Decoding the Language of Bees

AThe swarm that hangs from a branch in early summer is still described, even in otherwise careful books, as a court in exile: a queen who has left the hive with her retinue and who will presently choose a new address for it. The picture is charming and almost entirely wrong. A queen in a swarm neither inspects property nor issues instructions, and the decision is taken instead by perhaps three hundred elderly foragers, fewer than five in every hundred bees present, who leave the cluster to search the surrounding countryside and come back to advertise whatever they have found. They do so with the same figure-of-eight run that a forager uses to report a patch of flowers, its angle giving the bearing and its duration the distance. Within a few hours the scouts will have located a dozen or more hollow trees and wall cavities inside a radius of roughly three kilometres. The problem that then arises is not how a site is described but how eight or ten thousand animals settle on one of them, given that scarcely any bee ever enters more than a single cavity. The queen herself will simply be carried to whichever address the others have settled upon.

BThe explanation that held the field for four decades is elegantly simple and requires no comparison of options by anybody at all. A scout that has found a promising cavity dances for it with a vigour reflecting its quality, and on each return to the cluster she dances a little less, so that a mediocre site is abandoned by its own advocate within a few circuits while a good one keeps drawing new recruits. Support therefore accumulates where the evidence is strongest, without any bee weighing one address against another. Departure is triggered only by a headcount at the cavity itself: once the number of scouts gathered there passes roughly twenty, they fly back and the whole cluster lifts off. Between 1996 and 2001 marked colonies at Erlenbach chose the best available cavity in 32 of 36 trials, a record that seemed to settle the matter. The model is attractive precisely because it asks for nothing that a very small brain could not manage, since nothing in the account requires one bee to remember a second address or to trust the report of another.

CPerpetua Ojeda, of the Ravensmoor Institute of Social Insects, maintains that something further is at work whenever the choice is genuinely difficult. Filming clusters from two angles at once, her group catalogued a brief vibration that one scout delivers by butting her head against another, a message previously thought to mean nothing more than that foraging should cease. Of 1,900 such pulses recorded between 2014 and 2016, 84 per cent were aimed at bees advertising a cavity other than the sender own, and the recipients cut short their dances. Ojeda argues that the cluster is therefore doing more than piling up support, since each faction is actively suppressing its rivals, much as competing populations of neurons inhibit one another before an animal commits itself to a movement. She is precise about when this matters. Where one cavity is clearly superior the pulses are rare and the outcome is exactly what decay alone would predict; the mechanism becomes visible only when two sites are close in quality, a situation her team had to engineer with artificial boxes.

DJulius Kranz, who runs the field station at Erlenbach, is sceptical that any additional mechanism need be invoked. He points out that a stop signal is not obviously a vote against a rival, because a bee jostled inside a dense cluster receives such butts from every direction, and he contends that the frequencies reported are close to what chance would deliver in a crowd of that density. His own simulations, two hundred runs of a model containing nothing but dance decay and a threshold at the cavity, reproduced the recorded decision times to within a few minutes. The objection is a serious one, and yet it has a hole to which Kranz himself has drawn attention: when the two sites in his model are given equal quality, the swarm either splits or hangs undecided for days, which is not what real colonies do. His parameters, moreover, were fitted to a single swarm watched through one season, and he concedes that a simulation reproducing an outcome does not thereby establish the process that produced it. Nothing in the recordings, on his reading, obliges anyone to adopt a richer interpretation.

EThe dispute has narrowed to the question of what happens in the awkward cases. Ojeda does not claim that inhibition drives ordinary decisions, and Kranz does not claim that his model handles ties, so that between those two admissions there is a good deal more agreement than the exchange of published papers suggests. The reasonable reading is that decay does the work whenever the countryside offers one obvious cavity, which in most summers it does, and that the vibration exists to break deadlocks that would otherwise divide a colony in two. That is a description of an ordinary summer rather than a general law. What neither party disputes is the stranger point buried in the earliest observations. No bee compares anything, because no bee has anything to compare, and the assessment that looks so deliberate from outside is distributed across three hundred insects, not one of which holds more than a fragment of the information on which the swarm finally acts.