When I walked into the lecture hall last week, this
image looked like a swarm of bees to me. By the time, I walked out one hour
later, this is what I saw.
No, this was not the effect of a dream state or a
hallucinogen but rather the effect of Dr. Thomas Seeley’s work.
Tim Seeley, as he is known, is an ethologist and an
entomologist from Cornell and probably one of the world’s leading expert in the
phenomena of swarm intelligence. He has been studying the behaviors and social
structures of bees for many decades now. As far as scientific genealogy goes,
Tim has great Pedigree. He began his work on bees, as part of his doctoral
dissertation with Lindauer who was a student with Karl Von Frisch, the Nobel
laureate who pioneered the field itself. It was Von Frisch’s discovery of the
now famous waggle dance in bees, as a tool for communication that gave us our
first insights into the social structures of the insect world. Von Frisch found
that scout bees would return from a food source and would then dance around the
hive; soon after which, the rest of the forager bees would go hunting and
collecting.
By careful observation over years, Von Frisch found
that the duration, direction and quality of the dance would communicate to the
rest of the hive, details about the quality, location and distance of the food
source. This opened up a whole new field in biology as animals were long
considered to be incapable of communication, coordination, culture and
altruism. Instead, Von Frisch’s findings attacked the very foundations of human
uniqueness and demonstrated these tiny insects to be capable of all this and
much more.
Lindauer, Von Frisch’s student and later colleague,
followed up on this work and found that a young bee swarm would also choose its
future nest site by a decision making process that was based on reports of the
dancing scouts. Seeley was intrigued by the idea of swarm intelligence, by the
ability of a swarm as a whole to arrive at a decision that could potentially
make or break the future of the colony.
His first forays into this question began with
determining what the bees were looking for, when they were scouting the real
estate market. Honeybee colonies reproduce by budding, whereby the queen and
some workers (a few thousand) leave the nest and bivouac on a branch for a few
days depending on their resources. At this point, the swarm needs to locate a
new home for themselves that will ensure their survival. By chasing bee swarms
and by building potential hive sites of many different shapes and sizes as part
of his doctoral work, Tim found that the young swarm was looking for a nesting
site that was high, well insulated and yet roomy. It had to be high enough to
escape predators (at least 15 ft), big enough (atleast 10 gallons) to rear a
new brood and to stock supplies to last through the winter. For a honeybee
swarm, choosing a potential home is a momentous decision that might compromise
their survival itself. The task of hunting is thus delegated to the most
experienced workers who scout nearby locations and report to the swarm.
Lindauer had already shown that the scouts used the
waggle dance to “tell” the rest of the swarm about the new site and that the
decision was somehow arrived at by a collective process. Trying to understand
the dynamics of the swarm was a challenging and almost impossible task at the
time, even for someone like Seeley. In the early decades of the field,
ecologists were limited to the use of a simple tool set of notebooks, pens,
stopwatches, paint sets, wristwatches and little else. Tracking hundreds of
individual scouts, their movements, dances and patterns at the level of
multiple swarms was impossibility. Even someone as creative and ingenious as
Seeley was forced to wait for technology to catch up with his ideas.
Finally, in the last decade of the twentieth century,
as the camera and video recording techniques came of age, Seeley decided that
the time was ripe for delving into those forbidden questions. And the answer
lay in careful and meticulous observation. Seeley and his students would
manually label and mark the thousands of bees in a swarm, but this was only the
easy part of the job. They would then video record the entire house-hunting
process for a period of 2-3 days and at the end of it, they faced the enormous
task of decoding the videos to identify the underlying patterns. The watched
the scouts hunt at different locations, coming back to report and then the
entire decision making process. Each 16 hour film of the bees’ decision making
process would take atleast a month of eye-stinging labour to decipher. What
they found was fascinating indeed.
The experienced scout bees would head out and survey
the available real estate. They would the, come back and advertise the
potential locations and their qualities to their nest mates by performing the
waggle dance. The angle of the bees with respect to the sun would indicate the
direction of the site, while the duration of the dance would indicate the
distance. The quality of the site was indicated by the simple metric of the
number of dance repeats that the bee did. The more a scout bee advertised, its
location, the stronger the lobby it built. These waggle dances would then
recruit additional scouts to the site until a decision was made. Interestingly,
each scout bee would only visit one site on most occasions and would only lobby
or rather “dance” for it. Each time she returned from the site, the number of
dance circuits would progressively decline but each routine would recruit a new
wave of scouts, creating multiple, independent reports.
Seeley and his students began their studies suspecting
a consensus building mechanism to be involved – a democracy of sorts. What they
instead found was that the scouts themselves don’t pay any attention to the
consensus. They decision is made by a quorum, a critical number (20-30) of bees
being simultaneously present in the nest site – representing successful reports
from many independent scouts. Trying to understand the significance of the
number, Seeley and Dr. Kevin Passino, a professor of computer engineering at
the Ohio state university, came together and modeled this decision making
process. They fund that in a simulation, adjusting the quorum to only 15 bees
would result in quick but error prone decisions. Increasing the number to above
20 on the other hand produced slower but only slightly more accurate decisions.
It thus seemed that the bees had found the optimum number to balance the cost
of resource starvation with expected gain in accuracy. But the precise
mechanism underlying this “plebiscite” remained unclear, until of course, Dr.
Seeley and his students heard the head butts.
They found that in deciding which site to choose, the
scouts would also employ another tactic in addition to the
positive-reinforcements achieved by dancing. They also had a stop signal – a
booming head butt. When choosing between two nest sites, the scout bees
committed to each nest site would direct these head butts to the scouts
promoting the other box; thus setting up a cross inhibition between the two
populations of scout bees.
The honeybee stop signal is a vibrational signal
signal that lasts about 150 ms and has a fundamental frequency around 350 Hz.
It is typically delivered, by the sender butting her head against the dancer.
Although the dancer may not show an immediate response to the signal, the
accumulation of such stop signals is seen to increase the probability that the
bee will cease dancing. [Commonly used during foraging, the stop signal is
given when a forager bee is attacked in a food source and this reduces the
recruitment of colony resources to perilous food sources.]
To further understand the implications of such a
cross-inhibitory selection mechanism, Dr. Seeley and his colleagues further
modeled the decision making process. Testing individual models of no or
indiscriminate stop signals suggested the overwhelming possibility of reaching
at an impasse – an impossible deadlock. Such a stable deadlock would result in
the swarm never reaching a decision and thus starving to death. Simulations
showed that the deadlock persists even when the stop signaling is applied with
discretion but below a critical threshold.
When the stop signals are applied above a critical
threshold, the swarm is either able to randomly arrive at a consensus for one
the two equal alternatives or if the difference in the quality of the two sites
is significantly different, then the stop signals ensured a consensus for the
superior alternative. The existence of the stop signal and a high enough quorum
thus ensure that the bees arrive at an optimal decision most of the times.
Once the search committee has made up its mind and the
critical quorum of bees is established at the nest site, the scouts lobbying
for the site begin their final “move”. They begin with an auditory signal,
called as worker piping, that can actually be heard by us. Coincident with this
signal, the other bees in the swarm begin to prepare for flight. Bees fly with
wings moving at some 250 times per second and this requires their wing muscles
to be warm (35 deg C) and metabolically active. To do this, upon receiving the
piping signal, the bees warm up their flight muscles by disengaging their chest
muscles from their wings and vibrating them. The temperature of the whole swarm
rises up rather steeply in less than half an hour.
Worker piping and the pitch of the sound perfectly
match the beat frequency of a flying bee and some early experiments suggest a
critical link between the two. In the final few minutes before the take off,
the workers exhibit another rather strange phenomenon - a so called ritualistic
buzz run. Here, the bees run across the entire swarm with outspread wings and a
buzzing sound, almost invoking everyone into action. As the swarm is warm and ready, the buzz run
begins and finally the swarm lifts-off. It slowly hovers over its temporary
home for a minute or two and then starts proceeding towards the nest at speeds
ranging from 0.5 – 5 miles per hour. They finally stop near the goalm their
future home scented with the Nasonov pheromones from the scouts. In the absence
of these scent markers, the swarm finds its more difficult, not impossible, to
locate the entrance to the hive site. At this point, they are then guided by
these streaker bees, who repeatedly shoot towards the goal and guides the
others. The bees thus finally make it to their new home – and a young hive
begins to grow.
Although elegant and simple, this mechanism needs
further careful study to account for the individuality and subjectivity of the
bees. If in the early phase of the
scouting, a single bee dances more enthusiastically for a bad site or less
enthusiastically for a good site, then the balance of signals could tilt very
rapidly introducing a lot od stochasticity. In fact, as they individually tracked
these bees, Dr. Seeley and his group also noticed many differences between
them. They are all not the same like a cookie cutter and are rather quirky.
Some are really good dancers, some are not; some are really peppy and get going
early in the morning while some others have to be woken up. They have a lot of
personality and with the cumulative mechanism of this process, small vulnerabilities at the start could easily run amock
and alter the fate of the hive.
It would also be interesting to know if the hive
changes its criteria and its choices if they are running out of reserves and
how the choices are made? The rate of information arriving to the hive could
also bias the processing of the input and the cascade to the decision – say a
scout flies closer to the swarm and reports back sooner and builds up a bigger
lobby. The reliability of the responses of the individual bees to multiple
optimal or suboptimal sites is another interesting question that may be rather
revealing. All in all, like all great studies in science, Dr. Seeley’s raises
many more interesting questions, even as he finds the answers to them.
Interestingly, as Dr. Seeley and his students were
uncovering these details of the house-hunting process of the swarm, there
emerged a few parallels that are imperceptible to an otherwise untrained mind.
What Seeley saw was an emerging parallel between the bee swarm and the human
brain. In a swarm, each scout reported on a single find, much like a neuron
responding to a particular stimulus. “Both are cognitive entities shaped by
natural selection to be skilled at acquiring and processing information to make
decisions,” he says. In both these systems, decision-making is a competition
between mutually interacting populations of excitable units – neurons or
individuals – that accumulate noisy evidence for alternatives and, when
population exceeds a threshold, the corresponding choice is made. A common
underlying feature for both bees and neural decision-making is the existence of
cross-inhibition. A population that inhibits others proportional to its own
activation, thus ensuring that only one of the alternatives is chosen.
And so, as I stepped out of that lecture hall at the
end of that hour, my rewired brain could see more to the statement “buzz in the
brain” than was ever obvious to me.
The question is what do you see?
References:
1)
Stop signals
provide cross inhibition in collective decision making by honeybee swarms,
Seeley et al, Science, January 2012
2)
Swarm intelligence
in honeybees: Talk by Dr. Tim Seeley at the University of California, San
Diego.
3)
Decoding the language
of the bee: Nobel lecture, Dec 12, 1973, Karl Von Frisch.
4)
Swarm
Intelligence: How Tom Seeley discovered ways that bee colonies make decisions,
Part II, MEA McNeil
5)
How honeybees
break a decision-making deadlock, Science, 6 January 2012, By Jeremy E Niven
6)
You tube Lecture
by Dr. Tim Seeley
