BoundedRationality_TheAdaptiveToolbox.pdf

254
Thomas D. Seeley
ceased to dance for one site and later danced for another site after being recruited
to it. However, the precise mechanisms of the decision-making process in honeybee
swarms remained a mystery until recently, even though they were the subject
of much discussion (see Wilson 1971; Dawkins 1982; Markl 1985; Griffin
1992).
One thing that has favored the analysis of group decision making in honeybee
swarms is that this process is an external one: it occurs on the surface of the
swarm cluster where it is easily observed. The swarm's surface is where the
scout bees produce the waggle dances that advertise the sites they favor. Thus
one can monitor the building of a consensus among a swarm's scouts by videorecording
their dances. Recently, complete records of the scout bees' dances in
three swarms were made using swarms that consisted entirely of bees labeled for
individual identification (Seeley and Buhrman 1999). With each bee so labeled,
it was possible to follow each bee's history of dancing throughout the decision-making
process of each swarm. In each case, the videorecords yielded not
only a synoptic view of this process at the level of the whole swarm, but also key
insights into the rules of behavior used by individual bees in collectively producing
a decision.
Group-level View
The record of dances performed on one of the swarms is shown in Figure 14.2.
Each panel in this figure shows, for an approximately two-hour period, the number
of bees that danced for each potential nest site and the total number of waggle
runs that the dancing bees performed for each site. We can see that the time when
dances were performed was spread over three days and totaled some 16 hours of
active deliberations. During the first half of the decision-making process (i.e.,
what is shown in the first four panels), the scout bees reported all 11 of the potential
nest sites that they would consider. We can also see that during the first half
of the process the scouts did not advertise any one of the alternative sites more
strongly than the others; during the second half, however, one of the sites gradually
began to be advertised more strongly than all the others. Indeed, during the
last few hours of the decision making, the site that had emerged as the front-runner
(site G) became the object of all the dances performed on the swarm: by the
end there was unanimity among the dancing bees.
Besides the conspicuous transition from diversity to uniformity in the sites
advertised by the dances, we also see a crescendo in dancing at the end of the d&
cision making. If one compares the last panel with the seven prior panels in
terms of number of dancing bees, dances, and number of waggle runs, one sees
that the last panel has by far the highest number of each. Note too that the site that
was ultimately chosen (site G) was not the first site advertised on the swarm; it
was seventh out of 11 sites. Also, we see that many potential nest sites are advertised
only weakly and briefly on the swarm, i.e., by just one to three bees and often
in just one or two panels (sites C, E, F, H, I, J, and K).