Drop a book in water, fish it out and dry it in the sun: the paper goes brittle, it crumbles, the ink bleeds. DNA inside a bone buried for tens of thousands of years meets the same fate.
So for a long time, people from that far back were only shapes to us — the shape of a skull, of a tooth, of a stone tool. Who was related to whom was guessed from angles and ratios measured off bone.
From the 1980s some people tried to bypass shape and read the letters still left inside the bone. For the first twenty years it was almost all false alarms. When the road finally opened, what got rewritten was not the details. It was the main line.
Fold a sheet of paper over and over and tear it, and you end up with a pile of scraps carrying a few letters each. DNA in a bone does the same thing, only far more slowly: the long chain snaps at one point after another into shorter and shorter fragments.
How slowly can be measured. In 2012 someone picked a nearly ideal material: the extinct New Zealand moa, 158 bones, each independently dated, all buried in comparable conditions. The result was a very clean decay curve — a 242-letter stretch of mitochondrial DNA loses half of itself about every 521 years, at an effective burial temperature of around 13°C. Cold ground is much slower.
Besides breakage there is a second change, and it turned out to be the key to the whole business. One of DNA's four letters is called cytosine, written C; it slowly loses an amino group of its own accord and becomes something else, which a sequencing machine reads as T. This happens especially often at the two ends of a fragment, because the letters there sit on a loose single strand with no partner half to shield them. So real ancient DNA carries a fixed medical chart: fragments only tens of letters long, a markedly raised share of end-position Cs read as T — and how bad those two are has to match the age and burial temperature the sample claims.
Why does that chart matter? Because for this field's first two decades the enemy was not decay but contamination: the intact human DNA in the spray from one sentence spoken beside the bench outnumbers what is left inside a bone by several orders of magnitude. The 1985 paper reporting DNA cloned from an Egyptian mummy was a sensation; later re-examination concluded that most of it was modern human DNA that had got in. The 1990s brought insects in amber and "dinosaur DNA" from a coal mine — the latter, on inspection, resembled nothing so much as a human.
The way out was to treat that chart as a signature: authentication does not rest on looking clean; it rests on carrying the particular kind of dirt that time necessarily leaves.
The rest is legwork, and one piece of it paid off enormously. In 2015 someone compared different parts of the same remains: in ordinary bones and teeth only 0.3% to 20% of the DNA actually belongs to the individual, the rest being soil bacteria and contamination — while the dense little nub of the temporal bone that wraps the inner ear, the petrous, runs from 37% to 85%. A great many remains that had been written off as unreadable turned into usable samples overnight.
The method has now moved beyond bones altogether: in 2017 researchers pulled Neanderthal and Denisovan mitochondrial DNA straight out of cave sediments, at spots where no matching bone was ever dug up. The current age ceiling is a mammoth tooth from Siberian permafrost, around 1.2 million years old. Dinosaurs in amber remain impossible — that is still two orders of magnitude away.
In 1997 a short stretch of mitochondrial DNA was read out of the specimen dug up in Germany in 1856 — the first Neanderthal ever recognised as such. Mitochondria are small components inside cells, each with its own little loop of DNA that passes only from mother to child; there are many copies per cell, so that DNA survives more readily, which is why early ancient-DNA work could read nothing else.
That sequence sat well outside the range of every living person. At the time this looked like evidence for clean replacement: the people who walked out of Africa arrived, the resident Neanderthals went, and the two never mixed. The trouble is that mitochondrial DNA follows exactly one line — your mother's mother's mother. If there was interbreeding but that maternal line later died out, it sees nothing at all.
When the draft Neanderthal nuclear genome appeared in 2010, the conclusion flipped. What flipped it was not positive evidence but a harder question: if "clean replacement" were true, what would have to be equal? If everyone outside Africa descends from the same group that left, then every population outside Africa should be equally similar to Neanderthals.
It is not. Europeans and East Asians share more sites with Neanderthals than sub-Saharan Africans do — and the excess is not scattered about, it comes in blocks: whole stretches of tens of thousands of letters matching Neanderthals end to end. Similarity inherited from a common ancestor gets ground to powder by tens of thousands of generations of shuffling. Similarity in long blocks can only have been dropped in later.
The current estimate is that people outside Africa generally carry one to two per cent Neanderthal ancestry. In 2024 two independent studies pinned that mixing to roughly 47,000 years ago, lasting some seven thousand years.
Something stranger happened the same year as the draft genome. A finger bone came out of Denisova Cave in the Altai Mountains of Siberia, far too small to identify by shape. Only once its genome was read did it become clear that it was neither Neanderthal nor modern human, but a lineage nobody had known existed. This was the first time a human lineage was defined by a genome rather than by a skull you could set on a table. They came to be called Denisovans, and some populations in Oceania carry a far higher share of Denisovan ancestry. In 2014 it emerged that the gene variant associated with high-altitude, low-oxygen adaptation in Tibetans is arranged almost exactly as that stretch is in the Denisovan genome. Do not read this as Denisovans giving Tibetans the ability to live high up — the variant simply happened to enter this gene pool through one episode of interbreeding, and at altitude the people carrying it went on to leave more offspring.
The bluntest piece of all came in 2018: a two-centimetre bone fragment from the same cave, half of whose chromosomes look Neanderthal and half Denisovan — a Neanderthal mother, a Denisovan father, first generation. Very few archaic humans have had their genomes read at all, and one of them turns out to be a first-generation hybrid; that cannot be turned into a frequency, but it does suggest such things were no great rarity at the time.
Standing in front of a site a few thousand years old, it is hard not to assume that the people living there today descend from the people who lived there then. Europe has enough samples in the ground to test that sentence directly. The answer is: not really.
Present-day Europeans carry at least three layers, and each one arrived not as "a new technique was learned" but as a substantial replacement of people. At the bottom are the hunter-gatherers who had been there since the ice retreated. Around eight thousand years ago people carrying farming moved in from Anatolia (roughly today's Turkey), in many places largely replacing the existing population — farming reached Europe not as an idea travelling but as people walking. Around four and a half thousand years ago another group came from the steppe to the east: the Corded Ware population in Germany traces about three quarters of its ancestry to the Yamnaya of the steppe.
Go down and there is no pure floor. The piece on human evolution noted that sixty thousand years ago several kinds of human were on Earth at once; the sentence to add is that they did not merely coexist, they left shared descendants — and within our own species this repeated mixing has never stopped. "Pure ancestry" is an empty concept against this data: beneath every layer you take for the original inhabitants lies another that was itself replaced.
For that reason the conclusions here have to be read carefully. These percentages are not absolute quantities; they are similarities measured against some set of present-day reference populations, and swapping the reference set moves the numbers — when an ancestry test tells you "23% Northern European", it is saying "more like the people sorted into that box today", not "your ancestors lived in Northern Europe". At the other end, archaeology plus migration has always been favourite material for nationalism, and ancient DNA is precisely what closes that road: every lineage it produces leads, a few thousand years back, into another mixture.