You walk past the bins outside your building and a few stray cats are lying there. One bolts while you are still five metres off. Another stops and looks at you, and only shifts, reluctantly, when you come closer.
That distance — how near you can get — varies a lot within one group of cats, and you can measure it with a tape. Ten-odd thousand years ago the same variation existed in the wolves around human camps.
What happened next sounds like a story about someone kind. It is nothing of the sort.
Bones and offal that no one could finish piled up at the edge of camp. To a wolf that is a steady, easy food supply that does not have to be chased down. Whether it could be reached depended on one very concrete thing: how close to a human the animal could get before panic set in.
Wolves with a higher panic threshold ate a little more and raised a few more pups; those pups were, on average, a little less afraid again. Generation after generation, this group separated from the ones that stayed wild. Nobody decided anything here, and there was no "growing tame in order to get the scraps": the mutations that lower fear turn up at random, and the camp simply let the animals carrying them leave more descendants. The first step was probably not taken by people at all. The wolves walked over on their own.
One physical object pins the timeline down. In 1914 a grave was dug up near Bonn, Germany, holding a man, a woman and two dogs, dated to roughly 14,200 years ago. In 2018 someone re-examined the teeth of one of the puppies and found that at around 19 weeks old it had contracted canine distemper — a disease with a three-week course and very high mortality — and yet it lived another eight weeks or more. The disease record written into its teeth means somebody was feeding it throughout those eight weeks. That is no longer a scavenging relationship. Genetics gives a different kind of ruler: 27 ancient dog genomes sequenced in 2020 show that by about 11,000 years ago dogs had already split into at least five markedly distinct lineages, and splitting five ways takes a long time, so domestication itself is considerably older. Where it started, and whether it happened once or several times, is still argued about.
From 1959, Belyaev and Trut in Novosibirsk bred a line of silver foxes and each generation selected on a single criterion: reach a hand in, and does the animal flinch or snap? Not coat colour, not ears, not body size. A few dozen generations later, a string of things nobody had selected for had appeared — floppy ears, curled-up tails, patches of white in the coat, longer breeding seasons. Part of this string shows up in many domesticated animals, and it has a name: the domestication syndrome.
Why should turning one knob produce a string of outputs? The most popular explanation at the moment points at a group of cells. Very early on, the embryo sends out from its back a population of cells that migrate all over the body and end up building several unrelated things — the figure below shows where each of them goes. If what you selected for is "a weaker stress response", what you may actually have selected is "fewer of these cells, migrating more slowly", and then cartilage, pigment and jaw get dragged along with it. Nobody ever selected for ears.
Honestly, though: the explanation is elegant and the evidence is not yet solid. A public exchange in 2021 had critics pointing out that the candidate genes found in domesticated animals only show that structures derived from these cells have changed, not that other traits are dragged along by the same mechanism. It is unsettled.
The foxes themselves are also compromised. A piece published at the end of 2019 noted that Belyaev's silver foxes came from a Canadian fur farm that had already been breeding them for decades, where tameness had been recorded and white patches had appeared before the Russian experiment began. The textbook version — select for tameness alone and a whole set materialises out of nowhere — loses a large chunk of its force. But the core mechanism does not rest on the foxes: select one trait and you shift the traits tied to it, which every breeder meets daily.
When wild wheat and barley ripen, the ear breaks apart segment by segment and the grain scatters on the ground — reaching the ground is the only way it germinates next year.
A person with a sickle brings home the ears that have not broken apart, the ones still holding their grain. Those grains go back to camp and are sown the following year. So "the ear does not shatter on its own", a trait that means no descendants in the wild, became the only trait that could be passed on in human hands, and a few other things were selected along with it: seeds that germinate together instead of trickling out over several years, and thinner seed coats.
No generation here was "domesticating wheat". In the Near Eastern archaeological record, barley before 9500 BC is almost entirely the wild, shattering type; between 9000 and 8500 BC non-shattering rachis fragments reach about 30% at a few sites; only after 7400 BC do some sites turn up uniformly non-shattering. Two thousand years from end to end, and in between every generation was simply cutting ears.
Maize is starker: every grain of its ancestor teosinte sits inside a hard casing, and one gene change opens that casing — after which the grain can neither disperse nor protect itself.
One side effect is worth remembering on its own: only a small slice of the wild population was ever planted, so generation by generation the variation within a crop keeps shrinking. A whole field is one answer, and a pathogen that solves that one answer takes everything. The arms race from the Red Queen topic goes one-sided here — the pathogen iterates several times a year while we hold the crop's genome still. That is how Ireland's potatoes collapsed in the nineteenth century.
Set a modern human skull beside early fossils from 300,000 years ago and beside a Neanderthal, and the direction is fairly consistent: flatter brow ridges, less facial projection, a rounder braincase, smaller teeth. That string overlaps considerably with the domesticated-animal set. Behaviourally there is something even more striking: you spend every day packed into a carriage with hundreds of strangers and nobody fights, whereas unfamiliar chimpanzees at that density would kill each other.
This is what the self-domestication hypothesis is about: the selection pressure came from inside the group. One specific version has it that language and coalitions let a group jointly deal with its most domineering member, so the people quickest to lash out left fewer descendants. Nothing here runs "for the good of the group" — what does the work is individual descendant counts: the one who gets dealt with collectively leaves fewer.
The strongest supporting line is the bonobo. It is a close relative of the chimpanzee, split off roughly one to two million years ago with no human involvement anywhere in the process; it is markedly less aggressive towards its own kind, and its skull and teeth have shifted somewhat in a juvenile-looking direction. In 2012 it was formally written up as a naturally occurring case of self-domestication — and having that natural experiment means this set of changes does not need anyone to do the sieving.
Why is it still only a hypothesis? With dogs you can name who was sieving and which trait; here the selector is entirely inferred, and the trait list looks partly assembled after the fact — a rounder braincase has other possible explanations, such as a change in the pace of brain development. It is interesting, it has partial support, and it is far from settled.
One more thing has to be said plainly: this hypothesis is a comparison between species, asking why humans differ from chimpanzees on a scale of hundreds of thousands of years. Any reading that shifts it onto "which group of people is more domesticated than which" has no evidence behind it — one is a difference between species, the other a difference within one; the two scales simply do not line up.