Nobody Knows Where the Dog Came From

The dog was first. Before the sheep, before the goat, before wheat or barley or rice — and it is the only animal we domesticated while we were still hunting and gathering, thousands of years before anyone planted anything on purpose.

It has lived in our houses ever since. There are more written words about dogs than about most human beings. Every generation for two centuries has had somebody working on this.

And we cannot say where it happened, when it happened, or how many times.

This is a chain, and it runs from 1868 to March 2026. Every joint in it is marked — documented where the step is established, contested where serious people disagree, leap where we are the ones reaching. A chain is built to feel inevitable. This one is not allowed to.

What this chain is not. It is not an argument that the science failed. It is the opposite. The most striking thing in the whole sequence is a sentence in the most recent paper, written by people with 216 canid remains and a technique that did not exist a decade ago, saying plainly that they still cannot resolve it. That is what the work looks like when nobody is smoothing it.

The argument opens where most arguments in this field open, and the popular retelling gets it slightly and revealingly wrong.

You will read that Darwin wondered whether dogs came from one species or from some unusual mating — a wolf and a jackal, perhaps. He did consider it. But he finished the chapter, and he finished it with a verdict.

Several wild species, in several places. Not a question — a position, argued from breed antiquity, from geography, and from the resemblance of each region's dogs to each region's wild canids.

He was probably wrong. Every genome since 1997 points to wolves alone. But notice what the smoothing did: it turned a scientist's stated conclusion into a musing, which is a tidier thing to have been wrong about. Being wrong out loud is not the same as not having decided.

At Mallaha in northern Israel, under the floor of a house about twelve thousand years old, an elderly human was buried with a hand resting on a puppy.

Davis and Valla published it in Nature in 1978, and for a while it was the answer: dogs were domesticated in the Middle East, just before people took up farming. The evidence is a bone and a gesture, and the gesture is doing a lot of the work.

Later finds pushed the date back four thousand years and moved the place to Eurasia, so this stopped being the answer. It did not stop being what it is.

Nothing that came after has undone this burial. Every genome in the rest of this chain is arguing about a relationship that was already, demonstrably, this — close enough that somebody arranged a body that way and someone else agreed to bury it so.

Careful with the gesture. A hand on a puppy is powerful and it is not data about domestication timing. It tells us these animals were inside human ritual life by then. It does not tell us where they came from, which is the question this chain cannot answer.

Mitochondrial DNA from 162 wolves at 27 localities and 140 dogs across 67 breeds. The finding that has held ever since: the ancestor is the wolf. Not the jackal, not several canids. Wolves.

The paper's title is Multiple and Ancient Origins of the Domestic Dog, and its abstract says the divergence “suggested that dogs originated more than 100,000 years before the present.” That date did not survive. The figure that survived instead was 135,000 — the top of the paper's range, not the claim in its own summary, and the number that went into circulation and stayed there.

Now look at the author list. Nine names. Two of them are Peter Savolainen and Robert Wayne.

The dog wars did not begin as a clash of schools. They began inside one set of results, when the same evidence stopped supporting one story.

Savolainen's camp went large. Whole mitochondrial genomes for 169 dogs, and the control region for 1,543 dogs across the Old World.

The pattern is a gradient. All ten major haplogroups appear together in exactly one place — southeastern Asia, south of the Yangtze River. Seven in central China. Five in northern China and southwest Asia. Four by the time you reach Europe.

Highest diversity means oldest population, so: one origin, in that region, recently, from many wolves. The title states it outright — “mtDNA Data Indicate a Single Origin for Dogs South of Yangtze River, Less Than 16,300 Years Ago, from Numerous Wolves.” There is no hedge in it anywhere.

Why this is marked contested. The inference runs entirely through living dogs. Trade routes moved animals across Eurasia for millennia, and a place where many lineages ended up looks identical to a place where many lineages began. Wayne’s objection was exactly this, to a reporter: “It would be like concluding that humans arose in the United States because there’s so much genetic diversity here.”

The same year, and on the other rail entirely: Mietje Germonpré went through her own museum's archives in Brussels and found a canid cranium from Goyet Cave, in southern Belgium, that had been catalogued as a wolf.

She measured it. The proportions read as dog. Radiocarbon dating put it at 31,680 ± 250 years before present — more than twice the age of anything else then being called a dog.

If that identification holds, everything in this chain moves. Dogs would predate agriculture by twenty thousand years, and would have walked with people through the last glacial maximum.

The response was immediate and hard. Critics called the analysis premature and misleading, and said the specimen might simply be an unusual-looking wolf. Germonpré's own reply is the interesting one: it may have been an early dog that led nowhere — a domestication that started and then stopped, leaving no descendants.

She has described the field, in print, as more combative than any other subject in prehistory.

Six years later, Abby Grace Drake and colleagues put the Goyet skull — and a second contested specimen from Eliseevichi in Russia, dated 13,905 ± 55 years — through 3D geometric morphometrics.

Not lengths and widths. Hundreds of landmark coordinates, describing the whole shape at once: the angle of the snout against the braincase, the set of the eye sockets, the curvature above the muzzle.

Their conclusion is in their title. The fossil skulls contradict the suggested Palaeolithic domestication. Compared with wolf-like dog breeds, they report, Goyet and Eliseevichi have no cranial flexion, and no concavity on the dorsal surface of the muzzle near the orbits — two features dogs have and wolves do not.

This joint is marked contested and it stays that way, because Germonpré's group did not concede and the exchange is still running. Both sides are measuring the same bones with real methods and reaching opposite answers.

Why the disagreement is structural, not sloppy. Early domestication is a process, not a moment. There is no reason a population halfway through it should sort cleanly into “dog” or “wolf” on any measurement — and a method that must return one of two labels will return one whether or not the animal was one.

Wayne's answer to you are reading living dogs was to stop reading living dogs. Thalmann and colleagues sequenced complete mitochondrial genomes from 18 prehistoric canids across Eurasia and the New World.

Every modern dog, they found, is most closely related to ancient or modern canids of Europe. Molecular dating put the onset of domestication there at 18,800 to 32,100 years ago — among European hunter-gatherers, long before farming.

Savolainen's objection came straight back, and it was the mirror image of Wayne's: the study included no samples from southeast Asia. Like studying the geographical origins of humans without a single sample from Africa. Wayne's answer was that there were no ancient Chinese specimens to include — which he took as further evidence against an origin there.

Read the author list again, though. Germonpré is on this paper. So is Wayne. The Goyet skull's discoverer and the leading European-origin geneticist are co-authors here, and the Yangtze camp is not.

By this point Wayne had said in print that the two sides had reached an impasse and did not interact much. Two literatures, two sets of samples, two answers, and no shared bench.

Greger Larson and Keith Dobney had spent years on pigs, not dogs, and had watched the same mistake happen there: modern DNA suggesting many separate domestications, ancient DNA and skull shape together revealing that trade and interbreeding had faked most of them.

They made a funding case in 2012 that was partly a provocation — without dogs you don't have any other domestication; you don't have civilisation — and it worked. Three million dollars. Then they started phoning the people who were not speaking to each other.

By the end of 2013 they had fifteen collaborators, including Wayne, Savolainen and Germonpré, in a two-day meeting Dobney remembers as tense. Larson's pitch was to take the credit out of it: he has said he did not care where his name fell on the papers, he just wanted them out.

The collaboration grew to around fifty scientists and thousands of specimens. Both antagonists went on record saying they would accept being wrong. Savolainen: if I'm wrong, it will be a bit embarrassing, but science is about trying to find the truth.

The first big paper out of the new arrangement sequenced the complete nuclear genome of a dog that died about 4,800 years ago at Newgrange in Ireland, at 28-fold coverage, alongside 59 ancient mitochondrial sequences and 80 modern genomes.

It found a deep split between East Asian and Western Eurasian dogs. And then the surprise, which is the paper's own word for it: the split dates to about 14,000 to 6,400 years ago — after dogs already existed in both Europe and East Asia.

A split that postdates the animals it divides is not a family tree. The reading they offered: dogs may have been domesticated independently, in eastern and western Eurasia, from different wolf populations — and then East Eurasian dogs travelled west with people and partially replaced the European ones.

Both camps were half right. Both had been sampling a real population.

It did not settle. Wayne called the result plausible but too preliminary. And the paper's own framing is careful: suggests, may have been, possibly. The headline that travelled dropped every one of those.

Twenty-seven ancient dog genomes, up to 10,900 years old, co-analysed against 17 sets of human genome-wide data matched for age, place and culture. This is the paper the truce was built to produce.

Three findings, all stated in its abstract:

One. By 11,000 years ago there were already at least five major ancestry lineages — giving rise to dogs in the Near East, northern Europe, Siberia, New Guinea and the Americas. The diversity is Palaeolithic. Whatever happened, it happened long before.

Two. All dogs share a common ancestry distinct from present-day wolves, consistent with descent from a single ancient, now-extinct wolf population.

Three, and this is the strange one. There is “limited gene flow from wolves since domestication but substantial dog-to-wolf gene flow.” Every other domesticate keeps trading DNA with its wild relative. Not this one. For eleven thousand years the genes ran one way — out of dogs, into wolves.

Their own wording is that flow into dogs has been “so limited as to be undetectable at the current resolution of the data.”

The paper reports the asymmetry. It does not explain it. What follows is an explanation Greger Larson gave to a reporter, and it is marked leap because that is what it is — a hypothesis in an interview, not a result in a paper.

His reasoning: if your pig picks up some wild boar, you do not much mind, because you are going to eat the pig. A dog is different. A dog with wolf in it makes a worse guard, a worse hunting companion and a worse friend. So, he said, people get rid of the dog.

If that is right — and it is a plausible story rather than a finding — then the asymmetry in the genome is a record of eleven thousand years of removing the animals that did not fit, on a scale nobody was recording and no one intended as a programme.

We are not going to cash that. This is the joint where a chain like this usually reaches for the moral, and the honest position is that the mechanism is untested. Selective removal is one explanation. So is a demographic bottleneck, so is simple proximity: dogs lived where wolves mostly were not.

Marked leap, and here is exactly what is being leapt. The asymmetry is measured and solid. Why is not measured at all. A reader who takes only one thing from this spread should take that the finding and the story about the finding arrived from different places — the first from a genome, the second from a phone call.

Six months ago, the same collaboration published again: 216 canid remains, 181 of them from Palaeolithic and Mesolithic Europe, read with a genome-wide capture method that enriched the surviving DNA by ten- to a hundred-fold.

The oldest dog they could read is 14,200 years old, from Kesslerloch in Switzerland — an animal somebody had already called a dog from its bones alone, now confirmed from its genome.

And it shares ancestry with later dogs worldwide. Which kills a hypothesis:

It also already looked more European than Asian at 14,200 years — so the diversification had started well before that. Their inference: domestication is likely several millennia older again.

Then the sentence this whole chain exists for.

Here is what the chain leaves us with, and the last step is ours rather than anybody's data, which is why it is marked.

Two centuries of work. Two rails — bone and genome — that kept contradicting each other. Skulls re-measured and reclassified. A truce that had to be paid for. And a current answer that is, in the field's own words, we cannot resolve it.

Meanwhile the relationship never once depended on the answer. Somebody at Mallaha arranged a hand on a puppy twelve thousand years ago without knowing any of this. Nobody has ever needed the origin story to have the dog.

And that is worth saying out loud in a place like this, because where did it come from is a question people also ask about us. About a diagnosis. About a trait. About a family. The demand for a causal origin story arrives dressed as rigour, and it very often arrives before any actual willingness to sit in the same room.

The dog is the best-studied domestication in human history, and its origin is unresolved, and none of that has ever made a single dog less ours or us less theirs.