I A Fifteen-Year-Old Boy
In the spring of 1957, in Charnwood Forest near Leicester in central England, a fifteen-year-old boy went rock climbing with two friends. On the rock face of an abandoned quarry, he spotted a leaf-shaped impression. It was more than twenty centimeters long, with a central axis and regular branches down either side. It looked like a feather, or a fern frond.
The boy was Roger Mason. He made a rubbing and took it home. His father, a local minister, knew Trevor Ford, a geologist at the University of Leicester.
Ford went out to see it. The following year he formally described and named it Charnia masoni: the genus took its name from Charnwood, the species from Mason. The fossil is still on display in a Leicester museum. Mason went on to become a geologist himself, and later in life taught at the China University of Geosciences in Wuhan.
Mason was not the first person to see it. The year before, a girl of the same age, Tina Negus, had spotted a similar impression in the same forest.
Back at school, she asked her teacher about it. Those rocks were Precambrian, the teacher told her, and Precambrian rocks contained no fossils. She let the matter drop. It took half a century for her experience to enter the history of the fossil's discovery.
Twenty years earlier still, another boy in Leicester used to cycle into the countryside at weekends to look for fossils. His father was principal of University College Leicester. The boy grew up to make nature documentaries: David Attenborough.
Late in life, he said it had never occurred to him to look in Charnwood. Everyone knew the rocks there were Precambrian. There was nothing in them. We will come back to him in section four.
What “everyone knew” had a history. Darwin had worried about it in On the Origin of Species: the Cambrian fossils were already complex at their first appearance, and beneath them there was nothing.
By the middle of the twentieth century, it was textbook knowledge that rocks older than the Cambrian should contain no large, structurally complex organisms. A “large Precambrian fossil” landed squarely in that gap.
Similar objects had been reported several times before. Every time, they were turned away with one of two answers: the rocks had been dated incorrectly, or the object was not an organism at all.
In 1868, disc-shaped impressions were found in Newfoundland, now part of Canada. In 1872, Billings, a paleontologist with the Geological Survey of Canada, gave them a name. They came from rocks then thought to predate all complex life. Later scholars preferred to explain the discs as inorganic concretions.
Only when the type specimens were reexamined near the end of the twentieth century did that name regain its standing. Around the 1930s, German researchers described a group of large, soft-bodied fossils from South West Africa, now Namibia. There was no accepted framework for large Precambrian animals, so for years these specimens were interpreted as Cambrian.
The Australian discovery of 1946 came closest to breaking through. Sprigg, a twenty-seven-year-old geologist, was in the Ediacara Hills of South Australia's Flinders Ranges, checking whether several abandoned mines could be reopened.
The story goes that he sat down to eat his lunch, turned over a piece of sandstone, and saw disc-shaped impressions on it. His 1947 paper described them in its title as Early Cambrian jellyfish from the Flinders Ranges. After “Early Cambrian,” though, he inserted a question mark in parentheses. That question mark records a man believing two contradictory things at once: this looked too old, and being that old was impossible.
Charnwood was different. The age of those rocks had been established before the fossil was found. The evidence came from regional stratigraphic relationships and comparisons with the rocks above and below. Not a single fossil had been used.
The dating evidence and the fossil evidence were independent of each other. “Actually, those rocks are Cambrian” was no longer a way out. Only two options remained: accept that large, complex organisms had existed in the Precambrian, or prove that the impression was not an organism. The regularity of that leaf made the second route difficult too.
Newfoundland turned the problem from a single curiosity into an entire seafloor. In 1967, Misra, an Indian graduate student at Memorial University of Newfoundland, was mapping Precambrian strata on the south coast of the Avalon Peninsula when he saw hundreds or thousands of fossils on the coastal rock surfaces. That stretch of coast is called Mistaken Point. The name comes from old shipwrecks: in fog, captains would mistake it for another nearby headland, turn the helm, and run straight onto the rocks.
Many of those surfaces were covered in an instant by volcanic ash. As at Pompeii, whole communities were buried where they stood. What survives is a snapshot of the seafloor: every individual on a given surface was alive at the same time.
That makes it possible to measure their density, examine their arrangement, and compare the distribution of sizes. Mistaken Point became a World Heritage site in 2016. The cliffs along Russia's White Sea coast are another treasure trove; research there began in the 1970s.
Each spring, as the ice and snow thaw, another layer falls from the cliffs, exposing fresh fossil-bearing rock. This was once a shallow sea. Individual organisms are finely preserved, and their bodies can be linked on the same surface to the traces of their behavior.
The interval itself did not become an official unit of geological time until 2004. The reference section for its lower boundary lies in a creek valley in South Australia's Flinders Ranges. A bronze plaque set into the rock marks the line, at about 635 million years ago. For the first time in roughly 120 years, a new period had been added to the international geological timescale. It was named the Ediacaran.
Their age was settled. Their affinities were not. Within the same span of time stood a collection of utterly different bodies. This essay is about the repeated attempts to fit them into boxes, and why the boxes keep failing to hold them.
II What Was There
First, let us describe them, without rushing to decide where they belong.
One of the first major forms to be recognized was the frond-shaped body seen in Charnia. These are now collectively called rangeomorphs. The name comes from Rangea, a Namibian genus named in turn for Range, a German geologist who had surveyed the region.
Their defining feature is a body built by recursively repeating the same module. First-order branches grow from a central axis. Those branches consist of smaller second-, third-, and even fourth-order branches, each shaped like the whole.
Enlarge a little section and it still looks like the entire organism. A tree's twigs do not have the same shape as its trunk. In these fossils, every level repeats the same pattern.
Quantitative studies put their fractal dimensions at roughly 1.6 to 2.4, depending on the genus. Within a limited thickness, they packed in an enormous surface area. Living corals, sea fans, fungal mycelia, and plants can all grow branching forms, but none uses this particular construction, one level nested inside another, each repeating the whole.
They looked like plants, but did not live on sunlight. The communities at Mistaken Point inhabited deep water, hundreds of meters down and beyond the reach of light. That vast surface area was more likely used to absorb dissolved nutrients and oxygen from seawater.
They reproduced too. In 2015, researchers at Cambridge borrowed statistical methods used to analyze the distribution of trees in forests. They measured the positions of more than a thousand individuals on three rock surfaces in Newfoundland. One kind of spindle-shaped frond grew in clusters, each containing larger individuals ringed by smaller ones. The pattern resembled strawberry plants sending out new shoots on runners.
Frond-shaped organisms varied enormously in size. Some measured only centimeters. Some frond- and spindle-shaped individuals reached a meter or more, with the largest approaching two meters.
Among the flat, oval, transversely ribbed forms, Dickinsonia is the most conspicuous. Sprigg named it for Dickinson, then South Australia's director of mines. Its Chinese name still includes the word for “jellyfish,” a residue of an early misidentification.
It ranged from less than four millimeters to about 1.4 meters long. Thin and flat, with numerous riblike units arranged along its long axis, it resembled an oval air mattress. On a single rock surface, individuals of every size often lie scattered everywhere, from fingernail-sized to as large as a washbasin. This lets researchers reconstruct how it grew: new units were added one by one at one end, while older units widened as the whole body grew.
Whether the units on the two sides were exact mirror images or slightly offset has long been a central dispute. The difference sounds small, but it carries enormous weight. Strict mirror symmetry is a feature of the body segments found in annelid worms.
An arrangement in which a unit on the left faces the gap between two units on the right is almost unknown among living animals. So when we see transverse ribs, we must first ask what sort of ribs they are.
The characteristic threefold-symmetrical form is Tribrachidium. It is usually preserved as a circular low relief, from a few millimeters to five centimeters across. Three curved principal arms at its center sit roughly 120 degrees apart, all curling in the same direction, like a three-bladed pinwheel.
Fluid-dynamic simulations show that this body could disturb currents and make particles in the water settle in particular places. Living animals with bilateral, radial, or fivefold symmetry all have established places in the classification system; fivefold symmetry belongs to the phylum containing starfish and sea urchins. No living animal phylum has an adult body built on threefold symmetry.
Kimberella has clear bilateral symmetry and is associated with feeding traces. Its name commemorates Kimber, a fossil collector who died young, in 1964, during an expedition to central Australia. Oval to pear-shaped, it had a front and a back, a left and a right. Large individuals could exceed ten centimeters.
Repeatedly found alongside its body fossils are sets of paired scrape marks arranged in fans. Successive scrapes change their radius, like someone sitting in one place and raking the ground in front of them, turning a little with each stroke. What Kimberella scraped was the microbial mat covering the seafloor. Most importantly, the scratches are often directly connected to a body: we can see who did the scraping.
For the vast majority of trace fossils, we do not know who made them. Match a body to its traces and the information moves up a level. It tells us more than that something moved: it gives the direction of feeding, the repetition of the movement, and the way the working organ met the ground. The scraping organ itself, however, has not survived in the fossils.
Another form lived half buried in mud. Namibia's Ernietta had a three-dimensional, sack- or cup-shaped body. Many individuals are preserved with half the body inserted into the sediment. Reconstructions must show them standing in the mud.
These forms are often discussed alongside others with a quilted appearance, rows of tubes, or compartments, rather like a mattress stitched into squares. That outward resemblance is the entire reason for putting them together.
The tubular fossils are the most mixed assortment: clusters of slender soft tubes, centimeter-scale organic tubes, and Cloudina, with its cup-shaped tube segments fitted one inside another. Cloudina was already mineralized, one of the earliest known animals to grow a hard shell. Its name honors the American paleontologist Cloud.
Trace fossils grew more complex by degrees. The earlier beds already contain simple surface drag trails and traces of feeding beneath the mats. Between about 551 and 539 million years ago, more definite tracks of sustained movement and shallow burrows appear. Most Ediacaran traces are confined to the mat surface or the shallow zone immediately beneath it, from a few millimeters to a few centimeters deep. Organisms were not yet churning up whole sediment layers on a large scale.
Two numbers to finish with. Across the macroscopic fossil record, body size spans at least three orders of magnitude, from a few millimeters to nearly two meters. There is no typical size for an “Ediacaran organism.”
Nor is there a stable whole number for the genera and species. Specialist databases summarize the record as roughly two hundred described macroscopic species. Broader inventories of names published over the years yield about 247 genus names, of which the compilers consider roughly 164 valid. The lists also contain many synonyms and problematic fossils.
Naming bears human traces of its own. In 2018, a small disc from South Australia was named Obamus, after the recently retired American president. That same year, a tiny oval fossil the size of a raisin was named Attenborites, after the man who had not gone fossil hunting in Charnwood as a boy. How many members a group contains depends on how many names are judged valid, and that judgment is continually revised.
III The Boxes They Have Been Put In
The first approach was to look, one by one, for counterparts among living animal phyla.
In 1947, Sprigg described the South Australian discs straightforwardly as jellyfish. He sent his discovery to Nature, which rejected it. The following year, he presented it at the International Geological Congress in London. Hardly anyone took it seriously.
The turning point came with Glaessner, an Austrian-born researcher teaching at the University of Adelaide. In the late 1950s, he compared the Ediacara Hills fossils with similar material from around the world and confirmed that they predated the Cambrian. Sprigg himself later switched to oil exploration. He also bought a pastoral property at the northern end of the Flinders Ranges and turned it into a wildlife sanctuary.
For more than twenty years afterward, Glaessner and his colleagues followed a general scheme: discs and radial bodies were coelenterates such as jellyfish; fronds were sea pens; flat bodies with repeated units on either side were annelids or arthropods. The appeal was straightforward. If early members of modern phyla already existed in the Precambrian, the strange shapes could be explained as “familiar animals incompletely preserved.”
Dickinsonia was explicitly assigned to the annelids. In 1966, its transverse repeated units were interpreted as body segments; in 1972, it was reconstructed as a polychaete, like a ragworm. Yet the many specimens found later showed no convincing mouth, anus, gut, or parapodia, and no internal segmentation. The “segments” on the two sides might even be offset.
Charnia and its relatives were long placed among the sea pens. Both have a central axis with repeated “leaves” on either side. From a distance, they look much alike. Sea pens are cnidarians: each entire colony is made of many little polyps, each polyp with tentacles and a mouth. Today they stand in patches on muddy deep-sea bottoms. Some glow when touched.
Charnia's branches, however, are made of smaller self-similar modules. There is not a polyp to be found. From a distance they look alike. Up close, they are built in two entirely different ways.
One by one, these “ancestors of modern phyla” lost their places. In each case, some criterion had failed. Early studies relied above all on similarities of outline: discs looked like jellyfish, featherlike fronds like sea pens, transverse ribs like annelid segments.
Modern phylogenetics requires homologous characters that can be compared across groups. The second essay in this series discussed the distinction. Similarities can come from a common ancestor, or from separate lineages independently evolving similar forms. Only the first can be used to build a family tree. Many Ediacaran molds and casts preserve little beyond outlines, repeated units, and surface textures. Once “it looks like one” no longer counts, very few characters remain with which to diagnose a phylum.
The second approach started afresh. Seilacher, at the University of Tübingen in Germany, had spent his life studying trace fossils. Around 1989, he proposed that these organisms belonged to a wholly extinct major group separate from, and on a level with, the animal kingdom. It was later named Vendobionta.
He thought they shared an “air-mattress” architecture: rows of fluid-filled compartments, like a thin pad stitched together. The argument had a clear functional logic.
Many of the fossils were large but extremely thin. Repeated compartments could increase the ratio of surface area to volume, and many specimens showed no convincing mouth or digestive system. He therefore suggested that they might absorb dissolved organic matter across their entire surfaces, or obtain energy from symbiotic microbes. In this model, the enormous area, slight thickness, and repeated compartments all arose from the same physiological constraint.
The biggest difficulty was that this scheme could not hold everything either. Kimberella fed in a particular direction. Dickinsonia moved and left traces of feeding. Near the end of the period, there were tubular animals with hard shells. None fitted a uniform model of “no mouth, everything absorbed through the body surface.” Even among the fronds, three-dimensional reconstructions have found some branches open along their seams, rather than sealed like inflated bags.
The third proposal came from Retallack at the University of Oregon in the United States. From the 1990s onward, he argued that some characteristic Ediacaran fossils might be akin to lichens. In a 2013 Nature paper, he went further: some fossil-bearing beds in South Australia were ancient soils, he claimed, and the organisms had lived on land.
His evidence included cracks and concretions in the bedding surfaces, soil-like layering, and isotope patterns he interpreted as changes caused by soil formation. Some Dickinsonia specimens retained considerable relief after compaction, suggesting bodies tougher than soft aquatic carcasses.
The first objections came from sedimentology. These fossil beds contain ripple marks and storm-deposited sand layers. They are continuous with unmistakably marine strata, and some of the same organisms also occur in marine limestones.
The geochemical argument centers on whether soil formation is the only way to produce those chemical changes. Early diagenesis, seafloor microbial mats, and marine pore waters can also generate complex chemical differences. The model still has defenders, but it has not become the mainstream view.
A fourth proposal made them giant single-celled or multinucleate organisms. In 2003, researchers suggested that Ediacaran seafloors might have been dominated by microbial mats and giant protists. They linked some forms to a group of enormous foraminiferan organisms living in the deep sea today.
Such organisms really do inhabit modern deep-sea floors, with single cells growing more than ten centimeters across. This proposal needs no mouth, gut, or muscles. It can also accommodate enormous, multinucleate bodies that are thin and have a large surface area.
The difficulty is that molecular systematics places the living group on a branch that diverged relatively late within the foraminiferans. Linking it directly to fronds from 560 million years ago would push its history much farther back. In 2011, researchers rejected that simple correspondence, while leaving open the possibility that some compartmentalized fossils were large protists.
The fifth proposal was the plainest: some impressions were never a single organism at all. A 2007 study of several disc fossils suggested that some had formed when micrometer-scale filaments grew together in microbial mats.
The Newfoundland discs were first taken for inorganic concretions. Later work showed that many disc-shaped “genera” could instead be the holdfasts of different frond-shaped organisms, different states of preservation of the same organism, or structures formed by microbial mats. “One disc-shaped impression equals one disc-shaped organism” was among the first assumptions that taphonomy broke apart.
These proposals did not successively replace one another. Modern animal phyla, an independent major group, lichens, giant protists, microbial colonies: different proposals hold up in part for different genera. It is more like casting several nets over the same pile of things, each net catching a portion.
IV The Few We Have Placed
For a study published in 2018, Bobrovskiy, a doctoral student at the Australian National University, traveled to Russia's White Sea coast. The Dickinsonia he wanted were embedded in seaside cliffs tens of meters high. He flew in by helicopter, hung from ropes against the cliff face, and chiseled off the sandstone piece by piece. There was nothing out there but bears and mosquitoes.
He needed a particular kind of specimen. Over hundreds of millions of years, the White Sea rocks had never been deeply buried or intensely heated. Some Dickinsonia fossils still carried an exceedingly thin organic film.
There were no intact molecules in that film, only the carbon skeletons left after time had worn the molecules down. Essay five discussed cholesterol and its molecular relatives: common in eukaryotes, rare in bacteria. In animals, the predominant members of this family have twenty-seven carbon atoms. Algae and fungi mostly use related molecules with twenty-eight or twenty-nine.
Here was the result. In the film on Dickinsonia, the twenty-seven-carbon type accounted for about 93 percent. In the rock just a few millimeters above and below the fossil, it made up only about 10 percent.
The same piece of sandstone, a few millimeters apart, with utterly different molecular compositions. On that evidence, the team concluded that Dickinsonia was an animal. More than seventy years after it was named, there was finally a judgment that did not depend on its shape.
The judgment was “animal.” It did not specify a phylum. The team also made two checks for contamination from outside. The first was spatial: if the signal came from petroleum that had later seeped into the rock, the fossil and its background should look much alike. The compounds would not be enriched precisely along the fossil's outline. The second concerned maturity. The samples had undergone very little thermal alteration, unlike petroleum formed at high temperatures and subsequently transported into the rock.
Around 2025, researchers returned to those data. Colleagues had long raised two concerns. Could the unusual isomer ratios come from external petroleum? And the original study had not published the full absolute abundances and total organic carbon measurements, leaving some background corrections impossible to check independently.
The reassessment concluded that no contamination mechanism had yet been found that would naturally produce enrichment “only in Dickinsonia.” The signal still looked more likely to be original. It also acknowledged that the available data could not eliminate every possible form of contamination.
Cholesterol is not exclusive to animals. Some other eukaryotes can also synthesize small quantities. What carries real weight, then, is the overwhelmingly dominant proportion, the contrast between fossil and surrounding rock, and the agreement with other evidence.
Dickinsonia also left evidence of its own. On White Sea rock surfaces, Russia's Ivantsov found some individuals with a string of increasingly faint impressions behind them, each matching the shape of the body. The organism lay on the microbial mat, digested and absorbed what was beneath it through its underside, then moved a short distance and settled down again.
That string of impressions was its dinner table. In 2022, Bobrovskiy's team also analyzed molecules in Kimberella fossils. They concluded that it had a gut inside its body and had eaten algae and bacteria.
No sign of a gut could be found in Dickinsonia. It seemed instead to absorb food with its entire underside pressed against the ground. Two animals, two ways of eating, both written in molecules hundreds of millions of years old.
Kimberella followed a different route through classification. First called a jellyfish, it was later placed among the box jellyfish. In 1997, Russia's Fedonkin and America's Waggoner used a large collection of White Sea specimens to reconstruct a bottom-dwelling animal. It had a front and back, a left and right, a muscular foot underneath, and a tough covering over its back. They called it a “mollusc-like bilaterian.” Later still, its body and those paired scrape marks were repeatedly found together in Russia and South Australia.
The step to “bilaterian” rests on firm ground. “A close relative of molluscs” goes a step further. The fossils preserve no radula, mantle, or gills, and no typical shell. A cautious account today would say that it was very probably a bilaterian, and that a relationship to molluscs remains an important hypothesis.
China has an example too. In 2019, Chen Zhe and colleagues at the Nanjing Institute of Geology and Palaeontology, Chinese Academy of Sciences, discovered a segmented worm at Shibantan, in the Yangtze Three Gorges region at Yichang in Hubei. They named it Yilingia, after an old name for Yichang.
Its body was segmented and bilaterally symmetrical. Directly behind it ran the trail it had made itself. It was a bilaterian actively crawling about between 551 and 539 million years ago.
More precise placements have emerged in recent years. In 2022, Charnwood yielded another fossil: a slender stalk supporting a pair of cups, with traces of tentacles around their rims.
Researchers placed it in the cnidarian crown group and regarded it as one of the earliest known predators. Its genus name means “dawn lantern”; its species name honors Attenborough. The boy who had not come to Charnwood as a child finally had his name planted in the forest. In 2024, a study from South China also placed a sponge-like fossil within the sponge crown group.
These cases show that Ediacaran fossils can enter rigorous phylogenetic analysis when preservation is exceptional enough and enough diagnostic characters survive. Secure conclusions often stop at “animal,” “bilaterian,” or “animal with a skeleton.” Of the roughly two hundred described macroscopic species, only a small handful have independent molecular evidence, bodies matched to traces, or enough diagnostic characters.
V Boxes and Trees
To see where most of the others get stuck, we first need to look at how the boxes are made.
Kingdom, phylum, class, order, family, genus, species: the framework of these ranks comes from Sweden's Linnaeus. The first edition of his Systema Naturae, published in 1735, ran to just eleven folio pages. It used five ranks: kingdom, class, order, genus, and species. Phylum and family were added later; the rank of phylum was introduced by Germany's Haeckel in 1866.
Linnaeus wanted a catalogue of everything the Creator had made. Evolutionary theory did not yet exist, and species were considered fixed. Draw the boxes, then put things into them one by one.
Things could end up in the wrong box. In the first edition, whales went with the fish. Not until the tenth edition, in 1758, did Linnaeus move them into the mammals, on the grounds that they breathed with lungs, gave birth to live young, and fed them milk.
Some things burst the boxes as soon as they arrived. In 1799, a platypus skin was sent from Australia to London: a duck's bill, a beaver's tail, an otter's fur.
Examining it at the British Museum, Shaw suspected a practical joke, someone having stitched a duck's bill onto another animal. He is said to have taken scissors to it in search of the seam. Later, people learned that it laid eggs and also suckled its young. Today, it and the echidnas occupy their own distinct branch within the mammals.
Nesting lies at the heart of this system. Every group must sit wholly inside some higher group: an entire genus within a family, an entire family within an order. Half in and half out is not allowed. Today's animals are divided into thirty-odd phyla, each with a diagnostic checklist. Chordates must have a notochord; arthropods, jointed appendages and an exoskeleton; molluscs, a mantle.
After Darwin, the boxes were reinterpreted as a family tree. A formal group should correspond to “a common ancestor and all its descendants.” The question therefore changed from “What shall we call it?” to “Which surviving characters can establish the branch with which it shares an exclusive common ancestor?”
What this series calls a construct comes down, in the end, to drawing a line. What falls inside counts; what falls outside does not. Humans can draw the line, as with a set of criteria, or nature can draw it, as with a membrane. The construct in this essay is this set of boxes. Each box has its boundary drawn by a checklist of diagnostic characters.
Suppose you dig up a fossil. It has one feature that a living phylum subsequently retained, several more ancient features, and some peculiarities of its own that later disappeared altogether.
Does it belong to that phylum? Say yes, and it lacks the phylum's full set of characters. Say no, and it plainly shares something unique with that phylum.
Paleontologists developed a pair of concepts for situations like this. A crown group consists of the most recent common ancestor of every living member of a group, together with all that ancestor's descendants.
A stem group comprises the extinct side branches more closely related to that crown group than to any other living group, but which branched off before the crown group's common ancestor. Crown group and stem group together make up the total group.
Whales offer a good example. All living whales and dolphins, together with every descendant of their most recent common ancestor, form the whale crown group. Fifty million years ago, in what is now Pakistan, there lived a wolf-sized animal called Pakicetus.
In 1981, Gingerich of the University of Michigan in the United States published an account of its skull, found in rocks in northern Pakistan. It had four legs and walked along rivers, yet the bones in its ears already had a form unique to whales.
It was closer to whales than to any other living animal, but stood outside the whale crown group. It belonged to the whale stem group. Call it a whale, and it could walk. Say it was not a whale, and it had whale bones in its ears.
The term “stem group” was made for animals like this. When a phylum is defined as a total group, its stem members can perfectly well be included in that phylum.
The Ediacaran difficulty begins a step further back. A member deep in a stem group may carry just one or two characters that a phylum later retained. The rest may be ancestral features, along with peculiarities of its own that subsequently vanished altogether.
Compare it only with the full diagnostic checklist for modern phyla, and it looks like none of them. Extend a phylum indefinitely down the tree, and its traditional diagnosis becomes empty. A tree branches continuously; taxonomic ranks are separate boxes. Fit the first into the second and something will always get stuck between the boxes.
There is a deeper problem still. Many specimens do not preserve enough characters even to establish a position in a stem group. A branch, a disc, a row of transverse ribs can each evolve independently in different lineages.
Sandstone molds and casts are best at preserving overall outlines, relative dimensions, front and back, patterns of repetition, branching order, surface texture, means of attachment, and growth sequences reconstructed from individuals of different sizes. When a body and its trail lie on the same surface, we can add the direction of movement.
What they most often lack are precisely the internal characters most useful to phylogeny: germ layers, nerves, muscles, digestive tracts, body cavities, blood vessels, mouths and anuses, sensory organs. Whether a groove on the surface represents anatomy, a wrinkle from decay, or pressure from sand is often itself a matter of dispute.
Give the same fossil a different set of character codings and it can end up in an entirely different place. Some disagreements begin at the very step where a fossil is turned into data.
When something will not fit, there are two possible moves. Seilacher made the first: if the existing boxes cannot hold it, add a large new box on a level with the animal kingdom.
The modern approach using morphological groups makes the second move. Call them rangeomorphs, dickinsoniomorphs, or erniettomorphs for now. The names describe only the visible architecture of the bodies; they do not promise a monophyletic phylum or kingdom. First make clear how it was built. Leave what it was unanswered for the moment.
Adding another kingdom merely gives an unknown relationship a name. It supplies none of the missing information.
VI Three Assemblages
Ediacaran fossils are usually divided into three assemblages, named for their most representative localities: Avalon, White Sea, and Nama. The division goes back to a statistical study in 2003. Researchers grouped macroscopic fossil genera from around the world according to where they occurred and obtained three distinct clusters.
The Avalon assemblage is the oldest, roughly 575 to 560 million years old, represented by Mistaken Point in Newfoundland and Charnwood in England. Most communities lived in deep water beyond the reach of light. Stands of fractal fronds rose from the seabed, some anchored in the mud by a disc. The tallest stood more than a meter above the bottom, like a low forest without chlorophyll.
Very little crawled. Collecting fossils at Mistaken Point is forbidden. Researchers can only make molds on the rock surfaces, take photographs, and scan in three dimensions. Before visitors follow their guides onto the rocks, they must take off their shoes and walk in socks.
The White Sea assemblage dates to roughly 560 to 550 million years ago. Its classic localities are Russia's White Sea and South Australia. At Nilpena in South Australia, researchers have spent decades uncovering whole ancient seafloors, layer by layer. Across the rock surfaces lie Dickinsonia, Tribrachidium, Kimberella, and their traces, just as they were. Once exposed, the slabs are laid out again in their original positions in the field, like sections of a fitted floor. You can walk an ancient seafloor from end to end.
One tubular organism here grew in clusters. The researchers named it Funisia dorothea, after the lead researcher's mother. It is among the earliest known candidates for an animal that reproduced sexually. In 2026, these fossils formed a key part of the Flinders Ranges World Heritage nomination.
The Nama assemblage dates to roughly 550 to 539 million years ago and is centered in Namibia. Here the principal players changed: sack-shaped Ernietta, assorted soft tubes, and hard-shelled Cloudina.
Cloudina shells often bear regular little round holes, widely interpreted as holes drilled by predators. They are among the earliest known signs of predation. Cloudina is also abundant in rocks of the same age in southern Shaanxi, China.
The classic dickinsoniomorphs are noticeably scarcer here. Shibantan, where the Yilingia mentioned earlier was found, belongs to this interval too. Its fossils are preserved in limestone, an unusual setting in an Ediacaran record dominated by sandstone.
Ernietta often stood crowded together, half buried in sand. In 2019, researchers put their shapes into computer simulations of water flow. When individuals clustered together, those at the back could catch the recirculating flow stirred up by those in front. Food particles in the water were more likely to settle into their cup openings. They may have made their living by filtering water.
The differences between the three assemblages can have three sources.
The first is time. The statistical study found a broad sequence among the three clusters. Later high-precision dates from volcanic ash confirmed that typical Avalon communities were distinctly older. Volcanic ash is a natural clock. It contains small crystals called zircons, which incorporate a little uranium as they form. Uranium decays into lead at a fixed rate. Measure the ratio of uranium to lead and you can calculate when the ash fell, with an uncertainty of only a few hundred thousand years. At Mistaken Point, fossil beds alternate with ash layers, so each bed can be dated separately.
The second is ecology. From 2004 onward, researchers emphasized that Avalon-type communities favored deep, low-energy water, White Sea communities were mostly in shallow seas, and Nama communities were associated with another kind of shallow-water setting. Communities in different environments might have lived at the same time, each at its own depth.
The third is preservation. Deep-water volcanic ash, shallow-water sandstone molds and casts, and Nama's three-dimensional sand casts do not give every body shape the same chance of being preserved.
A statistical study in 2016 gave the answer some shape. Cluster and environmental analyses of a global database still produced three clear groupings. Differences in rock type and biases in burial and preservation could not account for all the distinctions. At the same time, the assemblages overlapped in age. Time, ecology, and preservation were all operating together.
Each can be tested: time through ash dates, ecology through sedimentary facies, preservation through taphonomy. The hard part is separating them. A genus may be missing from an assemblage because it had not yet appeared, because it did not live in that environment, or because it could not be preserved under those conditions. In a distribution table, all three appear as the same empty cell. The cell does not tell us why it is empty.
Names can skew the picture too. A study published in 2026 identified a community clearly belonging to the White Sea assemblage in the Mackenzie Mountains of Canada's Northwest Territories.
At the time, that region formed part of the ancient North American continent geologists call Laurentia. “White Sea assemblage” is useful as a statistical grouping. Its name comes from the place where it was first found, not from a claim that it lived only there.
Curves of early morphological diversity have the same problem. Researchers first code recognizable genera for visible features such as front and back, type of symmetry, levels of branching, and means of attachment. Multivariate methods then project them into a “morphospace,” whose expansion and contraction can be tracked through time. These curves are most sensitive to what goes in. Whether a genus counts as valid, whether different states of preservation of the same body have been split into several genera, whether a locality was sampled only on the undersides of beds or on their upper surfaces too: each decision changes the curve.
VII Death Masks and Silica
How, then, did these soft bodies survive in sandstone? How can coarse sand grains reproduce millimeter-scale details on a soft surface? Press your hand into a beach and the fine detail collapses at once.
The classic answer is the “death mask” model, systematically set out by Gehling in South Australia in 1999. The name comes from an old custom: after someone died, a mold of the face was made in plaster or wax to preserve their features. Both Napoleon and Beethoven left such masks behind.
While the organisms were alive, a continuous microbial mat covered the seafloor, the carpet discussed in the first essay. After death, they were quickly buried by sand. Bacteria in the mat broke down organic matter and released sulfides, which reacted with iron in the water to form a thin mineral film on the body's surface.
Before the soft tissues rotted away, this film fixed their millimeter-scale relief in place. Later, the sand above was cemented into a faithful cast. The mold was made by the mineral film growing rapidly over the corpse; the sand merely secured its shape.
In 2012, researchers tried it in a laboratory tank. They buried dead wax moth larvae in sand covered with a microbial mat. A few weeks later, a rim of mineral precipitate containing iron and sulfur really had formed around the bodies. In the control without a mat, fine details were much more poorly preserved.
It was a race between mineralization and decay. The conditions were narrow: enough iron and sulfur, sufficiently rapid burial, and a microbial mat. Lose any one, and the window closed. What survives, therefore, is often the body's outer surface. The soft tissue itself does not remain.
In 2016, researchers examining South Australian material found that many fossils lacked the pyrite predicted by this model. Instead, they found silica cement that had formed very early. They proposed another mechanism. Precambrian seawater was rich in silica; a thin siliceous film could rapidly precipitate on organic surfaces, and early silica cement could lock the sand grains in place.
The cement immediately adjoining the fossils had elemental ratios different from those of detrital quartz. Pore spaces had already been filled before compaction. Organisms such as diatoms and radiolarians, which remove large quantities of silica from seawater, had not yet appeared. The oceans contained far more silica than they do today.
Mistaken Point used another method again. There was no sand mold. Volcanic ash fell directly onto the living community, pressing the fronds down and leaving their impressions where they lay. On many rock surfaces, hundreds or thousands of fronds point in the same direction, the direction of the bottom current at the moment of burial. A standing forest was flattened overnight by the same rush of water, and stayed that way for more than five hundred million years.
Ediacaran-style preservation is a family of mechanisms, with different recipes in different places.
Microbial mats performed at least three jobs. They bound and stabilized the seafloor, preventing currents from immediately erasing fresh, shallow impressions. They created steep chemical gradients in microscopic environments, encouraging mineralization.
They also provided a tough organic interface that kept the sand above from mixing directly with the mud below. The mat was both the substrate on which many organisms lived and part of the machinery that preserved them.
These mats later became less common, most directly because more and more animals began grazing, burrowing, and turning over the surface sediment. Once a continuous, level mat was repeatedly pierced by burrows, it could no longer sustain that broad, sealed interface. The conditions that made these fossils were dismantled by the organisms that came later.
The window did not open and close all at once. Sandstone molds and casts of this kind are especially abundant after about 575 million years ago, but small numbers of similar examples occur in older and younger rocks too.
There are at least three concurrent explanations for the window's closure. Persistent bioturbation weakened the microbial mats. The ocean's silica cycle changed, so early silica cement no longer formed as readily.
The organisms themselves really did become extinct, leaving nothing to preserve. All three explanations can be true at once, but they make different predictions. The last requires the organisms actually to have died out. The first two do not.
This mode of preservation also selects strongly among bodies. The best candidates for faithful sand casts were bottom-dwellers at least a centimeter across, lying against the seabed, with enough area and toughness: broad, flat bodies, resilient outer coverings, raised textures. Tiny or extremely thin bodies, bodies that fell apart at a touch, organisms floating in the water without touching the bottom, and internal organs had much poorer chances.
It could also “edit” a three-dimensional organism into a two-dimensional exterior, preserving repeated surface textures in exquisite detail while deleting all the internal tissues. Parts of different firmness could be flattened, stretched, or wrinkled.
A “segment boundary” on the surface might therefore be an anatomical structure, or it might be a fold, an effect of decay, or a feature exaggerated by pressure from sand grains during molding. This is where historical arguments of the “transverse ribs must be annelid segments” kind went wrong.
The Ediacaran communities we see are a filtered list. The 2016 statistical study also reminds us that real biological differences remain behind that list. Known effects of rock type and preservation cannot fully explain the differences among the three major assemblages.
VIII The Box Around Them All
By the beginning of the Cambrian, Ediacaran-style soft-bodied fossils had almost vanished from the record. The Cambrian's lower boundary is also defined in Newfoundland, on a coast roughly two hundred kilometers west of Mistaken Point. Its marker is the first appearance of a particular burrow fossil: some worm dug through mud, turning as it went and leaving a series of branching tunnels. The Ediacaran begins with a bronze plaque and ends with a string of worm burrows. Where did its organisms go? There are at least three answers.
The first says that they really became extinct, driven out by environmental change. In 2022, researchers used a global database to estimate that roughly 80 percent of the units counted did not persist from the White Sea assemblage into the Nama assemblage.
The losses were selective. Survivors were mostly forms with greater relative surface area. The thinner and flatter they were, the better they survived; thicker bodies disappeared. The researchers interpreted this as an episode of oxygen depletion. Greater surface area makes gas exchange easier.
That 80 percent comes from the database of macroscopic fossils that could be assembled. The difficulty is that “oxygen depletion occurred” and “oxygen depletion killed these particular groups” are two separate inferences. Late Ediacaran records from different basins are not synchronized.
The second says that newly arriving animals replaced them. A burrowing animal can change oxygen levels in the sediment, pore-water chemistry, and the continuity of a microbial mat. It can destroy the habitat of organisms that depend on a stable mat surface without eating them at all.
The drill holes in Cloudina shells also show that predators had arrived. The problem with this explanation is scale. Burrows near the end of the period were still generally shallow. In many places, intact microbial mats and large soft-bodied organisms continued to coexist for a long time.
The third says that they were still there, but had become invisible. This is the “Cheshire Cat” model. The Cheshire Cat, in Alice's Adventures in Wonderland, slowly disappears until only its grin is left hanging in the air.
Its author, Carroll, taught mathematics at Oxford. The Red Queen discussed in essay six comes from his sequel. Sandstone molds and casts depend on stable microbial mats, particular minerals forming early, and very little disturbance. Once those conditions were gone, soft-bodied organisms of the same kinds could still be alive and yet become invisible in the record.
This model makes a testable prediction. In younger rocks that happen to reproduce the special conditions of preservation, Ediacaran-type forms should occasionally turn up. And they do.
In 1998, researchers reported characteristic frond-shaped fossils in sediments securely dated to the Cambrian. In 2000, similar fossils were reported from the Early Cambrian of California. Canada's 508-million-year-old Burgess Shale also contains a frond-shaped fossil that some researchers consider related to the Avalon fronds.
All three answers can be partly true at once, each accounting for different groups. In the Cambrian that followed, animals suddenly acquired shells, eyes, and legs. That is the next essay's story. These Ediacaran bodies stood at the threshold of all that bustle.
Finally, consider the largest box. The name “Ediacaran biota” has been in use for decades. In 2014, a scholar wrote an article titled “There Is No Such Thing as the ‘Ediacara Biota.’”
He pointed out that the term sometimes refers to an interval of time, sometimes to a collection of large fossils visible to the naked eye, and sometimes specifically to the strange soft-bodied organisms preserved as sandstone molds and casts. Algae, microbes, and skeleton-bearing animals from the same period are often excluded. The reason is neither ancestry nor morphology. They simply do not look like the things in those sandstone molds and casts.
The name draws a circle around things preserved at that time, in that way. One condition belongs to time; the other belongs to us. Asking “Which kingdom does the Ediacaran biota belong to?” may therefore mean asking the question of a collection. But a collection defined by preservation need not form a single evolutionary clade.
Return to this essay's construct, the boxes handed down from Linnaeus. Every box demands a diagnostic checklist, written around the members alive today. What a construct cannot accommodate and cannot make disappear, this series calls the remainder. The remainder here is this whole collection of organisms: roughly two hundred described macroscopic species, only a small fraction securely placed, and most of those no further than “animal” or “bilaterian.”
In the previous essay, the remainder lay in the coordinate axes. Those dozens of origins of multicellularity were fully visible, but a ruler without a natural unit had put them toward the back. Here, the remainder lies in the shape of the boxes.
These organisms are fully visible. Museum drawers hold them, thousands upon thousands of specimens. Their forms can be quantified, their growth reconstructed. Some even left traces of what they did.
What they lack is the set of characters that each of those boxes demands. To say “we have not found the full set yet” makes it sound as though the characters ought to have been there and simply failed to survive. The truth may be simpler. These organisms never had the whole set, because the set came together only later. The boxes were made to fit the traits once they had all come together. These organisms lived before that.
The criteria are still sifting. Genus names are still being merged and split. Some have already proposed abandoning the encompassing box itself.