Content uploaded by Thomas Seeley
Author content
All content in this area was uploaded by Thomas Seeley on Dec 16, 2014
Content may be subject to copyright.
Behav Ecol Sociobiol (2003) 54:511–520
DOI 10.1007/s00265-003-0664-6
ORIGINAL ARTICLE
Thomas D. Seeley · P. Kirk Visscher
Choosing a home: how the scouts in a honey bee swarm perceive
the completion of their group decision making
Received: 13 February 2003 / Revised: 6 June 2003 / Accepted: 24 June 2003 / Published online: 22 July 2003
Springer-Verlag 2003
Abstract This study considers the mystery of how the
scout bees in a honey bee swarm know when they have
completed their group decision making regarding the
swarm’s new nest site. More specifically, we investigated
how the scouts sense when it is appropriate for them to
begin producing the worker piping signals that stimulate
their swarm-mates to prepare for the flight to their new
home. We tested two hypotheses: “consensus sensing,”
the scouts noting when all the bees performing waggle
dances are advertising just one site; and “quorum
sensing,” the scouts noting when one site is being visited
by a sufficiently large number of scouts. Our test involved
monitoring four swarms as they discovered, recruited to,
and chose between two nest boxes and their scouts started
producing piping signals. We found that a consensus
among the dancers was neither necessary nor sufficient
for the start of worker piping, which indicates that the
consensus sensing hypothesis is false. We also found that
a buildup of 10–15 or more bees at one of the nest boxes
was consistently associated with the start of worker
piping, which indicates that the quorum sensing hypoth-
esis may be true. In considering why the scout bees rely
on reaching a quorum rather than a consensus as their cue
of when to start preparing for liftoff, we suggest that
quorum sensing may provide a better balance between
accuracy and speed in decision making. In short, the bees
appear to begin preparations for liftoff as soon as enough
of the scout bees, but not all of them, have approved of
one of the potential nest sites.
Keywords Apis mellifera · Group decision making ·
Honey bees · Nest-site selection · Quorum sensing
Introduction
Although one ordinarily thinks of decision making as a
process performed by individuals, sophisticated acts of
decision making are also performed by groups, most
notably those groups that have evolved a high level of
functional organization, such as colonies of army ants,
honey bees, and other social insects (reviewed by Bourke
and Franks 1995; Camazine et al. 2001). The decision-
making problems faced by animal groups include choos-
ing where to forage (Beckers et al. 1990; Seeley 1995;
Biesmeijer and Ermers 1999; Detrain et al. 1999), when
and where to move (Kummer 1971; Franks and Fletcher
1983; Prins 1996; Boinski and Garber 2000), which route
to take to a given destination (Goss et al. 1989; Strickland
et al. 1992), when and how to build a communal nest
(O’Donnell and Jeanne 1990; Franks et al. 1992; Jeanne
1996; Pratt 1998; Theraulaz et al. 1999), when to become
active (Cole and Trampus 1999), and whether or not to
attack another group (Hlldobler 1982; Adams 1990;
Boehm 1992).
Nest-site choice by a honey bee swarm is an impres-
sive example of group decision making. Swarming
usually occurs in the spring when a colony outgrows its
hive and divides itself by swarming (the biology of
swarming is reviewed in Michener 1974 and Winston
1987). The mother queen and approximately half the
worker bees leave the parental nest to establish a new
colony, while a newly reared daughter queen and the
remaining workers stay behind to perpetuate the old
colony. The swarm bees leave en masse, forming a cloud
of bees just outside the parental hive, but within about
20 min they coalesce into a cluster at an interim site,
usually a nearby tree branch. From here they choose their
future nest site. Several hundred scout bees fly from the
swarm cluster and search out tree cavities and other
potential dwelling places. The dozen or so scouts that find
Communicated by M. Giurfa
T. D. Seeley ())
Department of Neurobiology and Behavior,
Cornell University,
Ithaca, NY, 14853, USA
e-mail: tds5@cornell.edu
Fax: +1-607-2544308
P. K. Visscher
Department of Entomology,
University of California,
Riverside, CA, 92521, USA
suitable cavities report these locations by means of
waggle dances on the surface of the swarm, and other
scouts decode the dances, visit the sites themselves, and
may dance in turn. A process of recruitment and selection
ensues in which one site comes to dominate in visitation
and dancing, and the swarm takes flight again and moves
to the selected cavity.
One part of the decision-making process that remains
mysterious is how the scouts sense when a decision has
been reached. Scouts apparently do sense this, and change
their behavior, beginning to produce a vibrational signal
(the wings-together form of worker piping) an hour or so
before the swarm takes flight to move to the chosen
nesting cavity. This signal stimulates the other bees in the
swarm to warm their flight muscles to a flight-ready
temperature (35C) in preparation for the move to the
new home (Heinrich 1981; Seeley and Tautz 2001; Seeley
et al. 2003). However, we do not know how the scouts
sense when a decision has been made and thus when to
begin this flight-preparatory piping. Solving this mystery
was the objective of our research.
How might the scouts sense when a decision has been
made? One hypothesis, which has been frequently stated
(Lindauer 1955; Seeley and Buhrman 1999; Visscher and
Camazine 1999), is what we call the “consensus sensing
hypothesis”: the scouts note when all the bees performing
waggle dances are advertising just one site.
1
In support of
this hypothesis is the observation that swarms generally
move to their new dwelling only after the dancers have
reached a unanimous agreement. However, Lindauer
(1955) reported that 2 out of the 19 swarms that he
studied lifted off when there were two coalitions of
dancers advertising different sites. This suggests that
unanimity among the dancing scouts may not be neces-
sary for the scouts to sense that a decision has been made,
though it is also possible that these two liftoffs were
simply anomalies. Lindauer’s (1955) report of an occa-
sional liftoff with dissent, plus our own observations of
scout bees becoming numerous at chosen nest sites, led us
to a second hypothesis. This is the “quorum sensing
hypothesis,” related to quorum sensing in bacteria
(Shapiro and Dworkin 1997), and studied already in ants
(Pratt et al. 2002). In the quorum sensing hypothesis, the
scouts note when a site is being visited by a sufficiently
large number of scouts. Shortly before a honey bee swarm
lifts off to fly to its future home site, there are usually
numerous scouts at the chosen nest cavity, with ten or
more outside and probably even more inside (Seeley et al.
1979; Visscher and Camazine 1999; Seeley and Buhrman
2001). It should be noted that the consensus sensing
hypothesis proposes that the scouts sense a critical
variable at the swarm cluster, while the quorum sensing
hypothesis proposes that they sense a critical variable at
the nest site.
To test the two hypotheses, we recorded events at the
swarm cluster and at each of two nest boxes simulta-
neously as a swarm discovered and chose between the
nest boxes and started preparing for liftoff. Because it is
the scouts that initiate the preparations for liftoff, and the
shift to piping behavior reliably occurs shortly before
swarm liftoff, it seems reasonable to assume that this shift
in behavior indicates that the scouts have sensed that a
decision has been reached. Our aim was to observe
whether worker piping (i.e. preparation for liftoff) began
as soon as numerous scouts were visiting one of the nest
sites (even if many dances were still occurring for
alternative sites), as predicted by the quorum sensing
hypothesis, or whether liftoff preparations began only
after a consensus in dancing had been reached (often long
after scout numbers had built up at the favored nest site),
as predicted by the consensus sensing hypothesis. The
first result would contradict the consensus sensing
hypothesis while the second would contradict the quorum
sensing hypothesis.
Methods
Study site
The work was conducted at the Shoals Marine Laboratory on
Appledore Island, Maine (42580N, 70370W). This 39-ha island is
nearly treeless and bears only a few buildings, hence it has few
natural nest sites for honey bees. Here we could set out our nest
boxes and be confident that they would receive attention from the
scout bees of our swarms.
Swarm preparation
All our swarms were artificial swarms prepared from colonies kept
on the mainland. These colonies were headed by “Buckfast” queens
(Adam 1987). In making an artificial swarm, we first located a
colony’s queen and put her in a small cage (3.2101.6 cm). Then,
using a large funnel, we shook 1.0 kg of worker bees (some 7,500
bees, Mitchell 1970) from the frames of this colony’s hive into a
swarm cage (152535 cm) made of wood with wire-screen sides.
We also placed the caged queen inside the swarm cage. The caged
bees were then ferried to Appledore and kept in the shade for 48–
72 h (until copious wax scales appeared beneath the swarm cage).
During this time, we fed the bees ad libitum with a sucrose solution
(1:1 by volume, granulated sucrose:water). Finally we opened the
swarm cage and fastened the queen (still in her own little cage) to a
swarm mount (see Apparatus), and shook the workers onto the base
of the mount. Within an hour, the workers clustered around the
queen and behaved like a natural swarm. Prepared this way,
workers eventually choose a nest site, take off, and start to move
together to their new home. However, because in most cases we
kept the queen caged at the swarm mount, the workers were not
able to complete their move and returned to the swarm mount to
recluster around their queen.
Apparatus
Swarms were placed on a swarm mount that has been described
previously (see Fig. 1 in Seeley and Buhrman 1999). This mount
consists of a vertical board, on which the swarm clusters, and a wire
1
In naming our hypotheses, we have chosen the words “consensus”
and “quorum” based on their definitions in the Oxford English
Dictionary. Consensus: agreement in opinion; the collective
unanimous opinion of a number of persons. Quorum: a fixed
number of members of any body, society, etc., whose presence is
necessary for the proper or valid transaction of business.
512
screen (of 8-mesh hardware cloth with several passageways for the
bees cut into it) mounted vertically over the swarm’s surface so that
the outermost layer of the swarm is on the outside of the screen.
This apparatus facilitated video recording the scout bees’ dances,
which was done to determine when the dances became unanimous
for one of the nest boxes. Our video equipment consisted of an S-
VHS camera (Panasonic WV-F250B) and videocassette recorder
(Panasonic AG-7450) equipped with a time-code generator (Pana-
sonic AG-F745). The videotapes were analyzed using a videocas-
sette player with variable-speed playback (JVC BR-S525U).
To record worker piping within the swarm cluster, we mounted
two small, 5-mm-diameter, custom-made microphones (flat fre-
quency response from 20 to 6000 Hz), one on each side of the
swarm, on the rods supporting the screen of the swarm mount. This
positioned each microphone deep inside the swarm cluster. The two
microphones were connected to an amplifier whose two channel
output was recorded with a stereo digital minidisc recorder (Sony
MZ-R37SP).
To record the ambient temperature near the swarm cluster, we
used a copper-constantan thermocouple probe and a digital
thermometer (Bailey Bat-12). The probe was mounted on the
swarm mount, 2 cm below the board on which the swarm was
clustered.
The two nest boxes used in this study were specially designed
and constructed to enable us to count the scout bees at a potential
nest cavity, both outside and inside the cavity. Each was cube-
shaped, built of 2.0-cm-thick plywood, and sized to provide a 27-l
nesting cavity. Each had one entrance opening, a 2.0-cm-diameter
hole centered on one side wall, and had one of its side walls left
open. And each was bolted to the side of an observation hut, with
the open side aligned with a matching opening in the wall of the
hut. A sheet of 3.2-mm-thick glass provided a window-wall
between the nest box and observation hut. Each observation hut
(244113113 cm) was constructed of wood with lightproof joints
and had its inner walls painted flat black. Thus it provided a dark
viewing chamber from which we could scan the interior of the
attached nest box for scout bees. The entrance hole of the nest box
admitted sufficient light for observing the bees.
Experimental layout and data collection
We began each trial by placing a swarm on the mount located on
the covered porch of Bartel’s Hall and giving the bees sugar water
ad libitum from two feeder bottles. We had already positioned the
two nest boxes in locations that were the same distance (250 m), but
in different directions (angular separation of 58), from the swarm.
One was near Broad Cove while the other was 240 m away, in the
vicinity of Devil’s Glen (we used sites 2 and 5 shown in Fig. 2 of
Seeley and Buhrman 2001). To help the scouts locate the nest boxes
at the start of a trial, we inserted in the entrance opening of each
nest box a pheromonal swarm lure (Ecogen, Langhorne, Pa.)
suspended from a wire. Once the bees were settled on the swarm
mount, an observer was stationed at each nest box to record when it
was discovered by a scout bee and to be present for the start of data
collection.
Immediately upon observing a scout bee at either nest box, we
took several steps to begin data collection. First, the pheromone
lure was pulled from the newly found nest box and stored in an
odor-proof glass jar. Second, the observer at each nest box began
following a standard protocol for measuring the number of scout
bees at his or her nest box. Every 5 min, the person would make
four separate counts of the maximum number of bees seen
simultaneously at the nest box over a 30-s period, with two counts
based on the bees seen inside the nest box and two counts based on
the bees seen outside. When the number of bees inside or outside
rose above ten, precise counts were impossible, and the observer
noted the number of bees as a range (e.g. 20–25 bees). Third, the
observer at the swarm cluster turned on the video camera, so that
the dances of the scouts would be recorded from their start, and he
began following a standard protocol for measuring the level of
worker piping within the swarm. Every 5 min, he would make an
audio recording, lasting at least 60 s, using the stereo microphones
mounted inside the swarm cluster. He would also begin recording at
5-min intervals the ambient temperature of the swarm cluster.
Meteorological data were recorded automatically by a weather
station on Appledore Island that is operated by the National
Oceanographic and Atmospheric Agency.
Data analysis
The video recordings were analyzed to determine the number of
waggle runs performed for each nest box over 20-min intervals
before liftoff. The nearly 60 difference in direction to the two nest
boxes made it possible to tell which nest box was represented by
each dance. We made our counts of the waggle runs by reviewing
each video recording and, whenever a bee began to dance, noting
the time (indicated on the video record) and the location (specified
using a system of quadrats on the video monitor) at which this bee
began to dance, and then counting the waggle runs contained in this
bee’s dance. By noting the starting time and starting location of
each dance, we avoided double counting the waggle runs contained
in any dance, even though sometimes 10 or more dances were
performed simultaneously.
The audio recordings were analyzed to get estimates of the
percent time that worker piping was heard, at 5-min intervals before
liftoff. From each 60-s audio recording that we had made, we took
three 20-s intervals, and for each of these intervals we counted the
number of 1-s subintervals (N) in which worker piping was audible.
From these counts we calculated the percent time that piping was
audible during each 20-s interval ( N/20100). Finally, using the
three measurements for each sampling period, we calculated the
mean value of the percent time that piping was heard.
Results
Swarm 1
This swarm was placed on the swarm mount at 1800 hours
on 23 June 2002. Two days later, at 1620 hours, a scout
bee discovered the Broad Cove nest box and spent the
next 62 min there inspecting it excitedly. This bee,
however, performed no dances for this nest box on this
day, perhaps because she discovered it rather late in the
day.
Figure 1 depicts the results obtained the next day (26
June). The Broad Cove nest box was visited by a scout
bee starting at 0900 hours, but no dances were performed
for this site until 1101 hours. By 1120 hours, several scout
bees were visiting both the Broad Cove and the Devil’s
Glen nest boxes and strong dances were being performed
for both sites. By 1150 hours, there was noticeable piping
at the swarm cluster. The start of this piping came shortly
after the number of scouts at the Devil’s Glen nest box
had risen to approximately 15 total (about 12 inside and 3
outside), the number of scouts at the Broad Cove nest box
was at most 2 total, and nearly all the dancing had been
for the Devil’s Glen nest box. (For the 20-min period of
1126–1146 hours, 137 waggle runs were produced for the
Devil’s Glen nest box but only 4 for the Broad Cove nest
box). From 1146 to 1406 hours, when the swarm lifted off
and started to move to the Devil’s Glen nest box, the
number of scouts at the Devil’s Glen nest box remained
fairly steady at about 10–15 total, the number of scouts at
the Broad Cove nest box remained low except for a brief
513
time around 1300 hours when it rose to about 10 total, and
the piping persisted at a moderate level until the last half
hour before liftoff, when it rose to a high level. Curiously,
this swarm completed its decision making and performed
a liftoff without having a prolonged period of unanimous
dancing for the chosen site.
In summary, in this swarm the liftoff preparations
(worker piping) began at a time when the number of bees
at the ultimately chosen nest box had risen to 10–15 and
when the dancing had been, for about 20 min, virtually
unanimous for the ultimately chosen nest box. Liftoff
occurred without unanimity among the dancers.
Swarm 2
This swarm was placed on the swarm mount at 1900 hours
on 29 June 2002. On 2 July, at 1241 hours, a scout bee
discovered the Broad Cove nest box and at 1256 hours the
dancing for this nest box began. Somewhat earlier, at
1207 hours, dancing had also begun for an unidentified
site in a southerly direction, hence nearly 180 from our
nest boxes, probably in one of the old houses on the south
shore of the island.
Figure 1 depicts the results for the remainder of the
day, until liftoff at 1526 hours. The number of scout bees
at the Broad Cove nest box rose rapidly following its
discovery. Likewise, the first sounds of piping workers
were heard soon after the number of scouts at the Broad
Cove nest box had begun to grow. The Devil’s Glen nest
box was never discovered, so the counts of scouts there
remained at zero. The dancing was, naturally, unanimous
for the Broad Cove nest box, except when some dances
were also performed for the Old House site. It should be
noted that in this swarm, like in swarm 1, liftoff was not
preceded by a period of unanimous dancing for the chosen
site, though the dancing had been primarily for the Broad
Cove site. Shortly after liftoff, the swarm moved in the
direction of the Broad Cove nest box.
In summary, in this swarm as in the previous one,
liftoff preparations (worker piping) began at a time when
the number of bees at the ultimately chosen nest box had
become considerable (30–40 bees in this case), and when
the dancing was unanimous for the ultimately chosen nest
box. And, as with the first swarm, liftoff occurred without
there being unanimity among the dancers.
Swarm 3, day 1
This swarm was placed on the swarm mount at 1900 hours
on 5 July 2002. On 6 July, at 1040 hours, a scout bee
discovered the Devil’s Glen nest box. A few minutes
later, at 1106 hours, a scout bee discovered the Broad
Fig. 1 Results from monitoring
two swarms as their scout bees
chose between two potential
nest sites and then began
preparing for liftoff (indicated
by the start of worker piping).
At the swarm, we recorded the
intensity of worker piping every
5 min and the number of waggle
runs that were produced for
each site in 20-min intervals
before liftoff. At the potential
nest sites, we recorded the
number of scout bees outside
and inside each site every
5 min, except when one of the
sites was an unidentified loca-
tion in an old house
514
Cove nest box. The first dances for the Devil’s Glen and
the Broad Cove sites were recorded at 1223 and
1238 hours, respectively.
Figure 2 depicts the results obtained from swarm 3
over the next 2 days. Worker piping was first heard at
1255–1300 hours, shortly after the number of bees at one
of the nest boxes (Devil’s Glen) had risen to a moderately
high level, 15–20 bees, and the number at the other nest
box was still at a low level, only 0–5 bees. Then, for some
unknown reason, the number of bees dropped to almost
zero at both nest boxes. This lasted from 1300 to
1340 hours at the Devil’s Glen nest box and from
1300 hours to the end of the day at the Broad Cove nest
box. Simultaneously, the worker piping dwindled. It
started up again in the period 1350 to 1400 hours, which
is when the number of bees at the Devil’s Glen nest box
had returned to a moderately high level, 15–20 bees.
Although the number of scout bees at the Devil’s Glen
nest box remained high over the next 2 h of the afternoon,
peaking at 40–50 bees around 1600 hours, the worker
piping ceased suddenly when the sky darkened with
thunderstorm clouds at 1445 hours (shown in the inset
graph). No more piping was heard until 1520 hours, at
which time the sky brightened and the piping rose rapidly
to its highest level of the day. By 1540 hours, however,
the sky had again filled with dark storm clouds and
shortly thereafter the piping again ceased, even though
bees remained in abundance at the Devil’s Glen nest box
and strong, unanimous dancing persisted for this nest box.
In summary, on this day this swarm showed three
times when the worker piping started and three times
when it stopped. All three times when the piping started
were times when the number of scout bees at the Devil’s
Glen nest box had risen to at least 15–20 bees. Impor-
tantly, the first and second starts of piping occurred at
times when the dancing was nearly evenly divided
between the two nest boxes. Only the third start of piping
occurred when the dancing was unanimous for the Devil’s
Glen nest box. The first stop of piping occurred when the
number of bees at both nest boxes had dropped nearly to
zero. The second and third stops occurred when the
number of bees at the Devil’s Glen nest box remained
high but the weather had turned poor.
Swarm 3, day 2
This swarm resumed its decision making the next day,
which started out cloudy and cool but by 1000 hours had
become sunny and warm. As is shown in Fig. 2, the
number of scout bees rose rapidly at both nest boxes
starting around 1100 hours, and stayed high at both until
Fig. 2 Results from monitoring
one swarm over 2 days as its
scout bees chose between two
potential nest sites and then
began preparing for liftoff. Fig-
ure format follows that of
Fig. 1, except on day 1 when we
also include records of solar
radiation ( inset) because on this
day changes in the weather
greatly influenced the results
515
shortly before liftoff at 1204 hours. Because of a
malfunction in our recorder, we do not know when
exactly the piping started, but we do know that there was
no piping at 1100 hours and that by 1140 hours it had
reached a high level. Hence the piping must have started
sometime in the interval 1100–1140 hours, during which
the number of bees at each nest box was at least 10–20
bees. There was also strong dancing for both nest boxes
throughout this interval, and this division among the
dancers persisted until the liftoff, though in the last 20-
min period preceding liftoff the dancing was much
stronger for the Broad Cove nest box than for the Devil’s
Glen one (2,824 vs. 360 waggle runs).
By 1148 hours it was clear from the intense piping that
liftoff was imminent. Because this swarm still had
numerous scouts at both of the nest boxes, and so had
its “attention” split between the two sites, we wondered if
it would be able to move successfully to one of them. To
find out, we released the queen from her cage, whereupon
she began to walk rapidly over the surface of the swarm
cluster. At 1204 hours the swarm lifted off and the queen
was seen to be among the earlier bees to take flight. Once
in flight, the bees formed an unusually large cloud of bees
split by Bartel’s Hall, with some bees on the south side of
the building (the direction to Devil’s Glen) and some on
its north side (the direction to Broad Cove). The majority
of the bees seemed to be on the north side, and at
1209 hours these bees began moving slowly northwest, in
the direction of the Broad Cove nest box. Their movement
stopped, however, at 1211 hours, when the cloud they
formed had traveled only about 40 m from the swarm
mount. Simultaneously, a cluster of bees, apparently
composed initially of bees that had been on the south side
of Bartel’s Hall, started to form back at the swarm mount.
By 1215 hours, about half of the swarm bees had resettled
on the swarm mount and the queen was seen, having
alighted nearby. No bees arrived at the Devil’s Glen nest
box and only about 10 reached the Broad Cove nest box.
The swarm quickly reorganized itself for a second
liftoff followed by a successful move to the Broad Cove
nest box. By 1230 hours, the swarm cluster was rebuilt,
scout bees had begun to reappear at both nest boxes, and
dances were again being performed for both nest boxes
(see Fig. 2). Unlike in the morning, however, the buildup
of scouts at the two nest boxes was not balanced; the
counts at the Broad Cove nest box greatly exceeded those
at the Devil’s Glen nest box. Likewise, the dancing was
mainly for the Broad Cove nest box. Although we do not
know when exactly the worker piping restarted, we can
say that it was sometime between 1220 (when no piping
was heard) and 1240 hours (when much piping was
heard), hence it occurred when there were at least 10 bees
at the Devil’s Glen nest box and at least 40 bees at the
Broad Cove nest box. Finally, at 1313 hours, the swarm
lifted off again and moved to the Broad Cove nest box.
In summary, on this day this swarm performed two
liftoffs. In both cases, the start of the liftoff preparations
(worker piping) began when the number of scout bees at
both nest boxes had risen to at least 10–20 bees. The first
round of liftoff preparations began without a dance
consensus while the second round began when nearly all
the dances were for the chosen nest box. As for the actual
liftoffs, only the second one occurred when there was
clear unanimity among the dancers.
Swarm 4
This swarm was placed on the swarm mount at 1930 hours
on 12 July 2002. On 14 July, at 0928 hours, one scout bee
discovered the Broad Cove nest box and around
1030 hours two more bees evidently also discovered it,
for up to 3 bees were seen simultaneously there. At
1117 hours the dancing for this nest box began.
Figure 3 shows the results for the remainder of the day.
Shortly after the dancing began for the Broad Cove nest
box, the number of bees there began to rise. When it
reached a total of about 20 bees (15 inside and 5 outside),
at 1210 hours, the piping started. But when the number of
bees fell well below this level, at about 1220 hours, the
piping stopped. By 1240 hours the number of bees had
returned to a moderate level of about 25 bees (20 inside
and 5 outside), and the piping had resumed. Then, once
again, the number of bees at the Broad Cove nest box
dropped off, falling to only about 10 bees total by
1300 hours, and the piping again stopped. By 1325 hours
the count had risen again to about 20 bees (15 outside and
Fig. 3 Results from monitoring one swarm as its scout bees
discovered one potential nest site and then began preparing for
liftoff. Figure format follows that of Fig. 1
516
5 inside) and the piping again became audible. Yet once
more, starting at about 1330 hours, the number of bees at
the nest box began to decline and this was matched by a
decline in the piping, but this time neither variable fell to
a low level. Finally, starting at 1340 hours, the number of
scouts at the nest box surged to its highest level, some 35
bees total, the piping became loud and continuous, and
ultimately the swarm lifted off at 1403 hours. Throughout
this time, no scout bees appeared at the Devil’s Glen nest
box, which had been open and available until 1205 hours
when it was closed so we could concentrate on making
observations at the Broad Cove nest box. Also throughout
this time virtually all the dancing was for the Broad Cove
nest box; only occasionally was a dance performed for
some other, unidentified site.
In summary, this swarm seriously considered just one
of our nest boxes and so had unanimous dancing for this
site throughout its decision-making process. Also, this
swarm showed three times when the worker piping started
and two times when it stopped. All three times when the
piping started were times when the number of scout bees
at the Broad Cove nest box had risen to at least 20 bees,
and both times when the piping stopped were ones when
the number of bees at the nest box had fallen to only about
10 bees.
Discussion
Knowing when to start liftoff preparations
The main aim of this study was to learn how the scouts in
a honey bee swarm know when a decision has been
reached regarding the new home site, that is, when it is
appropriate for them to begin producing the piping signals
that will stimulate their swarm-mates to prepare for the
flight to their new home. Our results, summarized in
Table 1, enable us to draw two conclusions regarding this
puzzle:
1. The consensus sensing hypothesis is false. We ob-
served that a consensus among the dancers was neither
necessary (as shown by the results of swarm 3 on both
days 1 and 2) nor sufficient (as shown by the results of
swarm 4) for the start of worker piping.
2. The quorum sensing hypothesis may be true. We
observed that a buildup of 10–15 or more bees at one
of the nest boxes was consistently associated with the
start of working piping (as shown by the results of
swarms 1, 2, and 3, and especially of swarm 4).
Our observations are consistent with what has been
reported previously. Prior observations suggested that the
consensus sensing hypothesis is false. Lindauer (1955)
described how 2 of the 19 swarms that he studied took off
without their dancers having formed a consensus. Appar-
ently, the scouts in these two swarms, like those in our
swarm 3, did not need to sense a consensus among the
dancers to begin their preparations for liftoff. One could,
however, interpret Lindauer’s report of two swarms lifting
off without consensus as anomalies. We feel that our
finding that a swarm’s liftoff preparations often precede
the formation of a consensus provides thoroughly con-
vincing evidence against the consensus sensing hypoth-
esis.
Prior observations also suggested that the quorum
sensing hypothesis is true. Seeley and Buhrman (2001)
reported that in all five of the swarms that they studied
(see their Fig. 5), the liftoff was preceded by a population
of at least 10 bees outside (and an unknown number
inside) the chosen nest box during the last hour or two
before liftoff. Thus the scouts in each of these five
swarms could have sensed a buildup of 10–20 or more
bees at the chosen site when they began to stimulate their
swarm-mates to begin preparing for liftoff. The pattern of
scout bee buildup outside the chosen site reported by
Seeley and Buhrman (2001) matches the pattern described
earlier for one swarm by Seeley et al. (1979, see their
Fig. 1).
A rigorous test of the quorum sensing hypothesis
remains to be done with honey bee swarms. Although our
finding that the start of worker piping is tightly correlated
with the buildup of scouts at one of the sites (see Fig. 2
and 3) is strongly supportive of the quorum sensing
hypothesis, what needs to be done is an experiment in
which the number of bees at a nest site is manipulated
independently of other variables in the decision-making
process, and it is noted whether or not the scout bees start
their piping when and only when the number of bees at
the nest site exceeds a quorum requirement. Such an
experiment is planned. Already, however, the phenome-
non of quorum sensing has been conclusively demonstra-
ted in a study of group decision making during colony
emigration in the ant Leptothorax albipennis (Pratt et al.
2002). In this species, when a colony’s nest is damaged,
the ants perform a two-stage decision process in choosing
a new home. Initially, a minority of active ants in a colony
search for potential new sites and those that discover such
sites recruit others to their finds via the relatively slow
process of tandem running, in which a single follower is
lead all the way to the new site. The better the site, the
shorter the latency between discovery and start of tandem
Table 1 Summary of the conditions under which worker piping
started during the decision-making processes of four swarms of
bees. For each time that worker piping started on a swarm, we note
whether or not there was a dance consensus at the swarm cluster
and whether or not there was a buildup of scouts at the nest site. In
swarms 1 and 2, the worker piping started just once, but in swarms
3 and 4 the worker piping started several times
Conditions when worker piping started
Swarm Dance consensus? Buildup of scouts?
1 Yes, but temporary Yes, 10–15 bees
2 Yes, but temporary Yes, 30–40 bees
3, day 1 No, no, yes Yes, 15–20 bees each time
3, day 2 No, yes Yes, 10–20 bees each time
4 Yes, yes, yes Yes, 20+ bees each time
517
running, hence the more rapid the buildup of ants at the
site (Mallon et al. 2001). Eventually, the population of
ants at one of the prospective sites (usually the best site,
given the quality-dependent delays in initiation of
recruitment) exceeds a quorum requirement and the
active ants begin rapidly recruiting the passive majority
of the colony to this site via transports, in which
nestmates are simply carried. By performing experiments
with just one nest site, and in which they manipulated the
number of ants in this site, Pratt et al. (2002) were able to
demonstrate that ants use a quorum of nestmates (approx.
9–17 individuals) as the cue indicating when they should
begin bringing others to the site using the faster, transport
method of recruitment. Future studies in other social
insects may reveal that quorum sensing is an important
and widespread feature of the functional organization of
colonies.
In the context of group decision making by honey bee
swarms, we have learned that the consensus sensing
hypothesis is false, and that the quorum sensing hypoth-
esis may be true. If we tentatively assume that the quorum
sensing hypothesis is true in honey bee swarms, we face
intriguing questions of both behavioral mechanism and
functional design. With respect to behavioral mechanism,
there is the question of how a scout bee senses the number
of other bees at a nest site. One possibility is by visual
perception. For a human observer, 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. Measurements of the light level inside a model nest
cavity found light levels of 1–2 lx near the entrance
opening, which is where most of the traffic of scout bees
occurs, though less than 0.5 lx elsewhere inside the cavity
(see Fig. 15 in Seeley 1977). The threshold level of
illumination for bee flight (hence good vision) is approx-
imately 1.0 lx (Schricker 1965). Another possible means
of sensing the number of scout bees at a site is by tactile
perception. It is a curious fact that as soon as a site
acquires multiple scouts, they begin to make frequent
contacts with one another. Many of the scouts even start
to perform buzzing runs (first described by Lindauer
1955; see also Martin 1963 and Esch 1967) on the inner
and outer surfaces of the nesting site. That is, the scouts
make excited, zig-zag runs that are punctuated by bouts of
buzzing their wings and butting against other bees. So it
seems entirely possible that a bee could use the rate of
contacts with scouts in general, or encounters with buzz
runners in particular, as an indicator of the number of
fellow scouts at a site.
With respect to functional design, there is the question
of why bees do not use consensus sensing and appear
instead to use quorum sensing. This question is made
prominent by the fact that a consensus among the dancers
helps a swarm execute a successful move to a new home
site. Indeed, a consensus, or at least a near consensus,
among the dancers may be a requirement of a successful
move. In three reported instances of a swarm lifting off
when its dancers were strongly split between two sites
(Lindauer’s 1955 Balcony and Moosach swarms and our
swarm 3), the airborne swarm divided, stalled in its move,
and resettled. Two of these swarms went on to achieve a
dance consensus and a successful move, but one (Lin-
dauer’s Balcony swarm) lost its queen when it split itself
in midair and so experienced complete failure.
Why do the scouts not use consensus sensing and so
avoid the risk of their swarm suffering a fatal fragmen-
tation upon liftoff? One possibility is that it would be
exceedingly difficult or costly, or both, for the scouts to
sense a consensus among themselves as they perform
dances on a swarm. To do so would require that each
scout devotes much effort to polling her fellow scouts,
presumably by traveling over the swarm cluster, reading
some sample of the dances, and keeping a tally of the
readings. Moreover, the larger the swarm, the more
numerous the scouts, and perhaps the greater the
difficulty (or cost) of consensus sensing. Quorum sensing,
however, need not become more difficult with increasing
swarm size, because the quorum size could be fixed, and
so independent of swarm size. Another possibility is that
consensus sensing would greatly slow the decision-
making process. Franks et al. (2002) have pointed out
that there is a fundamental trade-off in decision making
between speed and accuracy. It seems likely that consen-
sus sensing would provide maximum accuracy in a
swarm’s decision making, for a swarm would not proceed
to liftoff preparations until the dancers were in agreement
over the home site. This maximum accuracy in decision
making would come, however, at the cost of taking extra
time, for in principle even a single scout bee that
mistakenly dances vigorously for an inferior site could
prevent liftoff and so prolong the decision making. There
can be no doubt that there is a cost associated with a
honey bee swarm taking more time to decide. A swarm
clinging to a tree branch is exposed to drenching rains and
is consuming its energy reserves; the 30–40 mg of rich
sugar solution carried inside each bee (Combs 1972).
Quorum sensing may provide a good balance between
accuracy and speed in decision making. With respect to
accuracy, the quorum requirement promotes a high level
of accuracy, for we observed that scout bees will not
begin producing piping signals for liftoff preparations
until the number of bees present at the site surpasses a
threshold of some 10–20 or more bees. Having a quorum
requirement this high helps ensure that liftoff preparations
are not initiated by scouts that have erred, judging a poor
site to be a good one. If a scout makes an error, and
recruits strongly to a poor site, her followers will likely
counter her mistake by judging the site less highly than
she, advertising it with weaker dances (if any), and so
putting off the start of piping. This idea has been
developed rigorously by Pratt et al. (2002) in their study
of group decision making during colony emigration in the
ant L. albipennis. By modeling the decision-making
process, they showed that a moderate quorum require-
ment helps a colony choose the best available site, by
preventing the launch of the rapid transport process to a
given site until numerous individuals have each been
518
convinced of the worth of the site. Thus, even a rather
moderate quorum requirement reduces the chances of
emigration to an inferior site.
With respect to speed of decision making, the
requirement of a quorum, but not a consensus, means
that preparations for liftoff can begin as soon as enough
scout bees have approved of one of the nest sites, even if
some others are still scouting other sites. We suggest that
the quorum size is a detail of the bees’ decision-making
process that has been tuned by natural selection to provide
an optimal balance between accuracy (favored by a large
quorum) and speed (favored by a small quorum). It is
likely that quorum size has also been influenced by the
need to have sufficient scout bees familiar with the
chosen site’s location to pilot the flying swarm to this site.
The use of a quorum requirement does seem to suffer
from at least one major weakness, in that it can lead to
liftoffs without a dance consensus. As discussed above,
such liftoffs do occur occasionally and they can be costly
mistakes. The swarm is apt to split after liftoff, thereby
failing to move to either site while at the same time
expending much energy. It may even lose its queen.
Presumably the reason that most liftoffs occur when there
is a consensus, even though a consensus is not required
for the start of liftoff preparations, is because there is the
strong positive feedback process of vigorous dancing for
the chosen site, which attracts more and more dancing for
this superior site (Seeley and Buhrman 2001). Moreover,
there is the steady attritional process of individuals
ceasing to dance, especially for the inferior, non-chosen
sites (Camazine et al. 1999; Seeley 2003; Visscher 2003).
Knowing whether to continue liftoff preparations
Our results also enable us to draw several additional
conclusions on the topic of what factors determine
whether piping (and thus preparation for liftoff) continues
once it has started. The continuation of piping may
depend on previously piping scouts continuing to pipe, or
on previously non-piping scouts starting to pipe, or both.
1. A consensus among the dancers is not necessary for
the continuation of piping and eventual liftoff (as
shown by the results of swarm 1, swarm 2, and swarm
3 on day 2).
2. A quorum of some 10–20 bees at one of the potential
nest sites appears to be necessary for the continuation
of piping (as shown by the results of swarm 3 on day 1
and swarm 4), except at the end of the period of liftoff
preparations. The number of bees at the chosen site
usually plummets shortly before liftoff and yet liftoff
preparations will continue and liftoff will occur (as
shown by the results of swarm 1, swarm 2, and swarm
3 on day 2).
3. Favorable weather is necessary for the continuation of
piping and eventual liftoff (as shown by the results of
swarm 3 on day 1).
Thus it seems clear that the scout bees, even after some
have started to stimulate their swarm-mates to prepare for
liftoff, remain sensitive to the weather conditions and the
number of bees at the chosen site. If the weather
deteriorates or the number of bees at the chosen site falls
to a low level (at least, during the early stages of liftoff
preparations), the piping will decrease. Such behavioral
flexibility is highly adaptive for it seems clear that only if
both the weather conditions remain favorable and the
nest-site decision remains strongly supported should a
swarm undertake the momentous act of launching into
flight to start its journey to a new home.
Acknowledgements The research reported here was supported by
the U.S. National Science Foundation (grant IBN02–10541), the
National Geographic Society (grant 7055–1), and the UCR
Academic Senate. We thank Marjorie Martin for letting us keep
our bees at her home at Kittery Point, Maine; Siobhan Cully for
spending many hours monitoring the nest box overlooking Broad
Cove; and Dr. James Morin for providing space and facilities at the
Shoals Marine Laboratory. This is contribution no. 113 of the
Shoals Marine Laboratory.
References
Adam B (1987) Beekeeping at Buckfast Abbey. Northern Bee
Books, Hebden Bridge
Adams ES (1990) Boundary disputes in the territorial ant Azteca
trigona: effects of asymmetries in colony size. Anim Behav
39:321–328
Beckers R, Deneubourg J-L, Goss S, Pasteels JM (1990) Collective
decision making through food recruitment. Insectes Soc
37:258–267
Biesmeijer JC, Ermers MCW (1999) Social foraging in stingless
bees: how colonies of Melipona fasciata choose among nectar
sources. Behav Ecol Sociobiol 46:129–140
Boehm C (1992) Segmentary ’warfare’ and the management of
conflict: comparison of East African chimpanzees and patri-
lineal–patrilocal humans. In: Harcourt AH, DeWall FBM (eds)
Coalitions and alliances in humans and other animals. Oxford
University Press, Oxford, pp 137–173
Boinski S, Garber PA (2000) On the move: how and why animals
travel in groups. University of Chicago Press, Chicago
Bourke AFG, Franks NR (1995) Social evolution in ants. Princeton
University Press, Princeton, N.J.
Camazine S, Visscher PK, Finley J, Vetter RS (1999) House-
hunting by honey bee swarms: collective decisions and
individual behaviors. Insectes Soc 46:348–360
Camazine S, Deneubourg J-L, Franks NR, Sneyd J, Theraulaz G,
Bonabeau E (2001) Self-organization in biological systems.
Princeton University Press, Princeton, N.J.
Cole B, Trampus FI (1999) Activity cycles in ant colonies: worker
interactions and decentralized control. In: Detrain C, Deneu-
bourg J-L, Pasteels JM (eds) Information processing in social
insects. Birkhuser, Basel, pp 289–308
Combs GF (1972) The engorgement of swarming worker honey-
bees. J Apic Res 11:121–128
Detrain C, Deneubourg J-L, and Pasteels JM (1999) Decision-
making in foraging by social insects. In: Detrain C, Deneubourg
J-L, Pasteels JM (eds) Information processing in social insects.
Birkhuser, Basel, pp 331–354
Esch H (1967) The sounds produced by swarming honey bees. Z
Vergl Physiol 56:408–411
Franks NR, Fletcher CR (1983) Spatial patterns in army ant
foraging and migration: Eciton burchelli on Barro Colorado
Island, Panama. Behav Ecol Sociobiol 12:261–270
519
Franks NR, Wilby A, Silverman BW, and Tofts C (1992) Self-
organizing nest construction in ants: sophisticated building by
blind bulldozing. Anim Behav 44:357–375
Franks NR, Pratt SC, Mallon EB, Britton NF, Sumpter DJT (2002)
Information flow, opinion polling and collective intelligence in
house-hunting social insects. Philos Trans R Soc Lond B
337:1567–1583
Goss S, Aron S, Deneubourg J-L, Pasteels JM (1989) Self-
organized shortcuts in the Argentine Ant. Naturwissenschaften
76:579–581
Heinrich B (1981) The mechanisms and energetics of honeybee
swarm temperature regulation. J Exp Biol 91:25–55
Hlldobler B (1982) Foraging and spatiotemporal territories in the
honey ant Myrmecocystus mimicus Wheeler. Behav Ecol
Sociobiol 9:301–314
Jeanne RL (1996) Regulation of nest construction behaviour in
Polybia occidentalis. Anim Behav 52:473–488
Kummer H (1971) Primate societies: group techniques of ecolog-
ical adaptation. Aldine, Chicago
Lindauer M (1955) Schwarmbienen auf Wohnungssuche. Z Vergl
Physiol 37:263–324
Mallon EB, Pratt SC, Franks NR (2001) Individual and collective
decision-making during nest site selection by the ant Leptotho-
rax albipennis. Behav Ecol Sociobiol 50:352–359
Martin P (1963) Die Steuerung der Volksteilung beim Schwrmen
der Bienen. Zugleich ein Beitrag zum Problem der Wander-
schwrme. Insectes Soc 10:13–42
Michener CD (1974) The social behavior of the bees. Harvard
University Press, Cambridge, Mass.
Mitchell C (1970) Weights of workers and drones. Am Bee J
110:468–469
O’Donnell S, Jeanne RL (1990) Forager specialization and the
control of nest repair in Polybia occidentalis Olivier (Hymen-
optera: Vespidae). Behav Ecol Sociobiol 27:359–364
Pratt SC (1998) Condition-dependent timing of comb construction
by honeybee colonies: how do workers know when to start
building? Anim Behav 56:603–610
Pratt SC, Mallon EB, Sumpter DJT, Franks NR (2002) Quorum
sensing, recruitment, and collective decision-making during
colony emigration by the ant Leptothorax albipennis. Behav
Ecol Sociobiol 52:117–127
Prins HHT (1996) Ecology and behaviour of the African buffalo:
social inequality and decision making. Chapman and Hall,
London
Schricker B (1965) Die Orientierung der Honigbiene in der
Dmmerung. Z Vergl Physiol 49:420–458
Seeley TD (1977) Measurement of nest cavity volume by the honey
bee ( Apis mellifera). Behav Ecol Sociobiol 2:201–227
Seeley TD (1995) The wisdom of the hive. Harvard University
Press, Cambridge, Mass.
Seeley TD (2003) Consensus building during nest-site selection in
honey bee swarms: the expiration of dissent. Behav Ecol
Sociobiol 53:417–424
Seeley TD, Buhrman SC (1999) Group decision making in swarms
of honey bees. Behav Ecol Sociobiol 45:19–31
Seeley TD, Buhrman SC (2001) Nest-site selection in honey bees:
how well do swarms implement the “best-of-N” decision rule?
Behav Ecol Sociobiol 49:416–427
Seeley TD, Tautz J (2001) Worker piping in honey bee swarms and
its role in preparing for liftoff. J Comp Physiol A 187:667–676
Seeley TD, Morse RA, Visscher PK (1979) The natural history of
the flight of honey bee swarms. Psyche 86:103–113
Seeley TD, Kleinhenz M, Bujok B, Tautz J (2003) Thorough warm-
up before take-off in honey bee swarms. Naturwissenschaften
40:156–260
Shapiro JA, Dworkin M (1997) Bacteria as multicellular organisms.
Oxford University Press, New York
Strickland TR, Tofts CMN, Franks NR (1992) A path choice
algorithm for ants. Naturwissenschaften 79:567–572
Theraulaz G, Bonabeau E, Deneubourg J-L (1999) The mecha-
nisms and rules of coordinated building in social insects. In:
Detrain C, Deneubourg J-L, Pasteels JM (eds) Information
processing in social insects. Birkhuser, Basel, pp 309–330
Visscher PK (2003) How self organization evolves. Nature
421:799–800
Visscher PK, Camazine S (1999) The mystery of swarming
honeybees: from individual behaviors to collective decisions.
In: Detrain C, Deneubourg J-L, Pasteels JM (eds) Information
processing in social insects. Birkhuser, Basel, pp 355–378
Winston ML (1987) The biology of the honey bee. Harvard
University Press, Cambridge
520
