Cenozoic Radiation: Not Inheritance, but Occupation

I The Day of the Boundary, and Who Survived

Sixty-six million years ago, an asteroid roughly ten kilometers across struck what is now Mexico's Yucatán Peninsula.

How do we know? At more than a hundred sites, a thin clay layer carries unusually high iridium, rare in the crust but much more abundant in meteorites. In 1980 researchers used that anomaly to propose an impact. Oil exploration had already detected the unusual structure beneath Yucatán; in 1991 researchers identified it as a large impact crater linked to the boundary event.

Below that clay layer, there are dinosaurs. Above it, there are none. Except birds.

We need to pause over those two words: “except birds.” Birds are not distant cousins of dinosaurs. They are dinosaurs. They branched off within the theropods, and the more than ten thousand bird species flying in our skies today are that branch's only surviving descendants.

Strictly speaking, then, not all dinosaurs went extinct. All dinosaurs other than birds went extinct. Whenever this essay says “the dinosaurs were gone,” that is what it means.

That line has consequently become one of the best-known boundaries on the geological timescale. The period before it is called the Cretaceous; it ended sixty-six million years ago.

The first interval after the line is the Paleocene, which runs from sixty-six million to roughly fifty-six million years ago. When this essay refers to “the boundary,” this is the one it means. What follows is the story everyone knows: the dinosaurs disappeared, and mammals took over.

Before we go any further, though, let us look at who was actually walking the Earth that day. It is a question people seldom ask, and the answer differs from the usual picture.

The mammals under the dinosaurs' feet were not a crowd of nearly identical little bug-munchers. The Mesozoic already had burrowers, climbers, gliders, semiaquatic mammals, dedicated plant-eaters, and meat-eaters. They belonged to several distinct major lineages, with differences far greater than those between a shrew and a squirrel today.

One of those lineages was the multituberculates. Their molars carried several rows of pointed cusps, like a set of tiny files, well suited to grinding seeds and tough plant tissue. They were highly successful in the Mesozoic and lived on long past the boundary. They did not merely squeeze through by the skin of their teeth: having made it through, they flourished for tens of millions of years before finally leaving the stage.

The gliders deserve a word of their own. As early as the Mesozoic, some mammals were spreading membranes between their limbs and gliding from tree to tree. This began far earlier than people generally imagine.

There were mammals in the water, too. We know of Mesozoic mammals with limbs suited to paddling and teeth suited to catching fish, living semiaquatic lives. Others turned their forelimbs into shovels for digging and lived underground year after year. Burrowing, climbing, gliding, swimming, eating plants, eating meat: this list was already in place while the dinosaurs were still here.

Researchers have measured the lower jaws of Late Cretaceous mammals, specimen by specimen, and plotted the results. One axis represents bite force, the other the kinds of food the teeth could process. Each species becomes a point.

If “all mammals in the age of dinosaurs were little bug-munchers” were true, those points should huddle in one small corner. In the actual plot, they spread out over a considerable area.

In other words, while the dinosaurs were still around, the mammalian branch had already run many experiments in what it was possible to eat.

One thing needs to be established now, because every later section will depend on it: on the day of the boundary, the starting line was not a blank slate.

Now for the protagonists. This essay follows placental mammals, and we first need to give that term something solid to stand on. All the debates ahead concern when this particular branch began to spread out.

Placentals include most familiar mammals: dogs, cats, cattle, horses, elephants, whales, bats, mice, humans. Living mammals have two other major branches. One is the marsupials, including kangaroos and koalas. Their young are born very immature and develop through prolonged nursing. Many species have a pouch, though not all do.

The other is the group that includes platypuses and echidnas. They lay eggs.

The name “placental” can mislead: marsupials also have functional placentas. The main contrast is not presence versus absence, but how development is divided between pregnancy and nursing. Placentals generally develop longer in the womb, yet vary greatly at birth. A foal stands within hours, while a newborn kangaroo is peanut-sized. But a newborn mouse is helpless too; a foal cannot stand for every placental.

There are now a little over six thousand mammal species worldwide. The overwhelming majority are placentals. Marsupials account for a little over three hundred species, while the egg-laying branch has only five left.

There is another reason to follow this thread: for the questions at issue, it has the richest body of evidence. On the molecular side, there are time-calibrated trees calculated from genomes. On the fossil side, there are studies that model large collections of fossil occurrence records together. On the morphological side, there are analyses that compile skeletal characters into enormous matrices and use them to calculate trees. All three approaches offer directly comparable answers about the same points in time. Finally, then, what is this essay going to do?

“Taking over” sounds natural enough. Quietly, though, it packs in three claims. There is a set of places. Those places stood empty after the dinosaurs vanished. The newcomers moved into them.

Each claim is stronger than the last. The first says the world is divided into compartments. The second says those compartments were empty at a particular moment. The third says the newcomers entered those very same compartments.

None of these claims is self-evident. Section Eight will give our conclusion, and it will have a slightly more interesting shape than either “right” or “wrong.” Of the first two claims, one holds up and one does not. With the third, even the way the question is posed is a problem.

The next seven sections move forward through time. Once we have followed them through and look back, “not inheritance” will hold up, while “occupation” will need qualifying.

II The First Million Years

The first million years after the boundary give us the most vivid part of this essay, because the record is remarkably detailed. How detailed? At several North American sites, recognizable Paleocene mammal assemblages can already be placed within roughly the first eighty thousand years after the boundary.

Eighty thousand years is a very short span in geology. The entire Cenozoic lasts sixty-six million years. Within that expanse, eighty thousand years takes up about as much as eight seconds in a two-hour film. And we can make out several changes of cast within those eight seconds.

Earlier high-resolution work in the same region had already shown rapid turnover in communities at the very beginning of the Paleocene. The boundary was not followed by a long silence, with life only slowly stirring much later. The cast began changing almost at once.

Such resolution requires several things to be present together. Sediment must have accumulated steadily at the site throughout the interval, without parts of the record being washed away. The rock layers must contain volcanic ash that can be dated directly, or some other marker of age.

And there must be enough fossils to collect and count a batch from every layer. Lose any one of those three conditions, and the resolution falls. Very few sites meet all three, which makes every one of them precious.

One site in Colorado is particularly good. Its rock layers lie one above another in an unbroken succession, so that you can count your way upward, layer by layer. A rock face that can be read continuously in this way is called a section. On this one, the following sequence can be read.

By about a hundred thousand years after the boundary, mammal species numbers had clearly rebounded. At roughly three hundred thousand years, animals weighing around twenty-five kilograms appeared: about the size of a medium-sized dog.

At roughly seven hundred thousand years, the largest had reached around fifty kilograms. Connect those three numbers, and you get a rising curve, climbing quite fast. So what was walking the ground at the time? Paleocene mammals looked like nothing alive today.

Researchers have measured their ankle bones and limb bones one by one. Ankle bones are particularly useful because they directly reflect how an animal moves: whether it runs fast, climbs, or slowly shuffles along the ground. The thickness of its limb bones reflects how much weight they carry and how forcefully it moves.

The result: in the robustness of their bones and the construction of their ankles, these Paleocene animals resemble neither their Late Cretaceous forebears nor any living group. They occupy a region of their own.

Another study, comparing combinations of functional traits in early Cenozoic North American mammals, reached the same conclusion. The Paleocene had distinctive combinations that later became rare and eventually disappeared. This was not a world of “modern animals, only smaller.”

If we could return to North America half a million years after the boundary, we would not find little deer, little wolves, and little squirrels. We would find sturdy animals, probably not especially quick, walking in ways that have no counterparts today. Many of those branches disappeared over the following tens of millions of years, leaving no living descendants; other lineages continued.

These animals barely appear in the familiar account, because that account cares only about who won in the end. They were neither winners nor losers. They were the actual inhabitants of that stretch of time.

It is worth keeping one fact in mind: the Paleocene lasted about ten million years. Ten million years is a long time, some three or four times the span from the appearance of the first members of our genus, Homo, to the present.

This was no brief transition between two ages. It was a whole world in its own right, with its own vegetation, food webs, and animals. And the whole of it gets pressed into the single sentence “Mammals rose after the dinosaurs went extinct.” While we are here, we should also say what that rising curve actually means. There is an entire system behind a number for body size.

To keep a fifty-kilogram animal alive, enough food must be available reliably over the long term, and that food must itself have a dependable source. In other words, seven hundred thousand years after the boundary, that land had already rebuilt a food chain capable of supporting fifty kilograms.

Work backward. The rebound in species numbers at a hundred thousand years means that more kinds of animals were able to secure a place. The arrival of twenty-five-kilogram animals at three hundred thousand years means that the food supply had reached a certain stability. The curve measures the largest animal, but the number it reports speaks for the entire system.

Put the two pictures in this section together, and we get a statement utterly unlike the familiar story: before the boundary there was no blank slate, and after it there were no scaled-down versions of today's animals. The interval in between had inhabitants of its own, and many of its branches have no living descendants—not every lineage ended there.

One last limitation, and only one. Resolution on the order of eighty thousand or a hundred thousand years comes from a handful of exceptionally good sites. It is not the normal standard of the terrestrial record as a whole. Move to another continent, and the same interval may be a blur. The curve above, then, is a curve on one rock face.

III When Did the Fan Open?

After that first million years of the Paleocene, a question arises almost by itself. When, exactly, did the lineages that eventually led to the modern groups of mammals split apart? Did the splitting begin only after the boundary, or had it already happened while dinosaurs were still around? This is the heart of the debate. Before answering, we need to establish two terms.

The first is “order.” An order is a rank in biological classification, smaller than a class and larger than a family. Three examples will make it clear. Carnivora includes cats, dogs, bears, and seals. Chiroptera consists of all bats. Proboscidea is the elephant order. Animals belonging to the same order generally share an ancestor tens of millions of years in the past.

The second is a picture. Imagine a fan slowly opening. Its handle is the last common ancestor of all living placental mammals. More than a dozen ribs branch out from the handle, each leading to one of today's orders. Each rib then goes on branching within itself.

There are three entirely different moments here, though ordinary language bundles them under one term. When the handle appeared is one question. When those dozen-plus ribs split from one another is a second.

When each rib began branching internally is a third question. Tens of millions of years may separate these moments, yet “the major branches of modern placentals formed in the dinosaur age” runs them together. Does it mean their common ancestor, divisions between orders, or diversification within orders? The answers need not agree. Mammals themselves originated long before the end-Cretaceous: this is not a debate about mammals first appearing after the dinosaurs died.

Two major studies in 2023 addressed different parts of the question. First, the genomic study. The 2023 study covered 241 placental species. Its 65-taxon locus-tree dating used 37 soft-bounded fossil calibrations; the concatenated 241-species analysis added another 23. The 37 anchors used below to illustrate the principle are not the total for the full analysis.

Those “calibration points” need a brief explanation. Gene sequences by themselves can tell us only how many differences have accumulated between two species, not how many years it took to accumulate them.

To convert differences into years, we need to find places on the tree where fossils allow us to say, “This branch must already have existed by this date.” Thirty-seven calibration points are thirty-seven such stakes in the ground. Here are the dates the study produced: the handle most likely appeared around 102 million years ago, deep in the Cretaceous, with the dinosaurs still present.

Several of the deepest splits date to between 94 and 96 million years ago. Most early splits within the ribs, however, fall between 65.3 and 53.6 million years ago: after the boundary.

The same study gave us both a “Cretaceous origin” and a “Paleocene unfolding.” There is no contradiction. The three moments each have their own date.

The second study used no molecules at all, only fossils. It modeled 22,381 fossil occurrence records from 539 mammalian families together, working backward to estimate when each family most likely originated.

Its result: placentals already existed in the Late Cretaceous. For at least seven families, even when the analysis was rerun with a different set of assumptions, the entire estimated interval for their origin still fell within the Cretaceous. For most modern orders, meanwhile, the origin intervals straddled the boundary or fell after it.

Two studies, one starting from genes and one from stones, point to the same shape: an old handle and a late opening of the ribs. What, then, is still in dispute? The middle part. On which side of the boundary did the ribs themselves separate?

The literature has several names for the alternatives. The best way to remember them is by asking which of the three moments crosses the boundary, rather than remembering who supports which model. The explosive model places all three moments near or after the boundary.

The long-fuse model puts the handle and a substantial share of the splits between ribs in the Cretaceous, with branching within the ribs after the boundary. The short-fuse model goes furthest, placing a great many splits within the ribs in the Cretaceous as well. An intermediate alternative is the soft-explosive model, which allows a few ribs to extend across the boundary but clusters the overwhelming majority near it.

The short-fuse model is currently the weakest, because no generally accepted Cretaceous fossils of modern orders have been found. The main contest is between the long-fuse and soft-explosive models. The fossil-only study itself explicitly notes that the older ends of its intervals support the long fuse, while the younger ends support the soft explosion. The same fossils cannot decide between them. One particular dispute shows the shape of the whole argument unusually clearly.

In 2011, researchers reported mammal teeth from a rock layer in Montana three hundred thousand years older than the boundary. The argument is not about whether the tooth is Cretaceous. Nobody doubts that.

The argument is about whether the group belongs inside or outside the fan's handle. Put the same tooth inside, and it proves that the handle had already appeared in the Cretaceous. Put it outside, and it is merely one of the long-extinct side branches below the handle. Not a word about the rock's age has changed, yet the conclusion shifts by an entire order of magnitude.

Finally, remember a discipline this essay will need repeatedly. Tiupampa in South America was long listed among the earliest Paleocene mammal assemblages. Another chronology, based on the directions of Earth's magnetic field recorded in the rocks together with the sequence of sedimentary layers, places it around fifty-nine million years ago, much later in the Paleocene.

These two dates cannot be averaged into sixty-two million years ago. They are two different interpretations of the chronology, not two measurements of the same quantity. When methods are the source of a disagreement, taking the midpoint is no compromise. It manufactures a third account that nobody proposed.

IV Growing Larger, and How Large

Before moving on, let us run through the new world's cast. Who eventually performed the kinds of roles the dinosaurs left behind?

Large terrestrial herbivores. The principal losses at the end of the Cretaceous were animals weighing several to several dozen tonnes. Several groups of mammals began growing larger during the Paleocene. In the Eocene that followed, roughly fifty-six to thirty-four million years ago, larger herbivores appeared. On different continents, different lineages eventually came to perform these roles.

Large terrestrial predators. Carnivorous mammal lineages, South America's large flightless predatory birds, crocodyliforms, giant snakes: the combinations differed completely from continent to continent. The next section is devoted to them.

Large marine predators. Sharks were there on both sides of the boundary and never left. Large bony fishes diversified again in the Paleogene. Whales, however, do not enter the record until about 53.5 million years ago.

The largest flying animals. Birds were present from beginning to end, and large seabirds appeared from the Paleogene onward.

There is one more line on the list, the easiest to skip and really the most important: the small and medium-sized tree-dwelling, ground-dwelling, burrowing, and omnivorous roles. Mammals had been performing these before the boundary. After it, surviving mammals, birds, and reptiles continued to do so. For this line, there is no way to write down “how long it stood empty,” because it was never empty.

With the cast accounted for, back to body size. Size is the most immediately visible change after the boundary, and here we have directly measurable data. Researchers have assembled the maximum body sizes of terrestrial mammals throughout the Cenozoic on each continent. What they track is the largest animal in each interval, not the average.

In the 2010 compilation and body-mass estimates used here, the largest land mammals near the boundary weighed about ten kilograms and later reached roughly seventeen tonnes: a rise of 3.2 orders of magnitude. Seventeen tonnes is an estimate within that dataset, not an immutable mammalian ceiling. Estimates for extinct giants have continued to be revised and disputed.

The timing differs in detail from continent to continent, but the pattern is a rapid early rise followed by a plateau once large body sizes are reached. How much is seventeen tonnes? More than twice the weight of today's largest African elephants.

We need to unpack that phrase, “a little over three orders of magnitude.” Going from ten kilograms to seventeen tonnes means growing 1,700-fold. Each time body mass doubles, the bones that support it, the muscles that move it, the heart that supplies it with blood, and the gut that processes its food all have to be reworked. These things do not scale at the same rate. The cross-sectional area of a leg bone fails to keep pace with the increase in body mass, which is why an elephant's leg and a mouse's leg are not the same shape at all.

Growing from ten kilograms to seventeen tonnes is not a matter of making one thing bigger. It means rewriting a whole bodily scheme again and again. And over sixty-six million years, that happened independently on different continents.

The details of timing vary from continent to continent, but the shape is the same: a steep early rise, then a leveling off at large body sizes, with no further climb.

Those intervening tens of millions of years produced some giants with no living counterparts. In Eocene and Oligocene Eurasia there were herbivorous mammals standing several meters at the shoulder. During its long isolation, South America independently produced large herbivore groups of its own. These were distantly related lineages, each making its own way from small bodies to large ones, and all eventually stopping at roughly the same order of magnitude.

So “mammals later grew very large” is entirely true. But the comparison that comes next determines the course of the second half of this essay.

Reliable estimates put the largest sauropod dinosaurs, those giants with long necks and long tails, at several dozen tonnes. Discussions of their upper size limit commonly range from sixty to eighty tonnes, or higher still.

Credible estimates for land mammals still fall below the roughly sixty-to-eighty-tonne scale of the largest sauropods. Using seventeen and seventy tonnes gives a difference of more than fourfold, but the ratio changes with the estimates. The robust point is the remaining gap between known extremes, not a law locking mammals below seventeen tonnes.

Now return to the curve itself. How fast did it rise? The question sounds perfectly natural, yet until a threshold has been specified, it has no answer.

Put the threshold for “large” in different places, and you get entirely different times. If it means “beyond the typical upper size limit of Mesozoic mammals,” that had already happened within a few hundred thousand years of the boundary.

If it means fifty kilograms, the local figure is about seven hundred thousand years. If it means a tonne, the wait is longer. If it means the several dozen tonnes of a sauropod, terrestrial mammals never crossed that threshold anywhere in the Cenozoic. In the oceans, the same problem becomes still more extreme. Here the threshold means the largest sauropods’ roughly sixty-to-eighty-tonne scale, not every estimate in the tens of tonnes.

Ask, “Were there large apex predators in the ocean?” and the answer is that the role was never vacant. Sharks were present on both sides of the boundary; the next section will address them directly. Ask instead, “Were there wholly aquatic, lung-breathing tetrapod predators ten or so meters long?” and we must wait for whales to evolve full aquatic life, more than twenty million years later.

The same sea, two ways of asking: one answer is close to zero, the other more than twenty million years. Before asking “How long did recovery take?” we must say what counts as recovery. Fifty kilograms, five hundred kilograms, one tonne, ten tonnes, seventy tonnes: these are five different questions.

And their answers are, respectively, hundreds of thousands of years, millions of years, longer still, tens of millions of years, and “never.” Roll them into “How long did the Cenozoic take to recover?” and you stir five different answers into an average that does not exist.

V Not Only Mammals

This next section is the most interesting part of the essay, because it dismantles an equation most people do not realize they are using: the places vacated by dinosaurs equal the places later occupied by mammals. In South America, birds filled the compartment marked “large terrestrial predator.”

Not flying birds. Birds that stood taller than a person, with pillar-thick legs, disproportionately huge heads, and a hook at the end of the beak. They ran fast, could not fly, and served as large predators in South America for tens of millions of years.

Giant flightless predatory birds are recorded mainly in South America. The oldest reliable records come from Argentina in the middle Eocene, more than forty million years ago. Later, one branch took advantage of the connection between the two continents to enter North America. It is a rare case of a southern group moving north and establishing itself. Section Seven will return to the statistical implications of that fact.

They spanned an enormous size range, from a few dozen kilograms to several hundred.

And they were not “miniature tyrannosaurs.” Researchers analyzed the stresses on the skull of an individual weighing about forty kilograms. The result: its skull could withstand forces acting up and down, but was poorly suited to resisting powerful side-to-side twisting.

The inferred killing method was therefore a rapid succession of downward strikes, rather than clamping onto prey and wrenching at it for a long time. Give the same ecological role a different body, and even the movements change.

A word about size while we are here. Within a single family, the smallest members weighed a few dozen kilograms and the largest several hundred, a difference of nearly tenfold. One study estimating running ability from the mechanics of the lower leg bones covered a mass range of forty to 350 kilograms. Two kinds of bird with a tenfold difference in body mass are unlikely to be doing the same thing in a food web.

South America at the time also had carnivorous relatives of marsupials and large crocodyliforms. Crocodyliform bite marks have even been found on the remains of large predatory birds. At least by the water, those birds were somebody else's food too. “Terror birds monopolized South America's apex-predator roles” therefore needs to be downgraded to a hypothesis. The tropics have something more startling still.

A fossil site in Colombia dating to roughly fifty-eight to sixty million years ago yielded a giant boa relative, estimated at about thirteen meters long and 1.1 tonnes. What does thirteen meters look like? Picture a bus broadside and you have roughly the length. In other words, only six to eight million years after the boundary, tropical wetlands already held a tonne-scale reptilian predator.

Its way of life was probably much like that of a large anaconda today: semiaquatic, lying in ambush. The dimensions, though, were on another scale entirely. Even today's largest anacondas reach only five or six meters.

Crocodyliforms have a longer story. Throughout the Paleogene and Neogene in South America, Africa, and Asia, they long performed the role of large predators at the meeting of land and water. More than once, they also evolved forms adapted to running on land. At the regional scale, “mammals alone took over the large terrestrial predator roles after the non-avian dinosaurs disappeared” has never been true.

In the oceans, meanwhile, the easiest mistake about “vacancies” concerns the survivors. Sharks did not abandon large-predator roles at the boundary. Analyses standardized for sampling suggest that neoselachians may have been affected much less severely than the traditional account implies. They continued to radiate through the Paleogene and attained high diversity in the Eocene. The large marine reptiles disappeared, but the ocean's apex-predator tier was never empty.

Stories are especially apt to omit this, because storytellers habitually fix their eyes on who disappeared. Sharks were present throughout the entire event. They needed no successors, and nobody could take over from them.

The change in bony fishes is clearer still. The relative abundance of shark skin denticles and bony-fish teeth in marine sediments shows that bony fishes rapidly gained relative importance in the open ocean after the boundary.

Close relatives of modern-style tunas, billfishes, barracudas, and other large, fast predatory fishes began to appear in the early Paleogene. Some researchers have called this turning point a new age of fishes.

That is no exaggeration. The open oceans of the Cretaceous held several lineages of large pelagic predatory bony fishes, which disappeared at the boundary. What followed was not the same cast returning, but a new set of ecological types that would lead to the fishes of today's oceans. What happened at sea was a change of cast, not the filling of vacancies.

Whales came much later. Their earliest reliable record dates to about 53.5 million years ago, twelve and a half million years after the boundary. Fully aquatic whales came later still. How long is twelve and a half million years? Four or five times the entire history of the genus Homo.

So whales did not arrive by driving the sharks out. They entered an ocean already home to large fish predators and added a new set of capabilities: breathing air with lungs, growing immense, diving deep, finding prey by sound.

One striking change was baleen filtration. Baleen whales still seek out and follow prey concentrations, but their baleen can strain large quantities of small prey from the water. Giant filter feeding was not their invention, however: Mesozoic seas already held huge suspension-feeding bony fishes, including pachycormids that persisted into the Cretaceous. Treating marine reptiles as the sea’s entire cast erases those predecessors.

Whales therefore did more than refill the “large marine animal” slot: they combined nursing, breathing, locomotion, and filtration into new ways of living. A new combination need not mean the ocean had never seen a similar function before.

VI Those That Never Returned

Now ask the question the other way around. Were there kinds of things that never appeared again afterward? There are three fairly strong candidates, and the first is the clearest.

Reliable estimates give the largest pterosaurs a wingspan of about ten to eleven meters, with body masses commonly on the order of two hundred kilograms. What does a ten- or eleven-meter wingspan mean? With both wings outstretched, the span was roughly that of a small airplane; standing on the ground, the animal was about as high at the shoulder as a giraffe. More important is how it got into the air. Pterosaurs could launch themselves with all four limbs. Their membrane wings and the way forces passed through their forelimbs were quite unlike a bird's.

The largest known flying birds were distinctly smaller. A conservative wingspan estimate for a giant pseudotoothed bird is about 6.4 meters. The giant flying bird from South America weighed about seventy kilograms, with a wingspan of six to seven meters. A flying animal with a ten-meter wingspan, weighing two hundred kilograms and still able to launch itself from the ground, never appeared again in the Cenozoic.

A glance at the mechanics shows why the difference matters. A bird gets airborne by pushing off the ground with its hindlegs and then relying on its wings. A pterosaur's forelimbs were both wings and front legs. It could use all four limbs together to catapult itself into the air.

One jumps with two legs, the other with four. At the upper extreme of body size, that difference determines whether takeoff is possible at all. Even when both are gliding, therefore, the two kinds of animal have different performance envelopes.

The second candidate is the large marine reptiles. Whales did eventually reenter the space defined by “large, marine, lung-breathing.” Add particular swimming styles, tooth arrangements, reproductive modes, and body plans, however, and whales and Cretaceous marine reptiles overlap without coinciding. Baleen whales occupy a wholly different place again, making their living by filtering vast quantities of small planktonic organisms.

Here we must also admit a fact that points in the opposite direction. The sea's top-predator tier was never vacant, yet the particular combination embodied in those Cretaceous marine reptiles, ten or so meters long, really did never return. They were neither fish like sharks, which spend their entire lives in water, nor mammals like whales, which entered the sea from land. They belonged to another route, and that route broke off.

The third candidate is the extremely gigantic terrestrial herbivore. The Cenozoic had seventeen-tonne giants, but the sixty- to eighty-tonne range never reappeared.

Supporting seventy tonnes takes more than just being big. Sauropods had an extraordinary combination of features. An immensely long neck let one sweep a broad stretch of canopy without moving from its spot. Hollow bones associated with air sacs reduced its own weight. And it did not chew: food went down whole and fermented slowly in an enormous digestive tract.

Only the conjunction of these features made that size range possible. They never came together again in the Cenozoic. At this point, a sharp question presents itself: what entitles us to say “never returned”?

After all, the pterosaurs are gone, but there are large seabirds. The sauropods are gone, but there are giant mammals. What exactly do we mean when we say that a place is “empty”? There is an answer, and it has three requirements, each harder to meet than the last.

First, we must specify a set of measurable dimensions rather than compare the names of groups: body mass, diet, the mechanics of teeth or beaks, locomotion, the substrate on which an animal operates, feeding height, swimming or flying performance, thermal physiology. Only if the animals of two ages genuinely overlap on those coordinates can we begin to speak of “the same place.”

The second requirement is harder and often skipped: the environmental coordinates themselves must be comparable. Two ecosystems separated by tens of millions of years may differ in temperature, vegetation structure, the spectrum of prey sizes, and the arrangement of continents. If the Cretaceous combination of resources no longer exists, we cannot say “that place is still empty.” It may no longer have the same domain of definition.

The third is to distinguish “the closest one” from “the one that genuinely coincides.” In morphological space, a later group can be the nearest neighbor of a Cretaceous group and still be very far away.

These requirements look exacting. What they actually do is keep a statement from becoming empty. If we say “a place has not been filled” without first specifying the dimensions that make it “the same place,” the statement could apply to any pair of different animals. It would therefore say nothing at all.

Revisit the three candidates with these requirements in mind, and our wording has to change. What really never returned was not “the entire pterosaur-style ecological niche,” but the combination of maximum body size, a particular way of taking off, and a particular way of feeding on land. Nor was it “all the effects of giant herbivores,” but the size range of individual animals weighing several dozen tonnes. Several herbivores of five to fifteen tonnes could together exert just as powerful a pressure through plant consumption, seed dispersal, and trampling.

Today's elephants do a smaller version of the same thing. They knock trees over, strip bark, trample paths, and keep ground that would otherwise turn to forest as open grassland. An elephant's main effect on a forest is not how much it eats, but how it changes what else can grow. Ecology has a name for animals like these: ecosystem engineers.

A seventy-tonne sauropod walking through a forest probably did the same kind of thing, only on a much larger scale. Yet ten seven-tonne animals working in aggregate might well be able to do it too.

What one seventy-tonne animal does need not be an indivisible whole.

VII If We Played It Again

By now, one large question has become unavoidable. If we rewound the tape to sixty-six million years ago and played it again, would we still end up with today's world?

That way of putting the question has a source. In a 1989 book, Gould proposed the thought experiment he called “replaying life's tape.” Notice that his question was not “Is evolution random?” It was: how sensitive is the outcome of history to the path already taken?

The best-known voice on the other side is Conway Morris, whose reply is not “History does not matter.” His argument is that physical, functional, and developmental constraints make the set of workable solutions much smaller than the set of logically possible ones. Different histories will therefore repeatedly arrive in similar functional regions. Eyes, streamlined bodies, flight structures, and digging forelimbs appear again and again in unrelated lineages.

Image-forming eyes are the most frequently cited example. They arose independently many times across the animal kingdom. Vertebrates and octopuses each produced camera-type eyes with remarkably similar structures, even though their retinas are installed the opposite way around. Flight, too, arose independently at least four times: in insects, pterosaurs, birds, and bats. Some solutions really are found repeatedly.

The thought experiment was later made into an actual experiment. One long-term evolution experiment began with twelve initially nearly identical bacterial populations and let them reproduce, generation after generation, under the same conditions. Only one evolved the ability to make extensive use of citrate in the presence of oxygen, and it did so only after more than thirty thousand generations.

The next step mattered even more. The researchers took frozen ancestors and intermediate generations out of storage and actually replayed the tape. When they restarted from later clones that had already accumulated certain prerequisite genetic backgrounds, the innovation was more likely to appear again. When they replayed from earlier ancestors, it appeared very rarely.

The result supports two levels at once: overall performance can be repeatable, while particular key innovations and genetic routes can depend heavily on history. Gould and his opponents each got one level right. They were not talking about the same level. Nature supplies near-replays too, and the classic pair is the thylacine and the wolf.

Researchers compared 101 specimens using 128 three-dimensional landmarks. Adult thylacine and gray-wolf skulls are strikingly similar, and even their developmental trajectories run strongly in parallel, with correlation coefficients reaching about 0.91.

The differences are just as systematic. The basicranium, auditory bullae, and details of the teeth are not copies. Skeletal modules with different embryonic origins did not converge to the same degree.

Further functional measurements even suggest that the thylacine's hunting style may not have been closest to that of wolves specializing in large prey. On several measures, it is closer to canids that hunt smaller prey. In a photograph it looks like a wolf. Measure it, and the resemblance becomes less convincing.

Convergence, then, does not mean that a whole animal turns into its counterpart on another continent. It means that certain structures performing similar functions draw closer to different degrees. Some parts are almost copies. Others retain their separate inheritances in full.

The same pattern appears elsewhere. Comparisons of carnivorous mammal skulls show that similar diets and hunting methods do indeed push different lineages toward one another. But the different regions of a skull face different developmental and mechanical constraints, so convergence is uneven. Some regions converge closely; others retain strong distinctions between lineages.

Gliding mammals offer another version of the same pattern. Different lineages independently evolved gliding membranes along the sides of their bodies, producing a powerful convergence in outward form. Quantitative skeletal morphospace, however, does not squeeze them into a single cluster. The pelvis, wrists, ankles, and proportions of the long bones all retain the marks of their respective ancestors. Functional results converge more readily than internal structures.

Island lizards provide a larger set of near-replays. The same kind of lizard radiated independently on different islands and repeatedly produced similar types: some living high in the canopy, some on tree trunks, some on the ground. Corresponding leg lengths, body sizes, and ways of moving also appeared again and again. This is one of the strongest quantified examples of convergence in nature.

At the level of protein-coding sequences, however, those same animals do not show convergence to match their outward resemblance. Looking alike does not mean being alike underneath. Finally comes a revision to the question of “who won,” and it is the point in this essay most worth remembering.

South America had been isolated for a very long time. It developed an entire fauna found nowhere else: its own ungulates, its own large sloths and armadillo relatives, its own carnivorous marsupials, and the great birds we met earlier. Then, around three million years ago, the Isthmus of Panama connected North and South America, and interchange in both directions intensified. This marks the major expansion of interchange, not its first crossing: some animals had crossed earlier.

Today's outcome looks clear enough. A high proportion of South America's living fauna has northern origins. Relatively few southern groups successfully entered North America and left a highly diverse legacy there. This used to be explained as the result of northern animals being better competitors.

In 2020, researchers modeled roughly twenty thousand fossil occurrence records, estimating dispersal, origination, and extinction rates separately. They found no systematic difference between northbound and southbound migrants in dispersal rates or rates of origination after arrival sufficient to explain the asymmetry. The real cause lay elsewhere. Native South American mammals suffered higher extinction rates in the Pliocene, shrinking the southern source pool from which northward migrants could come in the first place.

Another analysis that year used stable isotopes from more than four hundred fossils. Isotopes can reveal diet during the formation of the sampled tooth, not a complete lifetime menu. Animals that eat different plants leave different carbon-isotope ratios in their teeth.

The results showed that immigrant groups often had broad diets. Native and newcomer groups also partitioned resources clearly, rather than replacing one another neatly, one for one. They looked more like animals each finding an opening than animals pushing one another out. Who won in the end cannot tell us who was superior at the start.

VIII The Remainder in a Moving Coordinate System

We can now answer the opening question. “Not inheritance” holds up, and holds up firmly. No successor copied its predecessor's combination of traits. The maximum size range never returned. The extreme size range in flight never returned. Predatory functions were divided among mammals, birds, crocodyliforms, and snakes. And the ocean's survivors were there from beginning to end.

“Occupation” needs qualifying. The reason is this essay's real contribution. The word “place” presupposes something: a compartment that exists before its occupant and stays fixed while it is being occupied. Three lines of evidence fail to support that presupposition.

We saw the first in Section Two. The Paleocene had distinctive bodily schemes that never appeared again. This was not a matter of filling old compartments one by one. New forms grew into being and later disappeared.

The second came in Section Six: the environmental coordinates themselves change. If the Cretaceous combination of resources is gone, the place no longer has the same domain of definition. Calling it “still empty” describes the present with a coordinate system that no longer exists. The third line of evidence is the most important, and it has a name: niche construction.

Organisms alter vegetation structure, soil, nutrient cycles, fire regimes, hydrology, shelters, and their relationships with one another. What their descendants inherit is therefore not an untouched external environment.

Take a visible example. A beaver builds a dam, and a stream becomes a wetland. Every plant and animal that lives there afterward encounters that wetland, not the original stream. The beaver did not move into a ready-made place. It made the place itself, and not only its own place, but places for others too.

On a larger scale, the structure of terrestrial vegetation determines how many feeding heights are available, while soil depth determines whether burrowing is possible. Both have been shaped over long periods by organisms themselves. If the occupants in a radiation alter the abundance of resources and the selection pressures facing other species, then the space of opportunities moves as it is occupied.

Entering a space of opportunities and changing it through one's activities can happen at the same time. Some of these difficulties arise from the word “niche” itself. It never had just one definition: three traditions came to share the same word. In 1917, one researcher understood it in terms of the physical environments in which a species could exist.

In 1927, another emphasized a species' functional role in a community, especially its relations to food and enemies. In 1957, a third formalized the niche as a multidimensional space within which a species could maintain its population. This version is strongly centered on the species.

“Vacancy” thus admits two entirely different questions. “There is a role in the environment that nobody is performing yet” can be defined before an occupant appears. But if a place means “the set of conditions under which a particular species can maintain a population,” it is hard to say that a place belonging to that species already exists when the species does not.

An “empty ecological niche” is therefore not a settled technical term. It is the subject of a real conceptual dispute. One practical compromise replaces “a blank compartment already drawn” with ecological opportunity: the environment contains resources that can be used, and the capacities of existing organisms do not fully match those opportunities. This recognizes that an environment can offer opportunities before new occupants appear, without assuming that the complete place of a future animal already exists in advance.

The entry this essay adds to the table of remainders is therefore the thirteenth: the remainder lies in a coordinate system that moves as it is occupied.

It must be recorded separately from Episode Eight's “remainder in the coordinate axes.” In that case, the remainder arose from a choice: which axes the researcher selected. Change the axes, and you change the remainder. Here it arises from a process: the objects being measured rewrite the axes themselves. It is there whoever does the measuring.

A rule follows. Anyone wishing to say “a place has not been filled” must first supply three things: a list of functional axes; an argument that the environmental coordinates are comparable; and a distinction between “nearest neighbor” and “exact match.”

Without any one of the three, the statement is merely a category imposed after the fact, not a discovery. Alongside this stands another rule: “Cenozoic radiation” is best understood in the plural.

Branching within the fan's ribs was concentrated around sixty-five to fifty-four million years ago. The maximum size of terrestrial mammals increased markedly within the first million years, but reaching its worldwide maximum range took tens of millions of years. Whales entered the record only 53.5 million years ago.

Grasses make the point most clearly. The grass family already existed in the Cretaceous, yet truly open grasslands came much later, and not at the same time on every continent. In North America, open grassy habitats only began to spread gradually after thirty-four million years ago. The great expansion of the more drought- and heat-tolerant grasses of low latitudes was concentrated mainly between eight and five million years ago.

Studies using multiple indicators in East Africa push the local abundance of these grasses back earlier still. An ancient lineage does not imply the prior existence of an environment dominated by it.

The timing is uneven among birds as well. Some studies emphasize rapid morphological and phylogenetic differentiation in the early Paleocene. Others support a longer radiation extending continuously from the Late Cretaceous across the boundary.

Insects call for greater caution. Many major lineages of plant-eating and pollinating insects already existed in the Cretaceous. What is clearer after the boundary is the regional loss, recovery, and reorganization of relationships between plants and insects. Recovery in plant species richness and recovery in insect feeding types need not keep pace with one another. They are two curves, not one.

The exchange of animals between the American continents peaked later still, in the Pliocene and Pleistocene. These times share no single point of departure. Nothing in the evidence supports one globally synchronized “moment of succession,” accomplished once and for all.

Return to essay eighteen. The rocks preserve the history that happened; they cannot supply a second Earth as a direct control. That does not make counterfactual inference impossible. Bacterial experiments add another tool: restart from preserved ancestral samples and observe new evolutionary runs. That is replication, not literally rewinding the one past that occurred.

The price is scale: twelve populations, more than thirty thousand generations, one trait. We can restart a controlled experiment, not a whole chapter of Earth history. Models and evidence still constrain the larger question, but laboratory repeatability does not transfer intact. What we most want to ask happened on the scale we cannot directly restart.

There is no change of level to adjudicate in this essay. The unit of selection did not change across the boundary. What we have here is something else. When a construct claims that “a place is empty,” it is also claiming a coordinate system, and that coordinate system is being rewritten by the very process the construct describes.

Those that survived were not successors. They simply happened to be there, while the world continued changing shape beneath their feet.

Failing to cross did not mean being inherently inferior. That day changed which abilities helped: an old advantage might become a burden, an unremarkable trait a lifeline. Survival was not unrelated to traits; their usefulness depended on the environment. Crossing the boundary was no certificate of permanent superiority.

Nobody was waiting for a place, and no place was waiting for anybody. A “place” is what we, standing at the outcome and looking back, assign to the process.

We recognize today’s inhabitants, so it is easy to think the vacancy was waiting for them. Look back into the dust as it cleared. Nothing there knew that we would follow.