site located 1,500 meters to the west. Even here, however, the bees’ interest eventually faded; from 4:00 to 5:00 p.m., Lindauer recorded no new dancers for the western site. Only one site, the one located 300 meters to the southeast, held the bees’ interest all day, with new dancers for this site getting painted on the swarm each hour. Figure 4.2 shows that the rate at which Lindauer recorded new dancers for the southeast site grew gradually over the day, and that by 4:00 p.m. this site held a commanding lead. But it was not until the final hour of dance activity, between 4:00 and 5:00 p.m., that the recruitment of scouts to the southeast site utterly overwhelmed that to other sites, with fully 61 dancing newcomers advertising the southeast site and only two recruiting to alternative sites. The situation remained the same the following morning, with 83 out of the 85 new dancers in favor of the southeast site. Finally, at 9:40 in the morning on June 30, this swarm launched into flight and few 300 meters to the southeast to take up residence in the wall of a bomb-damaged building.

The overall pattern that Lindauer reported for the Eck swarm—initially new dancers appeared on the swarm indicating various possible nesting site, then bit by bit the new dancers became concentrated on one of the sites, and finally the swarm few of in the direction of the favored site—proved typical for the 17 swarms he watched. Occasionally, though, Lindauer encountered a swarm where the debate did not progress so smoothly. For example, sometimes the scout bees would find two sites that elicited strong dancing starting at about the same time. In this situation, both sites would receive plenty of fresh recruits and so new dancers for both sites would continue to appear on the swarm for many hours. Of course, this made it extremely difficult for the dancing bees to reach an agreement.


Figure 4.3 shows one example of how a prolonged debate arose when a swarm’s scout bees created two equal-strength groups of dancers. The Propyläen swarm’s saga began on June 11, 1952, when it left its mother hive at 2:14 p.m. Over the remainder of the afternoon, the scouts discovered and reported on 11 sites, one of which attracted the greatest number of dancers. We see that Lindauer labeled 15 bees dancing for a site some 900 meters to the northeast, but only 14 total for the other 10 sites. So at first it looked like the scouts in this swarm had come close to an agreement in less than three hours of discussion. On the morning of the second day, however, another group of strong dancers appeared, with each of its members promoting a site 1,400 meters to the southwest. Lindauer reported that “their dances were no less lively and continual, and for the observer it was a nerve-wracking task to witness the tug-of-war of these two dance groups that dragged on for two days with see-sawing prospects for victory.” Throughout the second day the appearance of new dancers for the northeast and southwest sites was nearly symmetrical, and it was not until late in the morning on the third day, June 13, that this symmetry began to break. For some reason, the strength of dancing for the southwest site must have weakened slightly, so fewer new dancers for this site were marked, and over the afternoon the mustering of new dancers for the northeast site surged ahead of that for the southwest site: first 25 versus 9 new dancers (12:00 to 2:00 p.m.), then 41 versus 7 new dancers (2:00 to 4:00 p.m.), and eventually 34 versus 0 new dancers (4:00 to 5:00 p.m.). We see that by the end of the third day an agreement had been reached, but, alas, it was too late in the day for the swarm to undertake its flight to its new home. (Swarm’s rarely take flight after 5:00 p.m., perhaps to avoid risking a move without a long period of daylight ahead. A swarm’s queen may need to make an emergency landing to rest, and when this happens it can take the workers in an airborne swarm over an hour to halt their group flight, locate their missing queen, and reassemble around her.) The Propyläen swarm’s delayed decision-making forced it to stay put until the next day, which turned out to be cool and rainy, so it was not until the afternoon of June 15, fully four days after the swarm left its parental hive, that the bees finally few of to their new residence in the northeast.


Lindauer even observed one swarm that failed to reach an agreement, that is, there never arose a dancer group that so dominated the deliberations that Lindauer marked new dancers for only one site. It was the Balcony swarm of June 22, 1952 (fig. 4.4). As in the Propyläen swarm, its scout bees got into a balanced competition, with one group of dancers advocating a site 600 meters to the northwest and a second group favoring instead a site 800 meters to the southwest. For four hours (12:00 to 4:00 p.m.) neither group managed to gain a decisive lead. Nevertheless, at 4:10 p.m. the swarm lifted of and then did something that Lindauer could scarcely believe even though he was seeing it with his own eyes. In his words, “The swarm…sought to divide itself. The one half wanted to fly to the northwest, the other to the [southwest]. Apparently, each group of scouting bees wanted to abduct the swarm to the nesting place of its own choice.” And each group partly succeeded, for half of the airborne bees started moving southwest toward the main railway station while the other half began trending northwest toward the Karlstrasse (see fig. 1.6). But neither group managed to keep going in its desired direction, perhaps because each lacked the queen, and eventually the two clouds of swirling bees reunited in the air where they had started. Then a remarkable tug-of-war between the two groups began. Over the next half hour, one group tried again to advance to the northwest, going of 100 meters before coming back, then the other group pushed 150 meters to the southwest, but then it too returned to the original site. At this point the bees resettled on the balcony where they had previously been clustered. Sadly, the swarm’s queen had become lost during the aerial tug-of-war, and over the next several hours Lindauer watched the swarm cluster gradually dissolve as the queenless bees drifted home to their mother hive.


The history of the ill-fated Balcony swarm highlights several features of honeybee swarms and their decision-making process. From the tragedy of the queen’s loss and the swarm’s dissolution, we are reminded that the survival of a swarm’s queen, who carries the new colony’s genes in her ovaries and her spermathe-ca, is absolutely critical to a swarm’s success. From the failure of the scouts to reach an agreement, we are shown how swarms are not infallible in their decision making. Occasionally, a swarm will produce a split decision, though usually this is just a temporary situation and the swarm manages to resolve the problem. We will see in chapter 7 that when a split decision arises, a swarm will normally take of, fail to move to either site, resettle, and then conduct further debate leading to an agreement. Lindauer watched 17 swarms of which just two (the Balcony and Moosacher swarms) produced split decisions. Only the Balcony swarm never reached agreement, and this was because it lost its queen, so it seems clear that complete failure in swarm decision making is rare. Finally, from the launch of the Balcony swarm into flight without prior consensus among the dancers, we are given a strong indication that dancer consensus, though conspicuous to the human observer, is not what the bees themselves monitor to know when they should switch from makinga decision to implementing a decision. How the bees actually make this switch will be revealed in chapter 7.

My Swarms

A great scientific discovery is one that gives rise to shining insights whose light dispels obscurities, opens up new paths, and reveals unknown horizons. Martin Lindauer made such a discovery when he found that a honeybee swarm chooses its future home through a debate in which the nest-site scouts express their arguments in waggle dances. He elucidated the function of dances on swarms, he blazed a trail toward understanding the swarm bees’ system of decision making, and most importantly he guided us into a whole new scientific territory: sophisticated group decision making by nonhuman animals.

Lindauer was certainly a pioneer in behavioral biology, and like all pioneers, the time and tools available to him were insufficient to explore fully the new terrain he had discovered. It is not surprising, therefore, that his investigation of honeybee democracy was incomplete in many ways. We can see this perhaps most clearly in the way he was limited by his equipment (notebook, watch, and paint set) to recording only the appearance of new dancers on his swarms. Ideally, he would have recorded all the dancers (new and old) that appeared on his swarms during each stage of the decision-making process, so that he could have acquired a full picture of the dynamics in the dancing for the alternative sites. This would have shown how the total number of supporters for each proposed site—not just the appearance of new supporters—changed over time, and how in the end presumably just one site was being advocated by the dancing bees. The records shown in figures 4.2 and 4.3 indicate that shortly before swarms few away the winning site was being advertised by all the new dancers, but these records don’t tell us whether in the end the winning site was being advertised by all the dancers, that is, all the old ones and all the new ones. Does the decision-making process in fact end cleanly, with essentially every dancer supporting a single site, the winning site? Lindauer suggested that it does; he wrote that the scouts for the losing sites ultimately “gave up their recruitment,” presumably by ceasing to dance, but he did not show that they stopped dancing. Also, he did not show when the advocates of the losing sites stopped dancing or how they stopped dancing. It would be, therefore, very informative to have a complete picture of the dancing on a swarm. It would also be extremely desirable to have complete records of the behaviors of individual dancers, so that one could track each dancer’s actions following her first bout of dancing. Does she perform multiple bouts of dancing? Is the total amount of her dancing related to the quality of the site that she is advertising? And if she stops dancing, how does she decide to do so? Does she quit on her own regardless of what is happening around her, or only after encountering another bee performing a more vigorous dance? Lindauer' s study is an amazing frst reconnaissance of how house-hunting honeybees perform their collective decision making, but it left unanswered countless questions about the rules of procedure that nest-site scouts follow when conducting their deliberations.

In 1996, I decided to tackle these questions. This was nearly 20 years after I had finished my PhD thesis research on the nest-site preferences of honeybees and the way they estimate the volume of a prospective nest cavity. Why hadn’t I started addressing the gaps in Lindauer' s work back in the mid 1970s? It was because I didn’t see how I could get the video recording equipment that I knew I’d need to go beyond Lindauer's analysis of swarm decision making. In those days, a color video camera, recorder, and monitor (they were all separate units back then) cost many, many thousands of dollars, a price that far exceeded the budgets of the small grants I could get as a beginner scientist. So I changed the focus of my research, but without abandoning the topic of how social animals make collective decisions. Rather, I switched to studying another form of collective decision making by the bees: how a honeybee colony wisely deploys its foragers among the kaleidoscopic array of flower patches in the surrounding countryside. This is a different sort of collective choice, for whereas a homeless swarm makes a “consensus decision” about which single option (candidate nest site) it will choose, a foraging colony makes a “combined decision” about how to allocate its foragers among multiple options (candidate food sources). I was attracted to the puzzle of colony decision making about forager allocation partly because it looked fundamentally similar to nest-site choice—being based on competition among groups of dancing bees advertising different options (food sources rather than nest sites)—and partly because it looked more tractable than nest-site choice. Swarming is an ephemeral phenomenon that lasts at most a few days, whereas foraging goes on all summer long. So for about 15 years I explored with pleasure how the bees in a hive work together as a unified whole in gathering their food, especially how they wisely distribute themselves among flower patches. In 1995 I summarized this body of research in a book, The Wisdom of the Hive, and with this behind me I looked forward to resuming my exploration of collective decision making in honeybee swarms.

The starting point was clear. I should obtain a complete record of the scout bees’ dances throughout a swarm’s choice of its new home in order to get a full picture of how a debate among scout bees unfolds. This broad description of the scout bees’ behavior at the swarm would fill the gaps in the description provided by Lindauer, and probably would also yield important discoveries, as indeed it did. Unlike Lindauer 40 years before, or myself 20 years before, I now had sophisticated video recording and slow-motion playback equipment that made it possible to build a comprehensive record of the scout bees’ dance activity. Also, I now knew how to label thousands of bees for individual identification with plastic color-number tags glued on the thorax and paint marks placed on the abdomen (fig. 4.5), a skill that I had honed in my studies of honeybee foraging. With each bee in a swarm so labeled, I figured that it should be possible to trace each individual’s history of dancing throughout the decision-making process. Success in this project would require, however, a formidable amount of painstaking labor. Thousands of bees had to be individually labeled for each swarm, the swarm and video equipment had to be tended throughout each swarm’s choice of a home, and the information about each dance (the dancer’s ID, the site being advertised, and the dance’s duration) had to be extracted manually from the video recordings. It was my immense good fortune to be joined in this endeavor by Susannah Buhrman, an extremely bright and indefatigable undergraduate student at Cornell. She proved an indispensable partner in this project. Working together throughout the summer of 1997, we achieved success.



Susannah and I eavesdropped on the deliberations among the scout bees on three swarms, and we obtained a complete record of the dances performed for each swarm’s decision making. Figure 4.6 shows the debate recorded on Swarm 1, which was set up at 10:00 a.m. on June 19. We see that nest-site scouts started reporting discoveries between 1:00 and 3:00 p.m. and that by the end of the day seven candidate nest sites (A–G) had been raised for consideration, though none of these had elicited enthusiastic support. The next day the scout bees were livelier. By midday, four additional sites (H–K) had been entered into the discussion and three sites—G (2,200 meters to the southeast), H (2,600 meters to the east), and I (4,200 meters to the south)—had received endorsements by multiple dancing bees. Site G appeared to be developing a lead, for nine bees had advertised it, but no site was yet dominant in the dancing. The situation changed markedly between 12:00 and 2:00 p.m. Now site I rose to prominence, supported by 23 out of the 25 dancing bees. This situation persisted for the remainder of the afternoon, though still two more possibilities (sites L and M) were presented and the dancing bees showed support for sites K, L, and M as well as site I until the end of the day. The next morning, however, there was a clear consensus among the dancing bees in favor of site I, and at 9:10 a.m. the swarm took of and few to the south, no doubt with site I its destination.

The debate on Swarm 1 proceeded in a manner reminiscent of what Lindauer described for his Eck swarm. During the first half of the decision-making process, the scouts reported numerous candidate sites located at various directions and distances from the swarm cluster. Then, in the second half of the debate, the dances of the scout bees quickly and smoothly became focused on just one site. Ultimately, there was virtual unanimity among the dancing bees, and the swarm moved to the agreed-upon site. It is worth stressing that figure 4.6 depicts, for each time interval, the number of total dancers for each site, not just the number of new dancers for each site. We can be confident, therefore, that the conclusion of this swarm’s decision making was marked by a real consensus among the dancing bees.

The most interesting scout bee debate that Susannah and I observed occurred in our Swarm 3; in this one there was a strong competition between two groups of dancers, and for many hours it was unclear which faction would emerge victorious (fig. 4.7). This swarm was set up at 2:30 p.m. on July 19, but it was not until the middle of the next day that scout bees started advertising potential nest sites with their dances. Six sites (A–F) were announced between 11:00 a.m. and 1:00 p.m., one of which (site A, 2,200 meters to the east) quickly developed a strong lead with eight dancing bees promoting it. Over the next four hours three more possible home sites (G, H, I) were entered into the discussion, and four sites (A, B, D, G) received strong consideration, each being advertised by several bees. Site A, though, was losing its initial lead as sites B (900 meters to the south) and site G (1,400 meters to the southwest) gained more and more support. Between 3:00 and 5:00 p.m., only four bees danced for site A, whereas 17 bees danced for site B and 10 did so for site G. At 5:00 p.m. it seemed that the contest among dancer groups on this swarm was still wide open. This situation changed dramatically during the remaining two hours of dance debate that day, for although bees performed dances for seven sites during these last two hours, including two new sites (J and K), only sites B and G received the support of multiple dancing bees. Susannah and I could see that the supporters of these two sites had managed to gain wide leads over the bees promoting the other nine sites, and we made bets on whether the B group or the G group would prevail the next day. I bet on site B and Susannah on site G. Whoever would win the bet would be treated to a triple scoop ice cream cone at the new Ben and Jerry’s in Ithaca.


The tension was high the next morning. We arrived at the lab shortly after sunrise, got our recording equipment set up before the scout bees could resume their debate, and waited eagerly to see who would win our little bet. For the first two hours, 7:00 to 9:00 a.m., we both remained hopeful, because both sites were being advertised by about a dozen dancing bees. Starting around 9:00 a.m., though, my optimism began to ebb, for the dancers supporting Susannah’s site G began to build a commanding lead over those for my site B, with 32 versus 17 bees respectively between 9:00 and 11:00 a.m., and 20 versus 4 between 11:00 and 11:54 (when it started raining). Somehow the site G bees had managed to overwhelm those for site B. The rain continued throughout the afternoon, during the night, and until about 8:00 a.m. the next day. The scout bees resumed dancing a little after 9:00 a.m., and now they showed unanimous support—73 out of 73 bees!—for site G in the southwest. Shortly before noon the bees few of to the southwest, and shortly after noon Susannah and I motored of to Ben and Jerry’s.

It was a great pleasure to watch the dance competitions among scout bees, but it was an even deeper pleasure to analyze the diagrams like figures 4.6 and 4.7 that we prepared many weeks later, after we had extracted all the information we needed from our 48 hours of video recordings. These diagrams gave us a crystal clear picture of the main features of the scout bees’ decision-making process. First, they showed that the bees’ debates tend to start slowly with an information accumulation phase during which scout bees put a sizable number of widely scattered alternatives “on the table” for discussion. In the three swarms that Susannah and I watched, the number of sites considered was 13, 5, and 11. These sites were located in various directions and at various distances (200 to 4,800 meters) from the swarm cluster, which indicates that the intrepid scouts from these swarms had searched some 70 square kilometers (about 30 square miles) of countryside for possible dwelling places. Most of the candidate sites were introduced during the first half of the deliberations, but as we can see with sites L and M in Swarm 1 (fig. 4.6), sometimes a few got introduced rather late in the discussion. Certainly a swarm does not manage to identify all its alternative options simultaneously, but as we will see in chapter 5, this asynchrony usually does not lead to swarms making poor decisions.

Second, the plots of the dance records showed that the scout bees’ debates end with all or nearly all of the dancing bees advocating just one site, that is, showing a consensus. A burning question is, therefore, how does the fierce competition among groups of bees favoring different options get transformed into a harmonious agreement? Specifically, how is it that the number of dancers builds up for one site (generally the best one, as we will see in chapter 5) while at the same time it falls to zero for all the other sites? We will see how the bees accomplish these things, using some nifty tricks, in chapter 6.

Third, our analysis showed that the bees’ decision-making process is a highly distributed and thus a democratic one, involving dozens or hundreds of individuals. Susannah and I observed 73, 47, and 149 bees performing dances in the three swarms we studied. These counts, however, probably underestimate the typical number of dancing bees in a swarm. This is because we used unusually small swarms—with only 3,252, 2,357, and 3,649 bees—to keep doable our task of individually labeling the bees. Swarms in nature generally contain 6,000 to 14,000 bees. The mean percentage of dancing bees in our labeled-bee swarms was 2.8 percent, which is similar to the 5.4 percent fgure for natural swarms reported by David Gilley, another dedicated Cornell undergraduate student who investigated the mystery of the identity of scout bees (next section). Given that 3 to 5 percent of the bees in a swarm participate in the dance debate, we can estimate that a typical swarm of some 10,000 bees will have approximately 300 to 500 individuals contributing to the decision-making process.

Intrepid Explorers

The profession of nest-site scout is performed only a few days each year, usually in the late spring or early summer. This fact, together with the fact that worker bees have short life spans—only three to five weeks during the warm months of the year—tells us that many generations of bees will pass without the need arising for individuals to explore for new accommodations. And yet when a colony prepares to cast a swarm, a small fraction of its workforce springs into action as nest-site scouts. These intrepid explorers are the prime movers of the whole swarming process. They determine when the swarm leaves its mother hive (as discussed in chapter 2), they make the swarm’s life-or-death choice of a suitable nesting cavity, they trigger the swarm’s takeoff to fly to its new home (see chapter 7), and they steer the swarm during its flight (see chapter 8). Who are these all-important bees? And what stirs them to action?

Evidently, scout bees are forager bees that have radically switched their behavior so that instead of seeking bright blossoms they search for dark crevices. The first evidence that nest-site scouts are reconfigured foragers came from an experiment conducted by Martin Lindauer. On May 11, 1954, Lindauer set up a colony in a locale east of Munich where flat felds stretch to the horizon and few trees and houses offer nesting cavities. There was, however, plentiful forage for the bees and within a week they began filling the combs in their hive with brood, pollen, and honey. Lindauer expected them to swarm shortly and eventually they did so, casting a swarm on May 27. Ten days before this, on May 17, Lindauer had set up a table 250 meters (820 feet) from the hive and on it he had placed a feeder filled with rich sugar syrup (granular sucrose dissolved in honey). In a few days, he had more than 100 bees from his hive foraging eagerly at his feeder, and he had each one labeled with paint marks for individual identification. Next, on May 22, he placed two artificial nest sites beside the feeding table: a straw skep and a wooden hive (fig. 4.8). Over the following few days, Lindauer started to see curious changes in the behavior of the bees visiting his feeder. First the eagerness of their foraging decreased. Fewer and fewer of his labeled bees made trips to the feeder, and those that kept coming visited less and less often. Sometimes they sucked only hesitantly at the rich sugar syrup. Eventually, on the morning of May 25, Lindauer noticed that his “foraging bees made only a pretense of coming to the feeding dish. They did sip very briefly at it, but then they flew up and buzzed around in the near vicinity for some time.” A knothole in a nearby oak tree drew their attention, as did the two artificial nest sites Lindauer had provided. Over the afternoon, six of his labeled foragers (Bee 73, Bee 100, Bee 106, Bee 113, Bee 119, and Bee 156) conducted 15 inspections of the skep and eight of the hive. There could be no doubt: some of his foragers had become scouts!


The second indication that nest-site scouts are converted foragers comes from a study conducted by Dave Gilley, the gifted undergraduate student who joined my laboratory and quickly fell in love with the bees. To earn an honors degree at Cornell, a student majoring in biology must write a senior thesis based on original research. Dave approached me in the spring of his junior year about attempting an honors thesis project with the bees. I suggested he probe further the mystery of who becomes a nest-site scout, and he happily accepted. Lindauer had shown that some scouts were previously foragers. Dave wanted to see if all or most scouts were previously foragers. If so, then the scouts should be among the oldest bees in a swarm, for it is well established that foragers are the oldest bees in a hive. Dave tested this prediction by setting up five small colonies of bees in early May 1996, and every three days, from May 5 to July 22, adding to each one a cohort of 100 bees that had just emerged from brood combs held in an incubator (0-day-old bees). All the bees in each age cohort were labeled with a paint dot of a particular color indicating the group to which they belonged. Over the next several weeks, as Dave filled the colonies with colorful bees, the bees filled their combs with brood, pollen, and honey. Then, one by one in June and July, the colonies swarmed. Once a swarm had settled into a cluster outside the laboratory building, Dave would watch it patiently for paint-marked bees performing dances, and each time he saw such a bee he would record her age and give her another paint mark (to avoid counting this bee again). Once he had sighted 50 or so nest-site scouts of known age, he collected the entire swarm, narcotized the bees with carbon dioxide before placing them in a freezer, and finally picked over the dead bees to count how many were in each age cohort. These counts enabled him to calculate the age distribution of the nest-site scouts that would be expected if they were drawn randomly from among all the known-age bees in the swarm. Figure 4.9 shows typical results from one swarm. We see that the nest-site scouts included many more older—that is, forager age—bees than would be expected if the known-age scouts had been drawn at random from the pool of known-age bees in the swarm. These findings support the idea that scouts come largely, if not entirely, from the ranks of a colony’s foragers. Both scouts and foragers make long-distance excursions from a central location (swarm or hive) and then must find their way home, so it is easy to imagine that bees with foraging experience make the best scouts.


Having foraging experience evidently prepares a bee for the special job of house hunting. This certainly cannot be the whole story, however, because many foragers never go searching for real estate opportunities. We now know that having certain genes also predisposes a bee to serve as a nest-site scout. Biologists have shown repeatedly, and in many animal species, that differences in behavior among individuals arise from differences in both their genes and their experiences, so it is not surprising that scouts and nonscouts in honeybee swarms differ in “nature” (genes) as well as “nurture” (experience). The need to have the right genetic stuff in order to become a scout bee was shown by two behavioral geneticists, Gene E. Robinson and Robert E. Page Jr., now professors at the University of Illinois (Urbana-Champaign) and Arizona State University. They established three colonies, each of which was headed by a queen that had been instrumentally inseminated with the semen of three unrelated drones (A, B, and C). The three sperm donors for each queen carried distinct genetic markers so that the investigators could determine which drone (A, B, or C) fathered any given worker in a colony. Robinson and Page then prepared artificial swarms (the method is explained below) from their colonies, set these swarms up outdoors, and collected about 40 scouts (dancers) and 40 nonscouts (nondancers) from each swarm. Finally, they conducted a paternity analysis of each collected bee and statistically analyzed their findings to see if the offspring of some drones were more likely to be scouts than were the offspring of other drones. In two of the three swarms they found that yes, the offspring of the three drones differed dramatically in the likelihood of becoming a nest-site scout. For example, in one swarm, one drone fathered over 60 percent of the scout bees even though he fathered less than 20 percent of the worker bees overall. One wonders what it was about this drone’s genes that gave his daughters their proclivity to set out as enterprising house hunters, going boldly where no bee had gone before.

Of course, it is only when a colony is in swarming mode that some foragers, especially those endowed with genes fostering exploratory behavior, adopt the special role of house hunter. How do these bees know when it’s the right time to change their occupation from forager to scout? One hint of how the bees might do this comes from what we humans must do to concoct an artificial swarm for an experimental study, such as the one just described by Gene Robinson and Rob Page. Basically, this is a matter of rendering a queen and a contingent of her workers homeless but not hungry. To do so, you first search through a hive of bees until you locate the queen and then you sequester her safely in a matchbox-size “queen cage.” Next, using a large funnel, you shake several thousand worker bees off the hive’s combs and into a shoebox-size “swarm cage” that has bottom, top, and ends made of wood, but sides made of window screen (for ventilation). At this point, you suspend the queen cage inside the swarm cage, so the worker bees have their queen, and you close the top of the swarm cage, so the bees are contained. Finally, you feed the caged bees lavishly by brushing sugar syrup onto the screen sides of the swarm cage. It is absolutely essential to feed the bees until they are sated and then to keep them full of food for several days. If you don’t, when you shake the worker bees from the cage you will see that the workers cluster conveniently wherever you have mounted the queen (still confined in the queen cage), but you will not see scout bees springing into action. I know this from failuresexperienced firsthand. When I began making artificial swarms, I sometimes made the mistake of not feeding a swarm sufficiently before setting it up. Then I would sit beside the swarm for days waiting for the scout bees to start their dancing, wondering why they didn’t. It seems that a critical stimulus for inducing a food collector to transform herself into a house hunter is that her stomach has been filled with food for a few days.

Lindauer observed this transformation from hungry forager to sated scout during the study described above that he performed in May 1954. The colony used in this study was housed in a glass-walled observation hive so Lindauer could watch the behavior of foragers both inside and outside the hive. When he established his sugar water feeder 250 meters from the hive on May 17, there was little natural forage available, and the bees that found his feeder danced vivaciously when they came home laden with sugar syrup. Over the next few days he labeled more than 100 foragers at the feeder. From May 22 on, however, the flowers of horse chestnut trees (Aesculus hippocastanum)provided plentiful nectar, the bees gradually filled their hive’s combs with honey, and when Lindauer's foragers returned home they had difficulty finding hive bees who would accept their loads of sugar syrup. It is now well established that when returning foragers experience difficulty unloading their nectar, they lose their zest for dancing and foraging. In the extreme condition of a strong colony that has its combs filled with brood and food (hence it is primed to swarm), it is likely that foragers will find it impossible to unload their nectar and will linger at home with their stomachs bulging. This forced inactivity may stimulate a few foragers, those who are constitutionally inclined to explore, to turn to nest-site scouting. I find it extremely suggestive that Lindauer started seeing some of his labeled foragers exploring his nest sites, not exploiting his feeder, a few days after he started noticing most of his previously active foragers sitting around idly, either in some quiet spot inside the hive or in the “beard” of bees hanging outside the entrance. Anecdotal observations like these are the perfect springboard for an experimental investigation designed to test conclusively whether it is a persistently full stomach per se, or something else associated with forced indolence, that informs foragers to become scouts. Students take note.