How we know what dinosaurs were really like

There's a limestone slope in central Portugal, once a quarry floor, that's covered in holes. From ground level they read as ordinary weathering. Climb the viewing platform and they resolve into lines — pairs of enormous round prints, one behind the next, running straight across the rock for well over a hundred metres.

Something walked there. It was heavy, it was going somewhere, and it did it in the Middle Jurassic.

No human has ever seen a living non-bird dinosaur and no human ever will. The last of them died about 66 million years ago, which is far enough back that the entire history of our own genus fits into the gap several dozen times over. And yet palaeontologists will tell you what these animals weighed, roughly how fast they moved, what they ate, that many of them had feathers, and in a small number of cases what colour they were.

The obvious question is how much of that is evidence and how much is confident guessing. It's a fair question. The answer is more interesting than either.

What a fossil is, and why there are so few

Start with the raw material, because everything downstream depends on how it got there.

A fossil isn't a preserved bone. In most cases it's a mineral copy. An animal dies, and if it's very lucky it gets buried fast — a flood, a mudslide, a collapsing dune — before scavengers, bacteria and weather take it apart. Groundwater then seeps through the buried skeleton, and over a long time minerals precipitate into the pore spaces and, in places, replace the original tissue outright. What you dig up is rock in the shape of a bone, and often heavier than the bone ever was.

Every step of that is unlikely. The animal has to die somewhere sediment is piling up, which rules out most forests, most uplands and most of the places animals actually prefer to live. It has to avoid being eaten and scattered. The chemistry has to cooperate. And then the rock has to survive tens of millions of years of not being melted, crushed, folded or eroded away — and it has to end up at the surface, in a place with no vegetation on it, at a moment when somebody who knows what they're looking at happens to walk past.

The fossil record isn't a sample of ancient life. It's a sample of ancient death in a handful of favourable settings, and it's biased in a way that's easy to state and impossible to correct for exactly. Big-boned animals near water are over-represented. Small, light, forest-dwelling things are largely missing. Most of the dinosaur species that ever lived have left nothing at all, and we've no way of knowing what we're not seeing.

Which is worth holding onto. The dinosaurs we know are the ones that died in the right places.

Reading the rock clock

A fossil with no date attached is nearly useless, and the dating doesn't come from the fossil.

The first half is simple and was worked out by Nicolas Steno in 1669: sediment piles up from the bottom, so in an undisturbed sequence, lower is older. That gives you order but no numbers. A fossil in one layer is older than a fossil in the layer above and that's all you can say.

The numbers come from volcanic ash. Ash beds interleaved with sedimentary rock contain crystals — zircon especially — that lock in uranium when they form and exclude lead. Uranium decays to lead at a rate that doesn't care about temperature, pressure or anything else, so the ratio inside a single crystal is a clock that started the day the ash fell. Find an ash layer below your fossil and one above it, date both, and you've bracketed the animal.

So nobody dates a dinosaur bone directly. They date the volcanoes that happened to erupt near it. That's how we know dinosaurs appear in the Triassic, spread through the Jurassic and Cretaceous, and stop — a run of about 165 million years, which makes the 66 million since their end look brief by comparison.

Radiocarbon has nothing to do with any of this, incidentally. It's useless past about fifty thousand years, and anyone offering you a carbon date for a dinosaur is selling something.

The evidence that records behaviour

Bones tell you what an animal was. They're nearly silent on what it did. For that you need trace fossils — footprints, burrows, nests, tooth marks, droppings — and these are the closest thing palaeontology has to a live recording.

Trackways are the best of them, because a trackway is a few seconds of an animal's life rather than the aftermath of its death. Stride length combined with an estimate of hip height gives a speed, using a relationship the zoologist Robert McNeill Alexander published in 1976 after studying how living animals of different sizes move. The results are sobering for anyone raised on films: most measured dinosaur trackways record a walk, because animals walk nearly all the time. The rare fast trackways come out at speeds a decent human sprinter would find respectable rather than terrifying.

Trackways also show animals travelling together, parallel, at the same speed, in the same direction. Some sites preserve smaller prints among larger ones. That's evidence for herding and possibly for juveniles being kept in the middle of a group — evidence, not proof, and the alternative readings usually involve animals following the same shoreline at different times, which is why individual sites get re-argued for decades. The famous Lark Quarry site in Queensland was read for years as a stampede fleeing a predator, and has since been reinterpreted more than once.

Nests changed the field outright. In the late 1970s, work at Egg Mountain in Montana turned up nesting grounds with eggs, eggshell trampled into fragments, and hatchlings whose leg joints looked insufficiently formed for them to have left the nest — alongside juveniles too big to be newborns. The animal was named Maiasaura, good mother lizard, and the argument was that at least some dinosaurs fed and guarded their young rather than laying and walking off. Later finds from Mongolia sealed the general point in a way nothing else could: several specimens of an oviraptorosaur preserved sitting directly on top of their nests, limbs folded round the eggs, in exactly the posture a brooding bird uses. They died doing it.

Then there are coprolites, which is a polite word for fossil droppings. They aren't glamorous and they're extremely informative, because they contain the actual meal — splintered bone in predators, wood tissue and conifer fragments in herbivores. It's one of the very few lines of evidence that reports directly rather than by inference.

Bones have growth rings

Cut a dinosaur bone into a slice thin enough to see light through it and put it under a microscope, and it stops being a shape and starts being a record. This is bone histology, and it's quietly rewritten more of the field than the headline discoveries have.

Bone is living tissue that lays itself down in layers. Growth slows or stops seasonally in many animals, leaving lines of arrested growth — rings, essentially, and countable like a tree's. Read them and you get an age at death, a growth rate, and whether the animal had stopped growing or was still going when it died.

The results were startling. Large dinosaurs grew fast, not slowly. Tyrannosaurus appears to have spent its early years at a modest size and then gone through an adolescent growth spurt of a couple of kilograms a day, reaching full size in its late teens and rarely living far past thirty. That's a bird-and-mammal growth pattern, not a reptile one, and it fed straight into a much larger argument about how these animals ran their metabolisms.

It also pruned the family tree. Plenty of specimens named as distinct species have turned out, on the microscope slide, to be juveniles of animals already known — the bone showing an animal nowhere near done growing, with the skull proportions to match. Whether Nanotyrannus is a real small tyrannosaur or just a young Tyrannosaurus has been fought over for decades and is still being fought over, with new specimens shifting the argument in both directions. It's an unresolved question, and anyone presenting either answer as settled is overselling it.

Guessing from living relatives, carefully

Dinosaurs left descendants and they left close relatives, and there's a formal method for using both. It's called the extant phylogenetic bracket, set out by Lawrence Witmer in 1995, and it's more disciplined than it sounds.

The logic runs like this. Birds are dinosaurs. Crocodilians are the nearest living things that aren't. Any extinct dinosaur sits on the family tree between the two. So if a trait is present in birds and in crocodilians, the simplest explanation is that their common ancestor had it and everything in between inherited it — which lets you infer it for a dinosaur with reasonable confidence, even though nobody watched one do it.

That gets you a surprising amount. Hard-shelled eggs. A four-chambered heart. Certain skull cavities and muscle attachments, because both living groups anchor the same muscles to the same scars. Building nests and tending them at least sometimes. Gastroliths — swallowed stones for grinding food — turn up in both, and turn up in dinosaur skeletons too.

What the method also does, which matters more, is tell you when to stop. A trait found only in birds doesn't transfer, because it might have evolved after the split. Singing doesn't transfer. Feather colour patterns for display don't transfer. Nor does anything requiring soft tissue that neither living group agrees on. The bracket is a fence as much as a tool, and the good papers spend as much space on what it forbids as on what it licenses.

The feathers are not a guess

For most of the twentieth century dinosaurs were drawn scaly, and the drawing was an assumption dressed as knowledge. Nobody had integument for most species. Reptiles have scales, dinosaurs were reptiles, so scales it was.

The link to birds had been on the table since the 1860s. Archaeopteryx came out of the Solnhofen limestone in Germany in 1861, two years after Darwin published, with feathers and a long bony tail and teeth, and Thomas Henry Huxley was arguing within the decade that birds had come from dinosaurs. The idea then went quiet for most of a century until John Ostrom described Deinonychus in 1969 — an agile, lightly built predator with a bird-like wrist that looked nothing like the tail-dragging swamp monsters of the textbooks — and set off the run of reappraisals that Bob Bakker later christened the dinosaur renaissance.

The physical proof arrived from north-eastern China in the 1990s. The lake beds of Liaoning were periodically smothered by volcanic ash so fine that it preserved soft tissue in outline, and out of them came fossil after fossil with unmistakable filaments and, later, unmistakable branching feathers. Sinosauropteryx in 1996 was the first non-bird dinosaur described with a filamentous coat. Others followed with feathers so structurally modern you could count the barbs, on animals that plainly could not fly and had no ancestors that flew.

That last point is the important one. Feathers came first and flight came much later, which means feathers didn't evolve for flying. Insulation and display are the leading candidates, and they're inferences, unlike the feathers themselves.

Feathered coverings are now solidly established across the coelurosaurs — the group containing tyrannosaurs, dromaeosaurs and birds. Beyond that group the evidence thins out, and here the honest position is uncertainty. Filaments and bristles have been reported in ornithischians on the other side of the family tree, which would push the origin of feather-like structures right back to the base of the dinosaurs, and whether those structures are genuinely related to feathers or a separate invention is not resolved. A fully feathered adult Tyrannosaurus is a defensible reconstruction rather than a demonstrated fact — large-bodied animals shed heat poorly, several tyrannosaur skin impressions show scales, and the argument continues.

How we know a few of the colours

Colour was assumed to be permanently lost. Every dinosaur you saw before about 2010 was coloured by an artist making a reasonable decision, and the greens and greys had no evidential basis whatsoever.

Then people started looking at melanosomes. These are microscopic pigment-bearing structures inside feathers and skin, and in living birds their shape correlates with colour — long and rod-like for blacks and greys, small and spherical for reddish browns, with a distinctive stacked arrangement producing the metallic iridescence you see on a starling. Where a fossil is preserved finely enough, the melanosomes survive as impressions or as carbon, and their shapes can be measured and matched against a reference set of living birds.

Actual published reconstructions exist for a small number of animals. Anchiornis, a small feathered dinosaur, was reconstructed in 2010 as mostly grey with white-spangled wings and a reddish crest. Sinosauropteryx came out chestnut and white with a banded tail. Microraptor turned out to have been glossy iridescent black. An extraordinarily preserved armoured dinosaur found in an Alberta mine, Borealopelta, showed reddish-brown pigment on its back and none underneath — countershading, the pattern animals use to flatten their own shadow, which in something the size of a rhinoceros implies it still had predators to worry about.

Set against the thousands of named dinosaur species, that's a handful. And the method has real limits. It needs exceptional preservation. It reads melanin-based colours only, so blues, most yellows and any colour produced by other pigments or by feather microstructure leave nothing comparable behind. There was a serious argument, published in 2014, that some of the structures being identified as melanosomes were fossilised bacteria instead, and while the balance of evidence has gone the other way, it forced the field to tighten its criteria.

So a picture of a ginger-and-white Sinosauropteryx tail is a claim with physical evidence behind it. A picture of a bright blue sauropod is a painting. Both can hang in a museum. They aren't the same kind of statement, and the caption ought to say which is which.

Birds never stopped being dinosaurs

All of the above converges on a conclusion that still lands oddly: birds aren't descended from dinosaurs in the way we're descended from fish. They are dinosaurs, in the same sense that bats are mammals.

They sit inside the theropod group, the same branch that holds Velociraptor and Tyrannosaurus. The shared features aren't superficial — hollow air-filled bones with the same pattern of openings, a wishbone, a backward-pointing first toe in some lineages, the same wrist joint that folds the hand sideways, the same brooding posture over a nest, eggshell with the same layered microstructure.

Which means the extinction was a pruning, not a clearance. When people say the dinosaurs died out, what they mean is that every dinosaur lineage except one small group of feathered flying theropods died out. There are something like eleven thousand species of them alive today, and one of them is standing on your bin.

The day it ended, and the argument about it

The killing mechanism is one of the better-evidenced events in deep time, and the case was built backwards from a chemical anomaly.

In 1980 Luis and Walter Alvarez and their colleagues reported that the thin clay layer marking the boundary at the top of the Cretaceous, wherever it was sampled, contained iridium at concentrations far above anything normal for the Earth's crust. Iridium is common in asteroids and scarce at the surface here, because most of ours sank into the core long ago. They proposed a large impact. It was not warmly received.

The crater turned up in 1991, and it had already been found once — buried under the Yucatán Peninsula in Mexico, spotted years earlier by geophysicists prospecting for oil who had no reason to connect it to anything. Chicxulub is roughly 180 kilometres across, dated to the boundary, and consistent with an object something like ten kilometres wide arriving at a speed nothing survives. Shocked quartz, glass spherules and a tsunami deposit ring the Gulf. It's an unusually complete forensic case.

What's still argued is whether that was the whole story. The Deccan Traps in India were erupting on a colossal scale across the same interval, pumping out gases capable of wrecking a climate on their own, and there's a live debate about how much of the extinction to assign to each, and whether the impact aggravated the volcanism. Palaeontologists also argue about how much groups were already declining beforehand, which is genuinely difficult to answer because the fossil record thins near a boundary for reasons that have nothing to do with biology.

The impact is not seriously in doubt. Its share of the blame is.

Why the picture keeps changing

Reconstructions get revised constantly, and that ought to raise your confidence in the field rather than lower it.

Take the earliest attempt. Richard Owen coined the word Dinosauria in 1842, and when the Crystal Palace sculptures went up in 1854 they showed Iguanodon as a heavy four-legged animal something like a rhinoceros, with a horn on its nose. The horn was a thumb spike that nobody had a complete enough skeleton to place. The sculptures are still standing in south London, and they're wrong in a wonderfully instructive way.

Take the tail. For most of the twentieth century, big two-legged dinosaurs were drawn upright like kangaroos with their tails dragging behind them. The trackways showed no tail-drag marks. None, anywhere, for decades, while the illustrations kept the tails on the ground. The evidence had been sitting in the rock the whole time, and the reconstruction eventually caught up: horizontal body, tail out behind as a counterweight, which is how they're drawn now.

Take Deinocheirus. A pair of gigantic arms came out of Mongolia in 1965 and there the matter rested for nearly fifty years, with reasonable people assuming an enormous predator. Fuller specimens described in 2014 revealed something nobody had drawn: a huge, humped, duck-billed, largely plant-eating oddity. The arms were the least strange thing about it.

And take the names. Brontosaurus was sunk into Apatosaurus in 1903 and stayed sunk for over a century, then a large 2015 reanalysis argued it deserved separating out again. That's still disputed. Nothing about it is a scandal — it's what happens when a discipline keeps checking its own back catalogue.

The dinosaurs of your childhood were drawn with confidence and were partly wrong. The dinosaurs in today's museums are drawn with confidence and are partly wrong in ways nobody has identified yet. That isn't a weakness of the science. It's the only honest thing a portrait can be when the sitter has been dead for 66 million years and left nothing behind but footprints, stone bones, and a few square centimetres of preserved feather.