Eusociality: The Unit of Selection Moves Up Again

I The Life of a Worker Bee

A worker honeybee chews through the cap of her hexagonal wax cell and crawls out on the twenty-first day after her egg was laid. Her mother is the queen, the colony's only one, capable of laying a thousand or two thousand eggs a day in the busy season. Her father was a drone. He died long before she was born.

Her life is fully booked, with changes of job largely determined by age. For the first few days she cleans cells, licking each newly vacated compartment ready for the queen to lay in again.

Next she becomes a nurse. Glands in her head produce a milky-white secretion that she feeds to the larvae: royal jelly. A little older, she begins to secrete thin flakes of wax from glands on the underside of her abdomen, and goes to work repairing the comb and building cells. After that she stands guard at the entrance, sniffing every bee that comes in and turning away those that do not belong.

A guard bee carries a barbed sting. If she stings a human or another mammal, the barbs lodge in the skin and will not come free. The sting and its attached venom sac are torn from her body, and she soon dies. Left in the skin, the sting keeps pumping venom. That is why, if a honeybee stings you, you should scrape it off as quickly as possible.

Only after about three weeks does she make her first flight to gather nectar. Foraging is the most exhausting work. A summer worker usually lives just five or six weeks, spending almost all of the last two or three in flight outside the hive. Her wings wear down to ragged edges. In the end she dies out in the open.

A common estimate puts the honey a worker produces in her entire lifetime at roughly one-twelfth of a teaspoon. Nor is this succession of jobs set in stone. If a beekeeper removes the foragers from a colony, young nurse bees will start gathering nectar ahead of schedule, within a few days.

Back in the hive, a forager performs a figure-eight dance on the comb. The direction of the straight run through the middle indicates the direction of the nectar source relative to the sun; the length of time she spends waggling along that run indicates its distance. The Austrian zoologist von Frisch deciphered this dance language. In 1973, that work earned him a share of the Nobel Prize in Physiology or Medicine with two other ethologists.

A strong colony may have twenty or thirty thousand to fifty or sixty thousand workers in summer. All are female, yet the vast majority never reproduce in their lives. All their labor goes into raising the queen's offspring: their own younger brothers and sisters.

There are only a few hundred drones in the hive. Larger than the workers, they have enormous eyes that meet on top of the head. They have no sting, collect no nectar, and do no housework. Their sole use is to fly out and mate with new queens from other colonies. When flowers grow scarce in autumn, the workers drag the surviving drones out through the entrance and refuse to let them back in.

Two kinds of honeybee are kept in China. One is the native Chinese honeybee, small and hardy in the cold, able to forage on scattered flowers in the mountains. The other is the Italian honeybee, introduced from the early twentieth century onward. Larger and more productive, it is now the mainstay of commercial apiaries. The social organization of the two is much the same.

Workers and queens do not come from two fixed genetic castes. The fertilized eggs laid by a queen need not have identical genomes, but they share the potential to develop along different female pathways. What a larva is fed, how much, and when helps steer gene expression and development. Sustained, appropriate queen feeding leads one way; the worker feeding regime leads another.

In 2008, researchers in Australia reduced the activity of an enzyme that adds chemical marks to DNA in bee larvae. Many larvae raised under worker conditions developed queen-like traits. Early nutrition and gene regulation can send reproductive development down different paths. That does not mean food rewrites every gene, or that all workers are incapable of laying eggs: some can still lay unfertilized eggs that develop into males.

This gave Darwin a considerable headache. In On the Origin of Species in 1859, he called sterile workers a special difficulty, one “which at first appeared to me insuperable, and actually fatal to my whole theory.” Natural selection works through individuals that leave more offspring. How can the traits of an individual that never breeds be passed on?

Darwin supplied an answer of his own, by way of cattle breeding. A beast with excellent meat is slaughtered and eaten; it can leave no more offspring itself. But the breeder can go back to animals from the same family for the next mating, and the quality of the meat will still be inherited. Selection can act on families as well as individuals. That analogy already contained the seeds of more than a century of argument.

Biologists call a society like that of the honeybee eusocial. The American entomologist Batra introduced the term in 1966 while studying sweat bees in India. It was later given three criteria: the same nest contains members that specialize in reproduction and members that reproduce little or not at all; adults of at least two generations live together in the nest; and members jointly care for young that are not their own. These three criteria were chiefly consolidated by the American bee specialist Michener in 1969 and by Wilson in his 1971 book The Insect Societies.

What this series calls a construct comes down, in the end, to drawing a line: what falls inside counts; what falls outside does not. The line may be drawn by people, as a set of criteria is, or by nature, as a membrane is. These three criteria for eusociality are a human line drawn across a broad expanse of social life.

Who falls inside? The list is longer than many people imagine. All ants, all termites, honeybees, stingless bees and bumblebees, some sweat bees, and the paper wasps and yellowjackets among the vespid wasps.

There are unexpected members, too. In 1977, Japanese researchers discovered that some aphids produce specialized soldiers that fight and never go on to mature and reproduce. These soldiers have thick forelegs and pointed horns on their heads. When a ladybird larva or a hoverfly maggot comes crawling toward them, they spring at it, stabbing with their horns and gripping with their legs, often dying along with the intruder. On acacias in arid Australia, several kinds of thrips live together inside galls. They have soldiers, too.

The sea has its examples. In 1996, the American biologist Duffy found snapping shrimp living in the channels of sponges on the coral reefs of Belize in Central America. Several hundred occupied a single home, but only one female laid eggs; the rest guarded the entrance and fought. In Australia, an ambrosia beetle that tunnels into living eucalyptus trunks likewise has workers that never lay an egg.

Most surprising of all are two mammals. Naked mole-rats from the dry regions of East Africa are about ten centimeters long, almost hairless, with wrinkled skin. They excavate underground tunnel systems that can run for several kilometers. A colony averages seventy or eighty animals and may approach three hundred, but only one female breeds. The Damaraland mole-rat of southern Africa leads a similar life.

The naked mole-rat's story has a prequel. In the 1970s, Alexander, an entomologist at the University of Michigan in the United States, speculated in a lecture that if eusociality existed among vertebrates, the animal would probably be a rodent that lived underground, ate large roots and tubers, and inhabited dry tropical clay soils.

Someone later told him that such an animal really did exist: the naked mole-rat. In 1981, Jarvis of the University of Cape Town in South Africa reported in Science that only one female bred in a naked mole-rat colony. Prediction and discovery had found each other.

Naked mole-rats are also exceptionally long-lived. An ordinary mouse of the same size generally lives only two or three years; a naked mole-rat can survive for more than thirty years in the laboratory, with a mortality rate that scarcely rises even in old age. Researchers usually call the breeding female the “queen.”

Their oddities do not end there. Naked mole-rat skin is almost insensitive to the pain of acid, and their body temperature rises and falls with the temperature of the burrow, much like that of a cold-blooded animal. In 2017, researchers put them in a chamber containing no oxygen at all. They lost consciousness for eighteen minutes, then revived when returned to air. Their hearts and brains survived by temporarily switching to breaking down fructose for energy, something rarely seen in other mammals.

This essay asks how the creatures on that list came to be, how many times they arose and why, what fossils they left behind, and what, exactly, biologists have spent more than half a century arguing about.

II Fifteen Times or Five

Start with termites. They look like ants, their Chinese name includes the word for ant, and they live much as ants do. Yet they are not close relatives at all. Ants belong to the order Hymenoptera, along with bees and wasps. Termites are related to cockroaches.

In 2007, three researchers at Britain's Natural History Museum used DNA to reconstruct the family tree of cockroaches and termites. They found termites nested inside the cockroach tree, closest to a group of wood-eating cockroaches. They called their paper “Death of an order.” Termites no longer qualified as an order of their own. They were a highly specialized group of cockroaches that had taken up social life.

Those closest relatives are Cryptocercus wood roaches. They live in decaying wood in the mountain forests of North America, the Russian Far East, and parts of China. A male and female stay together for life, sharing tunnels in the rotten timber with their young, which take several years to leave home. More importantly, these roaches depend on protozoans in their guts to digest wood. Newly hatched young lack those microbes and must acquire them by licking microbe-rich fluid from their parents' anuses.

Termites pass on their gut microbes in the same way.

In a large, mature termite colony, the queen's abdomen swells as thick as a finger and several centimeters long, a pale, swollen length of body. She can scarcely move; workers feed and clean her. She is estimated to lay more than ten thousand eggs a day and may live for well over a decade, with the king who founded the nest beside her throughout. One colony, one family; long-lived parents; children obliged to stay home to be fed. Put those conditions together and you are already close to a termite colony.

The entire termite lineage counts as a single origin of eusociality. All living termites are eusocial, and their common ancestor was already so. The same goes for ants: more than fourteen thousand living species, all social, count as one origin.

One origin, followed by an astonishing range of variations. The army ants of South America and Africa have no permanent nest. Hundreds of thousands, sometimes more than a million workers march together. At night they hook their bodies to one another to build a living, temporary bivouac around the queen and larvae.

The weaver ants of southern China make their nests from leaves in trees. A row of workers uses its mandibles to pull two leaf edges together, while others carry larvae back and forth in their jaws. Silk spun by the larvae stitches the edges closed. Ji Han's Plants of the Southern Regions, from the Jin dynasty, records people near Guangzhou selling these ants, nests and all, in bags, to be hung in citrus trees for pest control. The book is generally dated to AD 304, making this one of the world's earliest records of biological control.

Bees and wasps are far more complicated. Within the bee family Apidae, sociality may have originated once or twice in the branch containing honeybees, stingless bees, bumblebees, and orchid bees. Eusociality evolved several times in the sweat bee family, Halictidae, and several lineages later returned to solitary life. In the wasp family Vespidae, the sociality of paper wasps and yellowjackets has a separate origin.

Paper wasps build with paper. They scrape fibers from dead wood and bark with their mandibles, mix them with saliva, chew them to a pulp, and spread the pulp into layer upon layer of thin gray or brown paper. In 1719, watching wasps build, the French naturalist Réaumur suggested to his colleagues that people might try making paper from wood. European paper was still made from rags at the time.

Sweat bees are tiny, and often land on people's arms in summer to lick their sweat, hence the English name. They dig long, narrow tunnels in the ground and build brood chambers along both sides. This is probably the source of their Chinese name, “tunnel bees.”

Count all the way down the list, then: how many independent origins of eusociality have there been among animals? Earlier reviews often gave figures somewhere between ten and fifteen or so.

In 2021, da Silva of the University of Adelaide in Australia used a recent hymenopteran family tree, assigned each species' social life to one of seven grades, and reconstructed ancestral states. The result: eusociality in the broad sense had arisen independently at least fifteen times among bees and wasps, not even counting the origin in ants. But the kind in which caste differentiation is irreversible and individuals can no longer survive apart from the colony had evolved only five times across Hymenoptera as a whole.

The same study, the same tree, and both fifteen and five are right. They count different things. The first asks whether reproductive division of labor and cooperative brood care have appeared. The second asks whether an irreversibly sterile caste has appeared.

In the field, those two things look very different. The “workers” in a sweat bee nest have intact ovaries. When the queen dies, some of them start laying eggs. You can see it happen within days.

A red imported fire ant worker has no ovaries from the start. If the queen dies, she still has no way to become a mother. The first is not breeding for now. The second can never breed at all.

The return to solitary life is clearest among sweat bees. In warm lowlands, some species can raise two broods a year, with the daughters of the first brood staying on as workers. Higher up, the summer is too short for more than one brood, and the same species reverts to solitary life. At this end of the range, the door between eusocial and solitary life remains open. You can go in, and you can come out.

Those comings and goings can throw off the count. A lineage that is solitary today may have had a social ancestor and reverted later. If you code it as “always solitary,” a neighboring social lineage will be counted as one more independent origin. We met the same problem in Essay VIII when counting the origins of multicellularity.

III The Arithmetic of Sisters

The first person to work seriously through these accounts was Britain's Hamilton. In the early 1960s, he was a graduate student in London. Few thought much of his topic, and his supervisor paid him little attention. He often sat alone doing his calculations in the waiting room at Waterloo Station. In 1964, he published two consecutive papers in the Journal of Theoretical Biology. For the first few years, hardly anyone understood them.

Hamilton went on to become one of the most respected figures in evolutionary biology. In 2000, he traveled to the rainforests of the Congo to collect chimpanzee feces, hoping to test a hypothesis about the origin of HIV. He fell ill after returning to Britain. Initial reports spoke of malaria; an autopsy later established massive gastrointestinal bleeding as the cause of death. He was sixty-three.

An apocryphal joke captures his central idea. The British geneticist Haldane is said to have done a calculation in a pub and announced that he would lay down his life for two brothers or eight cousins. On average, a brother shares half your genes, so two brothers add up to one of you. A first cousin shares only one-eighth; it takes eight to make the same total.

Hamilton turned the joke into a rule, now expressed by a very short formula. If the benefit an action brings to someone else, multiplied by your degree of relatedness to that individual, exceeds the cost to you, a tendency to perform that action can spread through the population. It became known as Hamilton's rule. The theory behind it is kin selection.

“Altruism” here is only a technical term: reducing your own reproduction while increasing someone else's. It says nothing about intention. A worker bee does not know that the queen is her mother, and she cannot calculate a coefficient of relatedness.

Hamilton also found a seductive number. Honeybees, ants, and wasps have an unusual way of determining sex. Fertilized eggs develop into females; unfertilized eggs develop directly into males. A male has only one set of chromosomes, and every sperm he produces is identical.

Here is the consequence. Daughters with the same father receive an identical half of their genes from him. Of the half inherited from their mother, they share, on average, half. Add the two contributions and sisters have a relatedness of three-quarters. A female bee's relatedness to her own daughter, by contrast, is only one-half. Sisters are closer kin than daughters.

This seemed to explain everything. Helping your mother raise more younger sisters paid better than producing daughters of your own. No wonder sterile workers kept appearing among these insects. The explanation held sway in textbooks for decades.

Then its supports were knocked away one by one. First, a queen often mates with more than one drone. A honeybee queen leaves the hive on her mating flight and mates in the air with a dozen or more males. After mating, a drone's genitalia break off inside the queen, and he soon dies.

The queen stores their sperm, several million in all, in a sperm-storage organ, the spermatheca. For the next two or three years, sometimes longer, she releases a little whenever she lays a fertilized egg. The workers in a colony therefore belong to a dozen or more paternal families. Sisters who share a mother but have different fathers are related by only one-quarter.

Second, workers rear younger brothers as well as younger sisters. A female bee is related to her brother by only one-quarter. Put brothers and sisters into the same calculation and the apparent advantage may disappear.

Third, both male and female termites have two sets of chromosomes. They lack this special sex-determining system, yet are among the most intensely social insects. Fourth, well over a hundred thousand species of Hymenoptera have been described, and the vast majority are solitary. If three-quarters were really the decisive number, most should have become social long ago.

The study that truly shifted the question's center of gravity appeared in 2008. Hughes at the University of Leeds in Britain, with colleagues at the University of Sydney in Australia and the University of Sheffield in Britain, gathered mating records for more than 260 species of bees, ants, and wasps and placed them on a family tree. At each origin of eusociality, they asked, how many males had the ancestral mother mated with?

Across eight independent origins, every ancestral mother had mated with only one male. Mating with multiple males arose later in several lineages, after eusocial colonies had formed. What had been treated as an obstacle appeared after the society took shape, not before.

Boomsma of the University of Copenhagen in Denmark called this the monogamy hypothesis. His analogy was a couple that have children only with each other, all their lives. Their children are full siblings, related by one-half, just as closely as each would be related to its own offspring. Stay home and help your parents raise your younger siblings, or leave and breed for yourself: on relatedness alone, the two choices are a draw. If helping is even a little more efficient, staying may pay better.

The analogy works for termites, too. A termite nest is founded by a king and queen that remain together for life. All the offspring in the nest are full siblings. High relatedness does not require an unusual sex-determining system. Monogamy is enough.

Ant queens take the idea of making one mating occasion last a lifetime to its limit. A queen mates only on the day of her nuptial flight, stores the sperm in a tiny sac, and uses it one sperm at a time for years or decades afterward. In the laboratory, one black garden ant queen lived for almost twenty-nine years, still using the sperm she had stored when young.

Monogamy is not the whole answer either. Hymenoptera abound in monogamous, solitary species that have never evolved workers. How much good will helping do? What chance is there of succeeding alone? Where is the nest built, and do the larvae need continuous feeding? These ecological conditions decide which way the balance tips. Monogamy opens the door. Whether you walk through depends on the landscape beyond it.

IV Half a Century of Arguing Over the Accounts

There is another way to calculate all this, beyond Hamilton's rule, and it came from an even more unusual man. The American chemist Price had worked on the atomic bomb project when young, then as a journalist and for a computer company. He arrived in London in 1967 and taught himself genetics. Reading Hamilton's papers, he derived what is now called the Price equation. Hamilton was astonished when he saw it, and the two became friends and collaborators.

In his later years, Price grew deeply religious. He gave away his money and belongings to homeless men on the streets of London, and died there in 1975, by then destitute himself. Hamilton attended his funeral.

Price made another contribution as well. With Maynard Smith, he wrote a paper on why fighting animals so often stop short of all-out violence. Published in Nature in 1973, it brought game theory into the study of animal behavior. We will meet Maynard Smith again in Section VIII.

The Price equation lets us divide the change in a trait across a population into two parts. Take bee colonies. One part concerns differences between nests: do nests with more workers, working more effectively, produce more new nests? The other concerns differences within a nest: do workers that sneak in more egg-laying leave more offspring of their own?

The first part is called between-group selection, the second within-group selection. For a system of sterile workers to spread, the advantage between nests must outweigh the pressure to reproduce on the sly within them.

The same total account can be broken down another way: how much do I gain myself, and how much do I give my relatives, weighted by relatedness? That is Hamilton's kin selection. A company's profits can be broken down by department or by product. The total is the same; the ledgers look different. As early as 1992, researchers showed that in many models the two decompositions can be translated mathematically into one another.

The fiercest round of this accounting dispute erupted in 2010. Nowak, Tarnita, and Wilson at Harvard published a paper in Nature arguing that the mathematical conditions required by kin selection theory were much narrower than generally claimed. To study eusociality, they said, ordinary population genetics combined with a model of the colony life cycle would suffice. Inclusive fitness was unnecessary.

Wilson was the most unexpected participant on that side. The twentieth century's most famous ant specialist, he had published Sociobiology in 1975, the very book that carried Hamilton's kin selection across the animal kingdom. Thirty-five years later, he stood against it. He died in the United States at the end of 2021, aged ninety-two.

An accident in childhood helped set Wilson on the study of ants. At seven, fishing by the coast in the American South, he was struck in the right eye by a fish's dorsal-fin spine. He later lost the sight in that eye.

With the remaining eye, he could see clearly up close but poorly at a distance. He recalled that because he could not watch birds and mammals, he turned to the small things within reach, and so spent his life watching ants. He won the Pulitzer Prize twice, once for The Ants, written with Germany's Hölldobler.

In March 2011, Nature published a response. Its authors were listed alphabetically, beginning with Abbot of Vanderbilt University in the United States. It bore 137 researchers' names, a joint response of a scale rarely seen in biology.

They argued that the other side had misunderstood the mathematical reach of inclusive fitness and underestimated its predictions. Sex ratios, workers policing one another, the influence of mating frequency on cooperation: in all these cases kin theory had made predictions before the data arrived to test them. The same issue carried several other independent responses. Both sides went on publishing, and neither has conceded defeat.

At the end of all the shouting, three statements can be separated out. They do not carry the same weight. First, any change produced by selection can be divided up using either form of accounting. On that, the two sides largely agree. Second, any model can be expressed after the fact as some generalized form of Hamilton's rule, provided its coefficients may be defined to suit the circumstances. That, too, is broadly true.

The third statement is the narrowest: can you measure relatedness, benefit, and cost in advance, put those three numbers into the simple inequality, and predict the direction of evolution? When a worker's contribution depends on how many workers the nest already contains, for example when a minimum number must be assembled before a nest can be built at all, benefits and costs cease to be fixed numbers. The third claim becomes hard to sustain. This is where most of the substantive disagreement lies.

The colony life-cycle model used by Nowak and his colleagues follows a nest from its beginning with a single mother through growth, the production of new nests, and eventual death. Its calculation gave this result: a mutation that makes daughters stay as workers spreads readily only if it brings substantial benefits even while the nest is still tiny, with only two or three workers. The threshold lies with those first few workers, not in the decimal places of relatedness.

A bumblebee's life illustrates that threshold neatly. In spring, a queen starts a nest alone in an underground cavity. She goes out for pollen herself and incubates the eggs herself. During those weeks, whenever she leaves, the nest is unattended. A downpour or a mouse can undo everything. Once the first workers emerge, the queen can stay inside and concentrate on laying eggs. The nest's chance of survival rises sharply.

That conclusion does not clash with the monogamy hypothesis. Monogamy concerns the genetic conditions; the life-cycle model concerns the ecological returns. Both may be needed.

V How Far Down Has Conflict Been Pushed

Workers may live their entire lives without breeding, but that does not mean they lack the capacity. Most honeybee workers retain ovaries. Those ovaries are suppressed, first of all by the queen's scent.

In the 1950s, Butler in Britain discovered that the queen's mandibular glands secrete a substance that spreads through the colony as workers lick the queen and feed one another. It inhibits the development of their ovaries. He called it “queen substance.” Its principal component was later identified as a fatty acid.

When a queen grows old or dies, the colony hurriedly builds several extra-large queen cells and switches a few ordinary female larvae to a diet of royal jelly, raising new queens. The first to emerge makes a thin, high-pitched squeak that researchers call queen piping. She then seeks out the other queen cells and stings the rivals still inside them to death. In the end, only one queen remains.

If the queen dies and the colony cannot rear another in time, the workers' ovaries develop and they begin laying eggs. All their offspring are drones. With no new workers, the colony declines within a few weeks.

Even with a queen present, some workers slip through. A worker has never mated, so her eggs are unfertilized and, under that sex-determining system, can develop only into drones. In 1989, Britain's Ratnieks and America's Visscher discovered a second line of defense: other workers patrol the cells, identify worker-laid eggs, and eat them.

This mutual policing is extraordinarily efficient. More than ninety percent of worker-laid eggs introduced in experiments were quickly removed. The result is that in a queenright colony, only about one drone in a thousand is a worker's son. The main restraint on worker reproduction is other workers, not the queen.

Why should a worker eat her sisters' eggs? Kin selection offers a calculable answer. When the queen has mated with a dozen or more drones, a worker is less closely related to the son of a maternal half-sister than to the queen's son, her own younger brother. Replacing a sister's eggs with her mother's therefore pays on the kinship ledger.

The reasoning also yields a prediction. In species where the queen mates just once, workers are more closely related to one another, so mutual policing should be weaker. That prediction was made before the full data were available, and later comparative studies supported it.

Sex ratios provide another measurable conflict. In 1976, the Americans Trivers and Hare calculated that under this sex-determining system, a queen is equally related to sons and daughters and favors equal investment in the two sexes. Workers, however, are three times as closely related to younger sisters as to brothers, and favor a three-to-one bias toward females. They looked at ants and did indeed find greater investment in females. Mother and daughters share a nest, but not a preferred ratio.

The defenses can be breached. Some ants specialize in stealing pupae from other nests to raise as their own workers. They are called slave-making ants. In On the Origin of Species, Darwin described the blood-red wood ants he had watched himself in southern England. They raided a black ant species for pupae, which grew up to work in somebody else's nest.

In bumblebee nests, conflict can be visible to the naked eye. By late summer, workers start laying eggs of their own, competing with the queen to produce males. Queen and workers shove one another and steal each other's eggs to eat. Sometimes the workers mob and kill the queen. Researchers call this the competition phase.

One group, the cuckoo bumblebees, has no workers of its own. A female enters another bumblebee's nest, kills or subdues the resident queen, and makes the colony's workers raise her offspring.

Honeybees offer an even more thoroughgoing example. Workers of the Cape honeybee in South Africa can lay female eggs without mating, producing offspring that are almost clones of themselves. In 1990, beekeepers moved Cape colonies into northern South Africa. From then on, a clonal lineage descended from a Cape worker invaded the local African honeybee colonies one after another, laying eggs and bringing colonies down, like a parasite with wings.

Conflict lies hidden deeper still. Some red imported fire ant nests have a single queen; others have tens or hundreds. A long stretch of chromosome determines which route a colony takes.

In 2013, researchers found that this stretch functions like a “social chromosome.” Workers carrying a particular version kill queens that lack it. Native to South America, the red imported fire ant was found in Guangdong in 2004 and has since spread through many of China's southern provinces. Those soft little earthen mounds along field banks are its nests.

So even when workers have no ovaries at all, interests can still diverge: between queens in the same nest, between a mother and her workers, and between genes inherited from father and mother within a single worker. Conflict has been pushed very low. It has not been pushed to zero. The conflict inside multicellular bodies discussed in Essay VIII had the same shape.

VI What the Fossils Say

When did this arrangement first take hold? The fossil record gives some quite specific dates.

Amber begins as tree resin. A wounded trunk oozes sticky resin; small insects crawling across it become trapped, then covered by fresh layers. Buried in the ground, the resin hardens into amber over millions of years. It can preserve even an insect's bristles and compound eyes.

Some of the oldest ant fossils preserved in amber date to about a hundred million years ago, mainly from northern Myanmar and France. The Myanmar amber comes from the Hukawng Valley in Kachin State. About ninety-nine million years old, it is the resin of a tropical forest from the middle of the Cretaceous.

Among its inhabitants are bizarre “hell ants,” with mandibles like a pair of upward-curving sickles and a horn on the head. In 2020, researchers described a piece of amber about ninety-nine million years old in which a hell ant had caught a nymph between its mandibles and head horn. Resin had sealed the very instant of capture.

In 2016, Barden, newly graduated with a doctorate at the American Museum of Natural History, and his adviser there, Grimaldi, distinguished workers and a winged queen of the same ant species in amber of the same age. Several workers were also crowded into a single piece. This shows that by about a hundred million years ago, ants already had morphologically distinct castes and were very likely living in colonies.

That same year, researchers described termite soldiers, workers, and winged reproductives in Myanmar amber, likewise about ninety-nine million years old. Older termite fossils come from Early Cretaceous strata around 130 million years old in Spain, Britain, and elsewhere. There are also records from around the Jurassic–Cretaceous boundary in Transbaikalia, Russia.

Bees have a somewhat shorter history. A tiny insect described in 2006 as the earliest fossil bee comes from Myanmar amber about a hundred million years old. Less than three millimeters long, it still bears some wasp-like features. The oldest fossil stingless bee comes from amber in New Jersey in the United States, dating to more than sixty million years ago.

One intriguing number is the proportion of ants in amber. In Cretaceous amber, ants make up less than one percent of the insects. Tens of millions of years later, their share rises to about five percent in Baltic amber, then to more than twenty percent in the younger amber of the Dominican Republic.

Ants existed a hundred million years ago, but they became truly ubiquitous only after the dinosaurs died out. Why the long wait? One common explanation involves flowering plants. As angiosperm forests expanded in the Late Cretaceous, the leaf litter thickened and small insects grew more abundant, giving ground-foraging ants more prey and more places to hide. The explanation remains under discussion.

The amber along the Baltic coast is about forty million years old. Even today, pieces washed up by the waves can be found on the beaches, often with ants sealed inside.

China has a window of its own. Coal mines at Fushun in Liaoning produce amber about fifty million years old, enclosing ants, bees, and all manner of small insects. In recent years, Chinese researchers have also described large numbers of insects from Myanmar amber, including early ants and termites.

There is a limit to what amber can prove. A piece containing both a wingless worker and a winged queen establishes two distinct bodily forms. It does not establish that the worker's ovaries were empty. Reproductive division of labor can only be inferred backward from living relatives.

Nests can fossilize, too, as trace fossils. In the 1990s, structures from Petrified Forest National Park in Arizona, in the United States, were reported as Triassic “termite nests,” about 220 million years old. If genuine, they would place termite social life near the beginning of the age of dinosaurs. Later researchers reexamined them and concluded that termites need not have made them. Some may be traces of enormous plant root systems. Most researchers no longer accept the termite interpretation.

The traces left by living termites can be staggering in scale. In 2018, researchers used satellite images to count roughly two hundred million termite mounds in the dry woodland of northeastern Brazil. Each stands two or three meters high, ten or twenty meters from its neighbors. Together they cover about 230,000 square kilometers, an area comparable to Great Britain. Dating soil from the mounds showed that some had stood for nearly four thousand years.

These mounds are not nests. They are spoil heaps left by a termite species as it excavates tunnels. The Brazilian mounds contain an estimated ten cubic kilometers of earth, equivalent to four thousand Great Pyramids of Giza.

The Chinese have a long acquaintance with termites as well. The Han Feizi says that “a thousand-zhang embankment collapses through the burrows of mole crickets and ants.” This later became the proverb “a thousand-li embankment collapses through an ant hole.” Many of the “ant holes” in southern river embankments are actually termite nests. Even today, people along the Yangtze and Pearl Rivers are employed specifically to find and control termites in flood defenses.

Some large African termites build earthen towers several meters tall above their nests. A maze of passages runs through each tower: warm air rises by day and cool air sinks at night, flushing carbon dioxide out of the nest like a breathing lung. An office building completed in 1996 in Harare, Zimbabwe's capital, borrowed from termite-tower ventilation in its design and stays cool without conventional air conditioning.

Termites also grow mushrooms. A group of large termites in Africa and Asia builds spongy fungus gardens inside the nest from chewed-up plant matter, cultivating a specialized fungus.

When that fungus sends mushrooms up through the ground above the termite nest, they are the jizong mushrooms people in Yunnan eagerly seek out to eat during the rainy season. Beneath every jizong mushroom is a termite colony.

Ants keep livestock, too. Many guard herds of aphids, tapping their backs gently with their antennae. The aphid releases a drop of sugary honeydew from its rear; the ant drinks it, then returns the favor by chasing away ladybirds. You can often watch this on the rose stems in a garden.

South American leafcutter ants also cultivate fungi. They cut leaves into small pieces and carry them home, not to eat the leaves but to feed a fungus, which they eat instead. Molecular-clock estimates put ant fungus farming at some fifty or sixty million years old. A large leafcutter nest can house several million workers and hundreds of underground fungus chambers.

A molecular clock works by comparing differences in the DNA of two species, then using an approximately steady rate of change to work backward to when they diverged. The origins it dates are often earlier than the fossils.

For ants, molecular clocks commonly put the origin at around 140 million years ago, tens of millions of years before the oldest ant-bearing amber. Essay XIX will examine where that gap comes from.

Finally, consider their scale today. In 2022, researchers in Germany and Hong Kong, China, brought together nearly five hundred field surveys from around the world and estimated that Earth holds some twenty quadrillion ants. Their dry carbon mass amounts to roughly twelve million tonnes, more than that of all the world's wild birds and wild mammals combined.

Change the measure, however, and the conclusion changes. More than fourteen thousand ant species and around three thousand termite species are not many against the full diversity of insects. There are more than twenty thousand bee species, of which only a small fraction are social. In the vast majority, a female builds her nest and lays her eggs alone. Solitary bees and wasps have not vanished because they “failed to evolve society.” They have diversified into far more species than their social relatives.

So “eusociality is very successful” holds for abundance and mass, but not for species counts. A nest containing several million workers is a sizable lump on the scales. In a species count, it contributes just one. The measure you rank by decides who comes first.

VII Three Questions and a Spectrum

Essay X set out three criteria for deciding whether a change of level has truly taken place. Have units that once reproduced independently lost that capacity? Has a new unit appeared that reproduces as a whole? Have conflicts at the lower level been systematically suppressed? Now we can test eusociality against them one by one.

For the first criterion, we cannot say yes or no to “eusociality” as a whole. We must arrange its lineages along a spectrum. At one end are the primitively eusocial sweat bees, paper wasps, and bumblebees.

Their workers have intact ovaries. They do not breed because the queen suppresses them, the season is wrong, or circumstances in the nest do not allow it. When the queen dies, workers can take over egg-laying. At this end, the criterion is only partly met.

Honeybees fall in the middle. A worker cannot mate as a queen does and found a new female lineage. Yet most retain ovaries and can lay male eggs. To say that honeybee workers have “completely lost the capacity to reproduce” is inaccurate. More precisely, the vast majority never reproduce in their lives. The normal reproductive route has been blocked, leaving a narrow opening through which they can produce drones.

At the other end are some ants and stingless bees. In ants such as the red imported fire ant and the big-headed ant, workers develop with no ovaries at all. The queen may die, but they can never become mothers. Here the first criterion is strongly satisfied.

Even ants are not uniform. In Diacamma, a genus found in India and Southeast Asia, there is no conventional queen caste; mated workers, called gamergates, take on reproduction. In some species, the gamergate bites off a pair of small thoracic appendages called gemmae from newly emerged nestmates, preventing them from mating. These are not immature wing buds. Removing them does not remove the ovaries; it closes the route to mating and sexual reproduction.

Termites present another case. Their colonies are extremely stable, with clearly distinguished soldiers, workers, and reproductives. Yet in many termites, workers retain developmental flexibility. If the king and queen die, some workers can molt several more times and become replacement reproductives. Tight integration and irreversibility are two different things.

Naked mole-rats score lower still on the first criterion. The other females' ovaries remain undeveloped because the queen suppresses them. When she dies, several females fight fiercely for her place. The winner's body grows longer, and she begins to breed. In their close relative the Damaraland mole-rat, nonbreeding animals are not necessarily shut in at home for life, either. Some studies suggest that many leave when rain softens the soil, seeking mates elsewhere and setting up on their own.

The strongest positive evidence concerns the second criterion. A honeybee colony can regulate its temperature like a body. The brood area stays at around 34 or 35 degrees Celsius throughout the year. In hot weather, workers fan their wings at the entrance and sprinkle water inside to cool the nest. In cold weather, they cluster together and generate heat by vibrating their flight muscles.

No single bee is regulating that temperature. The colony is. An ant colony has a life of birth, growth, maturity, the production of new colonies, and eventual death. As early as 1911, the American biologist Wheeler published “The Ant-Colony as an Organism,” likening the colony to a body: queens and males were the reproductive cells, workers the somatic cells. Such a colony later came to be called a “superorganism.”

But a colony can “have children” in two very different ways. Honeybees swarm. The old queen leaves with well over half the workers and establishes a complete new colony elsewhere, as though one body had split directly into two.

The departing swarm first hangs from a nearby branch in a large, humming ball. Hundreds of scouts search for tree cavities, then return and dance on the swarm's surface to report what they have found. More and more scouts come to favor the same site. When support for one place reaches a certain number, the whole swarm flies there together. The American bee researcher Seeley made this process the subject of a book, Honeybee Democracy.

Most people in southern China have seen a nuptial flight. On a hot, close evening after rain in late spring or early summer, a great cloud of winged insects suddenly gathers around a light. Soon after landing, they shed their wings all over the ground. Locally called “big water ants,” these are usually termite reproductives on their mating flight. Once on the ground, pairs shed their wings and crawl off one behind the other, looking for somewhere to found a nest.

Most ants also make nuptial flights. The old nest releases thousands or tens of thousands of winged queens and males. After mating, a queen lands alone, sheds her wings, digs a hole, and starts a new nest. In between, the colony must pass through a stage consisting of just one mother.

For the third criterion, the evidence in highly eusocial insects is strong, as the figures discussed earlier show. Here, too, though, suppression means lowering conflict, not reducing it to zero.

Put the three together and primitively eusocial bees and wasps meet each criterion only in part. Ants and stingless bees with permanently sterile castes meet all three far more strongly. This explains the two numbers in Section II. Fifteen or more is the number of entries onto this spectrum. Five is the number of times the irreversible end has been reached.

VIII A Change of Level That Comes in Degrees

In 1995, Britain's Maynard Smith and Hungary's Szathmáry wrote a book together. Maynard Smith had begun as an aircraft design engineer and turned to zoology only in his twenties. He belonged to Hamilton's generation and was one of those who brought game theory into evolutionary biology.

Their book, The Major Transitions in Evolution, listed eight “major transitions” in the history of life: from molecules to cells, from prokaryotes to eukaryotes, from single cells to multicellular organisms. One entry read, in the original, “Solitary individuals to colonies (with non-reproductive castes).”

“Non-reproductive castes” is the crucial phrase. The entry does not say “from solitary life to any kind of social life.” It places nonbreeding members at the center of the transition. In 2015, West and colleagues at the University of Oxford in Britain revised the list. They kept this entry and explicitly made the suppression of internal conflict part of the transition process.

The previous changes of level in this series had relatively clean answers. In Essay V, a bacterium was incorporated into another cell and became a mitochondrion. All three criteria held.

In Essay VIII, a multicellular body grew from a single cell, and germ cells became distinct from somatic cells. Again, all three held. This time, the degree to which they are met varies along a spectrum.

At one end of the spectrum, red imported fire ant workers have no ovaries. The colony clearly matters as a reproductive unit, and comparing the survival and reproduction of colonies is particularly useful. That does not make individual-level selection disappear. At the other end, workers of some sweat bees may still take over reproduction, making differences between individuals more conspicuous. Both levels can operate at once; the difficulty is finding their relative importance in each case.

The Americans Queller and Strassmann proposed a memorable way to think about this. To judge how much something resembles an individual, ask two questions: how much cooperation is there inside it, and how little conflict? A cell, a multicellular body, a red imported fire ant colony, and a group of sweat bees can all be plotted on the same diagram using those two measures. No natural gap separates them.

The broad middle is not a transitional state. Primitively eusocial bees and wasps are not “things on their way to becoming ants.” They have lived in their own positions for tens of millions of years, and some have returned to solitary life. They are permanent residents of this spectrum, not travelers passing through.

So we cannot record the phase transition in this essay with a simple yes. The accurate entry is: completed in some lineages, only partly completed in others. This is not a fudge. It means that “eusociality” is itself a label, not a criterion. To judge a lineage, we must return to the three questions and examine the lineages one by one.

What a construct cannot contain, yet cannot get rid of, this series calls the remainder. This essay has two.

One is the spectrum itself: from restraint imposed through behavior, to physiological suppression, to a fate fixed in development, to a body that has no ovaries at all. The spectrum is continuous. The lines drawn across it by our three criteria have no natural cutting point.

The other is the accounting dispute. Both sides perform the same experiments in the laboratory and count the same nests in the field. They argue about whether to account for the same process in terms of self and kin, or between nests and within them. Part of this argument really is just a matter of bookkeeping. Part of it is not.

When benefits and costs can be treated as fixed numbers, the correspondence between the two ledgers is easier to see. If a worker’s contribution depends on how many workers are already present, those numbers change with the colony’s state; a simple inequality with fixed coefficients no longer suffices. But kin selection is not thereby restricted to fitting results after the fact, nor does group selection alone make predictions. Both frameworks can model such interdependence. What matters is how explicitly each describes colony founding and reproduction, which quantities can be measured, and which predictions could turn out to be wrong.

With one tasty cow, Darwin once moved selection from the individual to the family. A hundred and fifty years later, biologists are still arguing over whether selection falls on a worker, a family, or the hive as a whole. Those thirty thousand workers are still collecting nectar, feeding larvae, standing guard, and eating the eggs their sisters have laid on the sly.

The criteria are still sifting. The count of obligate transitions still shifts with the thresholds used in coding. Between not breeding for now and never being able to breed at all, there is still no agreed cutting point.