The Nest Without A Manager

AThe vocabulary that biology uses for ant colonies was fixed by analogy rather than by observation. Naturalists working in the language of monarchy gave the egg-laying female the title of queen, and the term superorganism, coined in 1911, kept the same shape while changing the figure: the colony was now a body, with castes as organs. Both images assume a centre. Yet a marked worker followed for a day gives no sign of consulting one. She attends to the ants she meets, to the odour of the tunnel and to the load she carries, and to nothing that could be called a report on the colony as a whole. The trouble is not that the old words flatter ants. It is that they smuggle in an answer to the very question the field is trying to ask, namely whether anything inside a nest of thirty thousand animals does the deciding.

BIngrid Halloran, an entomologist at the Tarnbridge Institute for Social Insects, holds that colony behaviour is the arithmetic of encounters and nothing besides. In the desert harvester species she studies, an ant waiting just inside the entrance will leave to forage only if returning foragers brush her antennae often enough; the threshold in her colonies sits near one contact every ten seconds. Because a forager comes back only when she has found something, the rate of contact carries, without anyone measuring it, the news that food is available. Between 2015 and 2021 Halloran's group filmed 74 colonies and logged roughly 1.2 million antennal contacts. A model containing this single rule, and no term for what any ant had done earlier that day, reproduced 81 per cent of the observed hourly foraging rates. Halloran calls the result bookkeeping rather than deflation: a colony that looks governed and one that is governed may be built by identical sums.

CAn account of that kind can be pushed until it breaks, and Halloran has done the pushing. If foraging is only the settled state of encounters already occurring, colonies ought to fail wherever the profitable move means abandoning a trail that is currently yielding, and to fail again wherever the correct response depends on an event that is already over. In staged arenas in which a rich feeder appeared only if a poorer one had been left untouched for forty minutes, her colonies succeeded on 23 per cent of trials, against 91 per cent on conventional shortest-trail tests of comparable length. Success in the staged arenas was confined to trials in which both feeders lay within the trail network the colony had already laid down. The asymmetry is the instructive part, since a honeybee colony finds the waiting problem straightforward and the trail problem hard. Halloran presents the percentages as a fence around one explanation rather than a verdict on the insect.

DMarcus Ilori, who runs the Rowan Valley Field Station, replies that a colony tested in a bare plastic arena, at constant temperature and with the nest reduced to a plaster chamber, is working in conditions its species has never met, and that a failure recorded there describes the apparatus as much as it describes the animals. His own colonies are kept for thirty days in soil nests with a humidity gradient before any trial begins, a practice now unremarkable in the field but treated as fussy when he adopted it in 2017. Under that regime, 58 per cent of his colonies began withdrawing foragers shortly before a hot, dry pulse that had been delivered every twenty-four hours was next due, and continued to withdraw them for three further days after the pulses had ceased. Ilori concedes two weaknesses. His eleven colonies came from a single population, and the colony-level state he calls memory was inferred from the timing of departures alone, since nothing yet permits it to be read directly.

EThe dispute has left the laboratory. Engineers borrowed the encounter rule as a design principle: a 2019 study applied it to a parcel delivery network and returned a schedule 9 per cent shorter in total distance than the operator's own, though it separated into isolated regions whenever a single depot was withdrawn. Transport authorities in four countries have commissioned comparable exercises, and one was rebuked for calling the output a plan drawn up by an insect society. Halloran notes, inconveniently for her own side, that nothing in her model settles whether individual ants differ from one another in any lasting way, because a local rule and stable individual differences are not competing options. Ilori answers that in practice the two cannot be held apart, since committees read a report of skill as evidence that the animal is understood, and a report of failure as evidence that there is nothing there. Both dislike the convention of publishing one photograph of a tidy trail instead of the ragged spread.

FWhat is emerging is less a winner than a division of labour between the two accounts. Where a problem can be solved by encounters finding their own level, the colony appears to solve it with nothing carried from one moment to the next, which would explain both the speed of the response and the readiness with which a colony abandons worked ground. Where something must survive an interval, attention has moved outside the animals altogether, to the pheromone films left on the trail and to the architecture of the nest itself, records held in the substrate rather than in any head. Whether such records do any work in open country, where rain and rival colonies erase them within hours, no field method now available can determine. What both researchers refuse is the assumption behind the old metaphors: that a society without a manager must be either a crude imitation of a government or nothing at all.