The End of the Permian: A Construct Emptied Out

I The Last Permian Sea

About 252 million years ago, most of Earth’s continental land had gathered into Pangaea, stretching from near the North Pole toward the South Pole like vast open pincers. Warm seas lay within its embrace, and a great ocean covered much of the rest of the planet. Blocks such as South China still lay apart in the seas to the east.

The name Pangaea, meaning “all the land,” first appeared in the 1920 edition of a book by the German meteorologist Alfred Wegener. For proposing that continents could drift, he spent the latter half of his life being treated by his peers as an amateur. Later geologists named the sea held in that embrace the Palaeo-Tethys, after a sea goddess in Greek mythology. What is now South China was then a piece of land adrift in that sea, surrounded by warm shallows. The limestone beds at Meishan settled out there, one after another.

Those shallows were busy. Reefs rose from the seabed, built chiefly by calcareous sponges, along with a kind of coral long since vanished and bryozoans. Brachiopods grew thickly in the reef crevices and across the surrounding bottom. Like clams, they had two shells; they opened them and used a ring of cilia to filter tiny particles from the seawater.

Crinoids swayed in the current like flowers on stalks. Foraminifera the size of rice grains rolled about on the muddy bottom, so numerous that they could accumulate into whole beds of limestone. Trilobites were still there, reduced to a scant few species. They had already been living in the sea for more than 270 million years.

The land was hardly quiet either. Gorgonopsians, two or three metres long and armed with sabre teeth, were its top hunters. Pareiasaurs, built like rhinoceroses and clad in bony plates, cropped plants beside the rivers. In the tropical swamps of South China grew Gigantopteris, with leaves of extraordinary size. Much of the coal in South China’s mines today is what became of those Late Permian forests.

The Permian began about 299 million years ago and lasted more than forty million years. It was the last period of the Palaeozoic, a word meaning “ancient life.” The event this essay concerns is precisely what marks that era’s end.

This, then, was the last Permian sea. The name Permian comes from Perm, at the western foot of Russia’s Ural Mountains. In 1841, the British geologist Roderick Murchison travelled there, found a distinct succession of rock layers, and named that stretch of time after the place.

He had already named the Silurian after an ancient tribe of Wales. The Chinese name for the Permian, Erdie, means “two divisions”: in Germany, the strata belonging to this interval fall into two clearly distinct units, one above the other.

Then, in what amounted to the blink of an eye in geological time, the sea emptied. The reefs collapsed. Not a single species of the reef-building corals remained. Brachiopods went from being the commonest animals on the seabed to a minority.

The rice-grain foraminifera almost disappeared. Trilobites made their final exit. Current estimates suggest that roughly 60 per cent of marine genera and 80 per cent of marine species failed to enter the next age. Whichever measure we use, this was the greatest extinction in the history of life on Earth.

The most famous figure is “96 per cent of marine species went extinct.” That number has a history. In 1979, David Raup of Chicago published a paper in Science. Raup was a palaeontologist who liked to think with computers. Back in the 1960s, he had used just a few parameters to calculate every possible coiled shell shape, and discovered that natural shells occupied only a small corner of that space.

He did not have a list of the marine species alive before the extinction. No such list exists even today. What he did have was information from higher up the hierarchy: families and genera in the fossil record, far more reliable than species. A family counts as alive if even one member survives, and recognising a family is easier than recognising a species. He knew that about half the families had disappeared.

He did something ingenious. He looked at living sea urchins: how many species, on average, belonged to a genus, and how many genera to a family. With that hierarchical structure in hand, he could work backwards. How many species would have to be killed to make that many entire families disappear?

The answer was “as many as 96 per cent.” Three years later, he and his colleague Jack Sepkoski used lists of marine fossil families to chart the five mass extinctions. The end-Permian was the deepest trough in the curve.

In 1984, they saw something else in that curve: mass extinctions seemed to recur about every twenty-six million years. Some astronomers proposed that the Sun had an unseen companion star, which stirred comets into the inner Solar System with every orbit. They even gave it a name: Nemesis, the goddess of vengeance.

That star has never been found, and belief in the periodicity itself has dwindled. He began by knowing how many families had died, then worked backwards to calculate how many species would have to die to account for them.

The original wording was “as many as”: 96 was the upper end of that calculation. Later work brought it down. One reason was that early compilations had folded a smaller extinction about eight million years earlier into the end-Permian event. Put two losses on the same account, and the drop will naturally look larger. That smaller extinction occurred about 260 million years ago, coinciding with a vast eruption of basalt in southwestern China. The cliffs at the Golden Summit of Mount Emei in Sichuan are those stacked layers of basalt.

In 2016, the American palaeontologist Steven Stanley separated the two extinctions again and found that about 62 per cent of marine genera disappeared at the end of the Permian. Working backwards to species, and allowing for the tendency of species in the same genus to die together, he arrived at about 81 per cent.

That estimate of roughly 80 per cent is still inferred from genera. The inference requires assumptions about how many species a genus ordinarily contains, whether species in the same genus die together, and whether the number of species preserved within a genus reflects the true number. Change an assumption a little, and the answer changes a little. “Roughly 60 per cent of genera and 80 per cent of species” is the more defensible wording today.

Figures at the family level are the least suitable measure of a disaster’s severity. As long as one genus remains in a family, it is recorded as “surviving.” A family reduced from a hundred genera to one is still perfectly intact in family-level statistics. In reality, it has lost 99 per cent.

II Sixty Thousand Years

The most precise clock for this extinction is in China.

In Changxing County, Zhejiang, there is a small hill called Meishan, where local people once quarried limestone. The beds of grey rock exposed on its slopes were the seabed 250 million years ago.

From the latter half of the twentieth century, Jin Yugan and others at the Nanjing Institute of Geology and Palaeontology sampled the site layer by layer, identifying the fossils in each bed one by one. In 2000, they published Meishan’s extinction record in Science: large numbers of species disappeared together within just a few thin beds of rock.

Shangsi, near Guangyuan in Sichuan, and several other outcrops in Changxing, Zhejiang, are also important sections for studying this extinction. In 2011, Shen Shuzhong and colleagues at the Nanjing Institute of Geology and Palaeontology brought together several sections from South China and narrowed the event to less than 200,000 years.

Many geological boundaries are defined by a reference point in a rock section chosen for worldwide correlation: a “golden spike.” Not every boundary is defined this way. Essay Twenty-Three will examine the practice.

In 2001, the International Commission on Stratigraphy placed the boundary between the Permian and the Triassic in Bed 27 of the Meishan section, marked by the first appearance of a particular tiny conodont fossil. Conodonts are toothlike structures from the mouths of small, eel-like animals that vanished long ago. Less than a millimetre long and quick to change shape through time, they are among the most useful rulers for dividing marine strata.

The Russian researcher Christian Pander first described these little objects in 1856. They look like rows of minute saw teeth, amber-coloured under the microscope. What kind of animal had carried them in its mouth puzzled palaeontologists for more than a century.

Not until 1983 did researchers find the impression of a complete animal with its full set of conodont elements, in Carboniferous rocks near Edinburgh, Scotland: a soft-bodied, eel-like creature a few centimetres long, with a pair of large eyes on its head and a ring of tiny teeth in its mouth. The “golden spike” is driven into Chinese rock. The site is now a geological park.

Meishan has something else, still more precious: volcanic ash. Closely spaced, thin layers of ash lie between the limestone beds. Weathering has turned them into white or yellow clay, instantly recognisable against the grey limestone.

This ash came not from Siberia but from volcanoes around South China. Essay Nine introduced its zircon clocks. They record crystallization and closure of the uranium–lead system, not a clock starting when ash hits the ground. Researchers identify crystals formed near eruption and check for later disturbance before using them to constrain a bed’s age. At Meishan, such clocks are unusually closely spaced.

In 2014, researchers at the Massachusetts Institute of Technology in the United States redated these ash beds. Bed 25 at Meishan is itself a layer of white clay just a few centimetres thick, altered from volcanic ash. Local geologists can pick it out at a glance from the grey-black limestone.

Stand before the Meishan section and you can see the line without an instrument. Break open the limestone below Bed 25 and you will often find brachiopod shells and foraminifera the size of rice grains. A few beds higher, the rock grows monotonous and fossils are few. The beginning of the main extinction is placed at the base of Bed 25, its end in Bed 28.

The volcanic ash immediately beside these two points yielded ages of 251.941 million and 251.880 million years, each with an uncertainty of only about 30,000 years. Subtract one age from the other: the main extinction lasted about 61,000 years, with an uncertainty of roughly 48,000 years.

When two ages measured on the same instruments against the same standards are subtracted, some systematic errors cancel out. So “these two beds are about 60,000 years apart” is more reliable than “this bed is so many years old.” The sedimentary interval at Meishan is also very thin, and the researchers regarded 61,000 years as the maximum duration of the main extinction.

Another approach counts the rhythms that astronomical cycles leave in the strata. It gives a duration of no more than about 83,000 years. Both routes lead to less than 100,000 years.

How short is sixty thousand years? About as long as it has been since humans left Africa. In the history of life on Earth, this was death arriving almost all at once.

During that interval, the seawater began to burn. The phosphate in conodont elements preserves the oxygen-isotope composition of the seawater at the time. Converted into temperatures, it indicates that surface waters around Meishan warmed from about 23°C to about 33°C during the main extinction, then stayed hot.

Sun Yadong of the China University of Geosciences and colleagues inferred from South China records that equatorial surface waters in the first few million Triassic years may have approached 40°C—bathwater temperatures, well above the usual conditions across most tropical seas today. Sustained heat combined with oxygen shortage severely restricts habitat and can kill many marine animals. But tolerance and survival time depend on the species, acclimation and duration of exposure; there is no universal deadline of hours or days.

Carbon was swinging violently too. Carbon has two stable forms, one light and one heavy. Living things favour the lighter one, so coal, oil and all manner of organic material are especially rich in light carbon.

If the carbon in seawater suddenly becomes lighter, a great quantity of carbon from organic material or from underground must have entered it. At Meishan, carbon isotopes began to shift towards the lighter form about 60,000 years before the extinction, indicating a large influx of carbon from organic material or magma into the ocean and atmosphere.

Within one bed, the carbon-isotope signal lurched sharply and then bounced back over a few thousand to ten or twenty thousand years, almost instantaneous in the geological record. The disruption of the carbon cycle as a whole dragged on for more than 400,000 years. Those three durations describe three different things: the beginning of the input, its most violent pulse, and the time the entire system needed to regain stability.

The speed at which carbon arrives matters more than the total amount. Earth has its own thermostat: when atmospheric carbon dioxide increases, the climate warms, rainfall increases and rocks weather faster. Weathering gradually removes carbon dioxide from the atmosphere and ultimately stores it in limestone on the seafloor.

This thermostat is slow. It takes tens to hundreds of thousands of years to catch up. If carbon pours in over just a few thousand years, the thermostat cannot respond in time; first the seawater grows hotter, more acidic and short of oxygen. That abrupt swing within ten or twenty thousand years is a record of precisely this failure to keep up.

Whether the main extinction unfolded in one act or several is still debated. High-resolution records from Meishan and several other South China sections describe the losses as two concentrated pulses.

But the sedimentary sequence at Meishan is very thin, and differences in oxygen depletion at different water depths may have shifted the last appearances at different sites. The premature last appearances discussed in Essay Eleven can also stretch a brief extinction into a long tail. The safer conclusion is this: the main crisis at Meishan fits within 100,000 years; how many pulses it contained is a question for thicker sections to answer.

III Siberia

At the same time, the other side of Earth was splitting open.

Today, the Putorana Plateau in northern Siberia is a tableland built of basalt, layer upon layer. Along the river valleys, the hillsides rise in enormous steps, one above another. Geologists call this kind of terrain “traps,” from the Swedish word for stairs. These steps are what remains of the volcanism at the end of the Permian. Today the plateau is a World Heritage Site and one of the most remote on Earth, accessible only by helicopter or water.

Its scale is difficult to imagine. A single lava flow could be tens of metres thick. One lay on top of another, and around Norilsk their combined thickness exceeds three kilometres.

Age models estimate that the first phase of volcanism discharged about two-thirds of the lava and volcanic debris in roughly 300,000 years: about a million cubic kilometres, enough to cover the whole of China with more than a hundred metres of basalt. In 2021, a volcano near Reykjavík in Iceland erupted for six months and produced about 0.15 cubic kilometres of lava, already enough to draw thousands upon thousands of visitors. A million cubic kilometres means eruptions like that occurring several million times in succession.

Count the deep sills, dykes and other intrusions, along with the parts later eroded away, and the whole magmatic system ran to several million cubic kilometres. The differences between “two million,” “three million” and “five million cubic kilometres” in the literature largely reflect whether the count includes only the lava or the entire system.

Part of that magma still supports a city today. Norilsk, inside the Arctic Circle, is one of the world’s northernmost major cities and one of its largest sources of nickel and palladium. The ore occurs in the bodies of magma that once forced their way underground. The city was built on permafrost by labour-camp prisoners in the 1930s. For a month and a half each year it sees no sun; to this day, no road or railway connects it to the outside world. People and supplies travel by air and along the northern shipping routes.

The connection between the basalt and the extinction came later. In 1991, researchers at Berkeley in the United States used argon isotopes to date the Siberian basalt and found that its age almost coincided with the end of the Permian. The volcanoes became the prime suspect.

Others looked for a killer in the sky. In 2001, researchers reported extraterrestrial helium trapped inside cage-shaped carbon molecules in the boundary clay. In 2004, a structure beneath the seabed off northwestern Australia was proposed as an impact crater. Other laboratories subsequently failed to reproduce or confirm these findings. Unlike the asteroid in Essay Eighteen, the end-Permian event still has no generally accepted impact crater.

The timing of the volcanism in relation to the extinction needs to be considered in three stages. In the first, about two-thirds of the lava erupted over roughly 300,000 years before the main extinction, yet the biosphere did not immediately collapse.

In the second, about 251.9 million years ago, eruptions at the surface stopped. The magma went underground and spread sideways between the rock beds, exactly overlapping the main extinction window at Meishan. In the third, eruptions resumed and continued beyond the extinction. Of the three stages, only the short middle one lines up with the dying. And that was precisely when the lava stopped coming up and began working its way underground.

Magma spreading sideways underground forces itself between layers of sedimentary rock, then cools into flat, sheetlike bodies called sills. The basin it entered was the Tunguska Basin, whose strata held salt, limestone, coal and oil. These happened to be the very materials least able to take the heat. Heated coal and oil release methane and carbon dioxide; heated limestone releases carbon dioxide; salt reacting with magma releases gases containing sulphur and chlorine.

Many people have heard the name Tunguska. In 1908, a small object from space exploded above these forests, flattening more than two thousand square kilometres of trees. Two hundred and fifty million years earlier, beneath the same land, scalding magma had crowded in.

If gases containing chlorine and bromine reach the stratosphere, they can eat away at the ozone layer, just as the Freons manufactured by humans in the twentieth century did. As we will see in Section Four, pollen and spores close to the boundary may bear traces of this process.

In 2009, Norwegian researchers proposed the sill hypothesis: the environmental effects might depend less on how much magma erupted than on what the magma baked. Dating in 2017 then aligned the extinction with the first sills. The same cubic kilometre of magma might crystallise deep underground, erupt onto barren basalt, or push into beds of coal and salt. The amounts of carbon, sulphur and halogens it sends into the atmosphere can differ many times over.

Different gases also do different things. A large pulse of sulphur dioxide forms tiny sulphate particles high in the atmosphere. These block sunlight and cool the surface for a few years, then fall with the rain as acid rain.

Carbon dioxide and methane, by contrast, remain in the atmosphere for a long time and warm the planet over the longer term. To speak vaguely of “volcanic gases” is to muddle together effects that work in opposite directions.

One piece of evidence is especially vivid. In 2011, Canadian researchers found distinctive particles in high-Arctic rocks from the end of the Permian: hollow, perforated spheres of carbon, almost identical to the fly ash emitted from the chimneys of coal-fired power stations today. They interpreted them as evidence that Siberian magma had set vast coal beds alight, sending smoke and dust on the wind across the Arctic Ocean of the time to settle into the sea thousands of kilometres away.

The hypothesis has gaps too. Widespread sill intrusion continued for about half a million years, an order of magnitude longer than the main extinction’s sixty thousand. The authors of the 2017 study acknowledged that a more concentrated release still had to be identified within that interval.

Many of the dated sills also clearly postdate the main extinction. Estimates of the total carbon released in gases range from thousands to tens of thousands of gigatonnes, an uncertainty of more than an order of magnitude. It depends on how much sediment was baked, how much carbon it contained, and how much of the gas actually reached the surface.

IV How They Died

Once the gases entered the atmosphere, a chain of changes followed.

The first brought heat and oxygen together. The hotter the water, the less oxygen it can hold. The higher the temperature, the faster a cold-blooded animal’s metabolism runs, and the more oxygen it needs. In 2018, researchers at the University of Washington and Stanford University in the United States first assembled laboratory measurements from dozens of species of modern marine animals: how hot and oxygen-poor could the water become before a crab or a fish could no longer sustain normal activity?

Then they put the two effects into a model. Warming squeezed the “aerobic space” in which animals could live from both ends, supply and demand. The pool was growing shallower just as the fish were growing thirstier.

The model made a testable prediction: extinction should be more severe at high latitudes. Cold water starts out rich in dissolved oxygen, and the animals there are adapted to that abundance. When the water warms, they lose the largest proportion of it. The researchers checked the fossil record. It matched. The model could also reproduce the greater part of the losses at genus level.

The animals in the sea tried to escape too. Song Haijun of the China University of Geosciences and his colleagues compiled marine fossils from before and after the extinction. They found that species became exceptionally scarce for a time in tropical shallows. The usual diversity gradient, richest at the equator and diminishing towards the poles, was almost flattened out. The equator had become too hot to stay in.

Something like this is happening today, on a much smaller scale. Syntheses by oceanographers show that the global ocean has lost about 2 per cent of its oxygen since the 1960s. Coastal “dead zones” are multiplying, places where bottom waters contain too little oxygen for fish and crabs to remain.

The second change was oxygen loss, and even toxicity. Records from iron, trace elements and organic molecules all indicate a great expansion of oxygen-depleted water. One of the most vivid pieces of evidence comes from bacterial pigments.

Green sulphur bacteria need both light and hydrogen sulphide. They can live only in a very particular place: somewhere sunlight reaches and hydrogen sulphide is present in the water. In a normal ocean, the two are far apart. Light is at the top, hydrogen sulphide in the depths, with a layer of oxygenated water between them.

In 2005, researchers found remnants of these bacteria’s pigments in boundary sediments at Meishan and in Western Australia. Poisonous water had risen all the way into sunlit depths. The Black Sea today is just such a stratified sea. Below about 100 to 200 metres, its water has no oxygen and is full of hydrogen sulphide. Green sulphur bacteria inhabit the thin layer at the boundary between the oxygen-bearing and sulphide-bearing waters. Only a minute fraction of the surface light reaches them there, but they can live on that little.

This condition varied by region and depth. The entire ocean did not turn into a single cauldron of hydrogen sulphide at once. Some researchers have also proposed that hydrogen sulphide escaped from the sea into the atmosphere, poisoning coastal land and even damaging the ozone layer.

There is one indirect piece of evidence for the ozone part of the story. Rocks near the boundary often contain malformed pollen and spores: four stuck together, unable to separate, their shapes distorted. In 2018, researchers grew a dwarf pine in indoor chambers with increased ultraviolet radiation. Its pollen developed the same deformities, then returned to normal once the irradiation stopped.

The third change was acidification. Boron occurs in seawater in two forms, whose relative proportions vary with pH. Organisms favour one form when they build their shells, so the boron isotopes in a shell can be used to work back to the pH of the water in which it formed.

In 2015 and 2020, British and German researchers, respectively, used this method on carbonates of the time and on well-preserved brachiopod shells. They read a clear acidification signal during the extinction; the British study placed it in the second extinction pulse. A shell 250 million years old had told people how acidic that sea had been.

The pH scale is logarithmic. A fall of a few tenths of a unit may not sound like much, but it means the concentration of hydrogen ions has doubled or more. For animals that must extract calcium carbonate from seawater to build their shells, that makes construction harder and existing shells more liable to dissolve. Since the Industrial Revolution, the pH of today’s ocean has already fallen by about 0.1. In some experiments, oceanographers have already observed signs of dissolution in the thin shells of tiny pteropod snails.

The fourth change was related to acidification, but distinct from it. As carbon dioxide increases in seawater, it also rises in animals’ body fluids, making blood and other fluids more acidic. This is called hypercapnia. People may feel dizzy and drowsy after a long time in a crowded meeting room with the doors and windows shut; if carbon dioxide builds up to a sufficiently high concentration, it can also alter the blood’s acid–base balance.

An animal might build no calcium carbonate shell at all and still suffer because it cannot expel carbon dioxide effectively. Another might carry a thick shell yet regulate its body fluids well. The same external change takes different forms of pressure in different bodies.

The atmosphere recorded this upheaval too. In studies published in 2021 and 2022, researchers used carbon isotopes in fossil plants and carbonates in ancient soils to estimate that atmospheric carbon dioxide concentrations increased four- to sixfold across the Permian–Triassic transition. Plants favour light carbon when they take in carbon dioxide. The more carbon dioxide there is in the air, the more pronounced that preference becomes. A leaf is a little gas meter.

These are not competing answers. Current syntheses join them into a chain: large quantities of carbon and sulphur entered the atmosphere, bringing warming, oxygen depletion and acidification. All three occurred almost simultaneously and compounded one another.

A chain with many factors carries a danger: whenever an observation fails to fit, another factor can be added to patch the gap. The model of heat-driven oxygen shortage has a special standing because it said in advance what we should see, and where. The fossil record then answered.

V Who Survived

Now for a different question: who survived, and why?

In 2007, Andrew Knoll of Harvard University in the United States and his colleagues proposed a framework. If the pressures were high carbon dioxide, warming, oxygen shortage and acidification, it should be possible to predict part of an animal’s risk of death from its bodily organisation: how it breathed, how its blood circulated, how fast its metabolism ran, whether it could regulate the acidity of its body fluids, and how much calcium carbonate its shell required. The method was to derive the traits survivors should share from the environmental mechanism, then check them against the fossils. It could be wrong.

The most dangerous combination was a heavy calcium carbonate skeleton, a very low metabolism, and weak respiratory and circulatory systems. Reef-building corals, calcareous sponges, many brachiopods, bryozoans and echinoderms fell towards this end. Most spent their lives fixed to the bottom, slowly filtering water with cilia. They had no specialised respiratory pigments, scarcely regulated the acidity of their body fluids, and needed large amounts of calcium carbonate for their shells.

At the other end were animals that actively ventilated, had higher metabolisms and more developed circulatory systems, carried oxygen-binding pigments in their blood, and could regulate the acidity of their body fluids: many molluscs and arthropods, along with fish. These traits did not guarantee that a genus would survive. They merely changed its chances.

In 2011, the American researchers Matthew Clapham and Jonathan Payne put this idea into a multivariate statistical analysis. They considered skeletal mineralogy, physiological buffering capacity, way of life and geographical range together, asking which best predicted whether a genus went extinct. The physiological selectivity was pronounced.

A wide distribution, on this occasion, offered surprisingly little help. Ordinarily, the more widely a group is distributed, the harder it is to wipe out: a local disaster cannot kill all of it. At the end of the Permian, the pressures reached around the globe. The insurance policy of a broad range stopped paying out.

In 2022, another study used more than 1,200 marine genera from South China and let machine-learning algorithms identify for themselves which features best predicted extinction. The algorithms had not been told the researchers’ hypothesis. They picked out depth range, the number of species within a genus, shell composition and physiology. Later, researchers tested the oxygen-carrying capacity of respiratory systems across more than a thousand genera. They also directly measured tolerance of warming and low oxygen in living relatives, then fed those results into extinction simulations. The selectivity they obtained resembled the pattern in the fossil record.

Crinoids were almost wiped out. Only one lineage crossed the boundary, and every crinoid in the sea today descends from it. Sea urchins came close too. The very animals Raup used for his conversion almost went extinct themselves; all living sea urchins may descend from just one or two lineages that made it across.

The brachiopods offer an especially clear example, and one connected to the sieve discussed in Essay Eleven. Their shells are mostly calcite, and their fossil record is the most complete of any marine animals. Their losses are therefore also the clearest to see.

Consider a few groups in detail. Rugose and tabulate corals had built reefs in Palaeozoic shallows for more than two hundred million years. At this boundary, not one species remained. Fusulinid foraminifera were only the size of rice grains, shaped like little spindles. In the tropical shallows of the Late Permian, they were so abundant that their remains alone could build bed upon bed of limestone. At the boundary, they were almost entirely wiped out.

The Chinese character for these creatures, 䗴, was coined by the geologist Li Siguang. As a young man, he studied at the University of Birmingham in Britain; later, his research on the fusulinids of northern China earned him a doctorate there. In 1927, he published a monograph on northern China’s fusulinids, using them to subdivide the coal-bearing strata of the region. Coal prospectors gained another ruler.

He later became the first minister of geology of the People’s Republic of China. A foraminifer the size of a grain of rice had been abundant enough in the Permian sea to pile into mountains. More than two hundred million years later, it helped people find coal inside mountains. Brachiopods, the commonest filter feeders on the Palaeozoic seabed, suffered far greater losses than many molluscs.

Trilobites deserve a separate word. They were immensely important in the Cambrian and Ordovician seas, but by the end of the Permian only a very few species remained. Their end shows that longevity by itself guarantees nothing. It cannot, however, serve as proof that “even the most flourishing can die,” because they had long ceased to flourish.

Those that crossed the boundary each had their own arrangement of traits. One kind of brachiopod survived with a thin, tongue-shaped shell made of calcium phosphate and chitin rather than calcium carbonate. Its close relatives still live in coastal mudflats in China and Japan, where local people dig them up to eat. Their common Chinese name means “sea bean sprouts”: a little shell trailing a long, translucent stalk, like a bean sprout. The general appearance of this group has changed very little from the Cambrian to the present, and they are often called living fossils.

Bivalves suffered heavily, but several lineages made it through. They actively pumped water over their gills and had more active blood circulation. Some gastropods, ammonoids, crustaceans and conodont animals also had members that crossed the boundary.

Only a very few ammonoid lineages remained. Yet in the Triassic they diversified into hundreds of genera and filled the ocean once again. How much of a group remains when it crosses the boundary and how much it can grow back afterwards are two different matters.

The rulers of the seabed changed too. In the Palaeozoic, the bottom belonged to brachiopods. From the Mesozoic onwards, bivalves and snails gained the upper hand. Go to the shore and collect shells today, and almost everything you pick up will be a bivalve or a snail; brachiopods have become hard to find. The American palaeontologist Stephen Jay Gould compared the two groups to “ships that pass in the night”: one receding, one advancing, their moment of passage at the end of the Permian.

The roll of survivors reads like a list of those who happened to be equipped suitably in those particular respects. Some genera in groups with strong buffering capacities still went extinct; some members of groups with weak buffering survived.

The model predicts differences in probability. One or two counterexamples cannot overturn it. Genera within the same major lineage share both ancestors and bodily structures. Treating them as independent data points exaggerates the role of physiology, so modern analyses include their relationships in the model too.

VI On Land

Something enormous happened on land as well. We simply cannot see it as clearly as we can in the sea.

In the Late Permian, the high latitudes of the southern continent were covered by forests dominated by Glossopteris, a tree with tongue-shaped leaves. In 1912, the British explorer Robert Falcon Scott and his party died on their return journey after reaching the South Pole. Months later, the search party found their tent. On a sledge were about sixteen kilograms of rocks collected beside a glacier. Among them were fossil Glossopteris leaves.

They had hauled those stones with them to the very end. The fossils were among the early evidence that Antarctica had once been forested, and that the southern continents had once been joined.

Glossopteris takes its name from its tongue-like leaves. It was a seed plant, with some seeds borne on the leaves themselves. In the Late Permian, it grew everywhere from South America, South Africa and India to Australia and Antarctica. That map of its distribution is itself evidence that the southern continents were once connected.

In 2019, researchers working in Australia’s Sydney Basin used pollen, spores and high-precision dates from volcanic ash to establish that the Glossopteris forests there collapsed about 370,000 years before the main marine extinction at Meishan. South Africa’s Karoo Basin had long been treated as a classic section showing synchronous extinctions on land and at sea. High-precision ages published in 2020 placed its biotic turnover earlier. In 2026, another study argued that the classic “extinction horizon” itself did not hold up. Combining hundreds of metres of strata, high-precision zircon ages, palaeomagnetic evidence from several sites and isotopes from ancient soils, the study concluded that the classic extinction line did not represent a single simultaneous surface.

That position is still disputed. A 2024 terrestrial record from the ancient tropics of South China, meanwhile, gave the opposite sequence: the crisis on land followed the one in the sea.

Around Dalongkou in Jimsar, Xinjiang, another continuous sequence of river and lake deposits runs from the Permian into the Triassic. It contains dicynodont bones, pollen, spores and volcanic ash, and is an important Chinese section for studying the terrestrial side of the event.

Taken together, these sites suggest earlier losses of high-latitude vegetation, a challenge to synchrony in the Karoo, and later losses in tropical South China. There is no single extinction line on land that all the sections jointly confine to those sixty thousand years. High-latitude forests, tropical vegetation and inland tetrapods could, after all, have faced different pressures and responded at different speeds.

The existing record cannot yet give a robust global percentage for plant extinctions. Giant relatives of dragonflies, with wingspans of about seventy centimetres, still flew through Permian skies. They did not fly into the Triassic. The insect record comes from only a few exceptionally well-preserved fossil localities, and should not be converted into a global percentage either.

Large quantities of slender, filamentous microfossils have also been found in terrestrial beds near the boundary. At first, some researchers interpreted them as fungi: vast forests had died, and the rotting wood had fed a fungal bloom.

Others later argued that the filaments resembled algae. The “fungal spike” story remains suspended between interpretations. The next essay examines the multimillion-year Early Triassic coal gap, commonly estimated at roughly five to seven million years depending on definitions and dating. It records the absence of coal-forming ecosystems, not the disappearance of every forest.

Our own ancestors were on that line as well. On land at the end of the Permian lived a group of cynodonts ancestral to later mammals. They were the size of a cat or smaller.

A fossil from the earliest Triassic of South Africa’s Karoo Basin preserves a cynodont curled up in its own burrow, with an injured amphibian squeezed in beside it. Synchrotron scans allowed researchers to see everything inside. Retreating underground may have been one way to endure the worst of it.

The most famous survivor on land was Lystrosaurus. Its scientific name means “shovel lizard,” referring to its flattened face. A pig-sized herbivore belonging to the dicynodonts, it had a horny beak like a turtle’s and a pair of small tusks in its upper jaw.

After the extinction, it almost monopolised the Early Triassic of South Africa’s Karoo. In some beds, more than 90 per cent of the vertebrate fossils belong to it. It has been found in South Africa, India, Xinjiang in China, and Antarctica.

In 1969, the American palaeontologist Edwin Colbert found Lystrosaurus bones on an Antarctic mountainside. A land animal could not have swum across the ocean; it became powerful evidence that the southern continents had once been joined. In 2020, researchers sectioned Lystrosaurus tusks found in Antarctica and saw bands of tightly packed stress lines in the growth record, much like those left by hibernating animals today. Through the long Antarctic polar night, it may have endured in a hibernation-like state. The answer to how it survived may lie in this ability to slow its body down.

Gorgonopsians did not cross the boundary. The places they vacated were soon taken by another branch of reptiles, the archosaurs. From them would later emerge crocodiles, pterosaurs and dinosaurs. When Lystrosaurus was everywhere on the post-extinction wastelands, the distant ancestors of dinosaurs were probably still just some of the small reptiles among them, slight in build and long in limb.

The rules of selection differed between land and sea. Among marine animals, the amount of calcium carbonate in the body, the mode of respiration and the regulation of body fluids can be used to test the crisis model.

Terrestrial tetrapods and plants were not immersed in seawater’s carbonate system. Their risks were more likely to arise from a combination of heat, drought, vegetation change, wildfire, toxic gases and ultraviolet radiation. There is only one true common denominator: having been abundant, and having lasted a long time, guaranteed no passage across the boundary. Glossopteris had once filled the forests of the entire southern continent. Several long-established lineages of Karoo tetrapods also disappeared during this interval.

VII The Sea After the Disaster

The first post-extinction seas were unsettlingly empty. There were no reefs on the bottom, no broad beds of brachiopods, and fewer animals to stir the mud and sand. In many places the sediments preserve fine lamination, layer after layer undisturbed. That is what a seabed with very little animal activity looks like.

Immediately above the boundary in many shallow-water sections of South China, something appears in the rock: microbialites. These are stones grown by microbes such as cyanobacteria, which bind sediment layer by layer. They resemble cauliflowers or steamed buns, and occur in strata of the same age in South China, Iran and Turkey.

They were everywhere in the seas before the Cambrian. Later, as animals multiplied, grazing and disturbance of the sediment pushed them back into a few extreme environments. The extinction cleared away the animals that ate them, and suddenly they spread across the shallow seafloor again. Essay Nine described how later animals dismantled the microbial mats of the Ediacaran. Here, the dismantlers were gone, and the mats came back.

Among animals, a few species became the leading figures after the disaster. A small bivalve called Claraia, with a thin, flat shell, carpeted the Early Triassic seabed. Some shale bedding planes in Guizhou and Guangxi are almost entirely covered in its shells. It can be found from China to the Alps.

Large, complex reefs took longer to return widely, often an estimated eight to ten million years. Yet small local sponge–microbial reefs already appeared in the Early Triassic; animal-built reefs were not wholly absent. By the Middle Triassic, scleractinian corals—the group that builds reefs today—were becoming important builders. They cannot simply be treated as surviving Palaeozoic rugose corals; their precise ancestry remains debated.

New leading players in the sea arrived quickly too. Rocks about 248 million years old near Chaohu, Anhui, have yielded a small, short-snouted marine reptile, one of the earliest members of the ichthyosaur lineage. Only about four million years after the extinction, reptiles from the land were already back in the sea hunting prey.

Tongue-shaped brachiopods and some small gastropods were also especially common at this time. Palaeontologists call them disaster taxa: with other animals gone, they spread rapidly over the emptied seafloor, then retreated to more ordinary positions as ecosystems slowly recovered. Recovery took millions of years. That is the next essay’s story.

The hottest places stayed empty longest. For a time in the first few million years of the Triassic, fish and reptile fossils were almost absent from both sea and land near the equator. These animals remained at the cooler middle and high latitudes. The equator became a belt where life was thin on the ground.

The survivors shared another look: small. After the extinction, individuals in many groups became noticeably smaller. Borrowing the name of the land of tiny people in Gulliver’s Travels, some researchers call this the “Lilliput effect.” Heat, oxygen shortage and scarce food might all favour small bodies, or prevent animals from reaching full size.

One set of species dies out while another does not. Does that amount to selection at the species level? In 1975, the American palaeontologist Steven Stanley proposed species selection: regard species as units that can be “born” and can “die.” If certain species are more likely to give rise to new species or less likely to go extinct because of properties that belong only to the species as a whole, such as its range or the way its populations are distributed, then over long periods those differences can alter the composition of a lineage.

In 1986, a further distinction divided this into two levels. Some species leave more descendant species than others: that is sorting. Only when the difference can be explained by properties belonging to the species as a whole does it count as species selection.

The end-Permian offers a concrete test. Geographical range is a property of an entire species or genus. It makes no sense to ask whether an individual animal has a wide distribution.

If range can consistently protect a group, it becomes a candidate for species selection. But, as Section Five showed, range was not particularly useful this time. The respiratory pigments, circulatory arrangements and shell compositions that did matter are traits carried by each individual animal.

When a genus disappears because all its members have bodies built that way, the result is sorting at the level of genera and species, but the cause can be explained from individual physiology upwards. Those who regard mass extinction as a higher-level process have reasons of their own: species and genera really are the statistical units of extinction; a species’ range, population structure and ecological breadth really do alter its probability of extinction; and mass extinctions can erase whole batches of species in an extremely short time. The other side demands stricter conditions. If what actually determines survival is the capacity every individual shares to tolerate heat and oxygen shortage and to regulate acidity, then disappearance at the species level is simply the aggregate of causes operating at the individual level.

Sorting occurred. That did not shift the unit of selection to the species. A crowd of people can be sifted by height without the crowd thereby becoming a new organism.

VIII Survival Is Not Superiority

What this series calls a construct comes down, in the end, to drawing a line. What lies inside counts; what lies outside does not. Humans may draw that line, as with a set of criteria. Nature may draw it, as with a membrane.

The line in this essay is the map of divided labour that the Palaeozoic sea had spent two hundred million years drawing: who built reefs, who filter-fed, who crawled on the mud, who swam in the water. The reefs, the meadows of brachiopods, the seafloor paved with rice-grain foraminifera: the whole structure was emptied out in sixty thousand years.

Were those left behind, then, the better ones? In 1991, Raup wrote a short book with that very question in its title: Extinction: Bad Genes or Bad Luck?

His point was that much of the dying in a mass extinction might have little to do with the ordinary meaning of “good” or “bad.” It could be like a hail of bullets, who gets hit depending partly on luck. He offered another comparison: on a battlefield swept by gunfire, the surviving soldiers need not be the best trained.

In the book, Raup divided extinction into three kinds. One was the field of bullets: who lives and who dies has nothing to do with the body. Another was fair game: those well adapted in ordinary times also survive well in a disaster.

The third he called wanton extinction. Death is selective, but the traits selected have no connection to what ordinarily counts as good or bad, and may even reverse it. The end-Permian evidence comes closest to this third kind.

It was clearly not pure luck. As the preceding sections showed, respiration, circulation and shell composition all changed the chances of crossing the boundary.

Yet it was not “good” in the ordinary sense either. An animal able to sustain its metabolism in water at 33°C, low in oxygen and high in carbon dioxide, did have higher fitness in that environment. That was its body, not luck. That marking scheme belonged to a particular crisis environment, with some pressures continuing into the Early Triassic—not to an unchanging scale valid for every age.

Rugose corals are also known as tetracorals, a name derived from the arrangement of their internal partitions in multiples of four. Many grew as solitary individuals, shaped like a curved cow’s horn, hence the familiar name “horn corals.” They can readily be knocked out of Palaeozoic limestone.

The corals that build reefs in today’s seas belong to another group, with partitions arranged in multiples of six: the hexacorals. They did not begin building reefs until the Middle Triassic. Rugose corals had been extraordinarily successful in the seas for more than two hundred million years before that.

They built the Palaeozoic reefs, ranged widely and existed in immense numbers. At that particular moment, a heavy calcium carbonate skeleton combined with relatively weak respiration happened to land in the highest-risk box. They had not become worse. The environment had changed the marking scheme.

“Dominant before the extinction” can also be broken down into testable parts: how common a group was in one place, how often it appeared in the fossil record, how widely it ranged, and how many genera and species it contained. Range has already been tested. This time, it did not help much.

Frequency of occurrence also reflects the intensity of sampling and must be read with care. A group containing more genera is, on purely statistical grounds, more likely to leave at least one survivor.

Taken together, none of these measures shows that “the more flourishing, the safer.” A standard came into force for a few tens of thousands of years, and a bodily design that had lasted more than two hundred million was judged unfit.

Most of those unfit bodies never came back. Rugose corals, fusulinids and trilobites remained only in the rocks. Brachiopods survived but never regained their place as rulers of the seabed. An examination lasting a few tens of thousands of years determined the shape of the oceans for the next 250 million.

Gould asked another question: if we rewound the tape of life and played it again from the same starting point, would the result be the same? The end-Permian’s answer is that, on the replay, changing only the temperature, oxygen and seawater chemistry during those sixty thousand years might replace more than half the list of survivors.

The part that a construct cannot accommodate and cannot eliminate is what this series calls the remainder. In this essay, the remainder does not lie outside the evidence. The vanished groups have names, distributions and abundances. They have been studied for two hundred years.

Nor does it lie in the sieve. At Meishan, volcanic ash beds are closely spaced, the carbonate sequence is continuous, and conodonts subdivide the strata finely. The record of this event is among the best in the entire Palaeozoic. The remainder lies in something we think we have but do not: a ruler that stays the same across environments.

If such a ruler existed, “who survives an extinction” would be a question of better and worse, and the roll of survivors could be read as a ranking. In the material we actually have, every measurement yields a set of traits bound to a set of conditions.

Change the temperature, oxygen and seawater chemistry, and the ranking of the same groups may change. That can be tested: take the same groups into another extinction and see whether their ranking by risk holds.

Researchers have begun doing this. So far, the results show no stable ranking that carries across events. Essay Eighteen will take the same question into another extinction.

The criteria are still sifting. That percentage is still being recalculated. “Best” still has no definition that works across events.