Your hands go first, which is the humiliating part.
You walk past a schoolyard, or a playground, or the strip of concrete behind a church hall where somebody's birthday is happening, and two kids are doing the thing with the palms. Clap, cross, clap, hit her hands, hit the backs of her hands, faster, faster. Before you've identified the tune, or agreed to identify the tune, something in your forearms has started keeping time.
Uninvited.
You don't get a vote on this.
Then the words show up, and the words are appalling. They always are. Every clapping rhyme on the planet is about a death, or a body, or somebody's boyfriend behaving badly, or a sailor who went to sea sea sea, delivered at speed and with excellent diction by seven-year-olds who would be startled to hear they're transmitting anything. Ask one what a plague rhyme is. She has no idea. She's just doing the hands.
Nobody taught you that. Not a teacher, not a parent, and nothing anybody handed you off a shelf. If an adult had taught it to you it would have died on contact, because the whole mechanism depends on no adult being anywhere near it.
Two people went and measured this. In the 1950s Iona and Peter Opie, married, working mostly out of their own house, collected schoolyard rhymes and riddles and skipping songs and jeers from around 5,000 children at 70 schools across England, Scotland and Wales, and published it in 1959 as The Lore and Language of Schoolchildren. (They collected the insults too. Pages of them. Filed by region. The insults are the best thing in the book and I would like that on the record.)
What they found is the part worth sitting with. The material moved child to child, out of adult hearing, and it moved fast and it moved far, turning up nearly word for word in towns hundreds of miles apart with nobody adult carrying it between them. And it was conservative. Adult culture is never conservative like this. Children don't improve on the rhyme. Get one wrong in front of a nine-year-old and you'll be corrected before you've finished the line.
They also settled an argument the adults of the period were very confident about. The assumption at the time was that radio and television and the whole entertainment apparatus were about to wipe children's own traditions out, and the Opies went and demonstrated that the traditions were sitting there in the playground, unbothered, absorbing the new material and carrying on. Kids took the tune off the television and put their own words on it. Kids had been doing exactly that to hymns for four hundred years.
So the folk transmission everyone assumes is vague and drifty is, in the one place anybody has had a proper look at it, ferociously accurate.
Anyway. Hold on to that, because somebody ran the same question at a much longer range, using software built to trace the ancestry of species.
The 2016 study, and what it actually did
In 2016, Sara Graça da Silva and Jamshid Tehrani published a paper in Royal Society Open Science with the title "Comparative phylogenetic analyses uncover the ancient roots of Indo-European folktales." Phylogenetics is the branch of biology that reconstructs family trees from present-day evidence, working backwards from what's here now to what must have been there before. It's the machinery behind every diagram showing you which animal is cousin to which other animal.
They pointed it at fairy tales. Yup.
The setup: take 50 Indo-European-speaking populations, take 275 tale types from the Tales of Magic section of the folklore index, and code each population for whether it has each tale. Then put a language family tree underneath as the model of how those populations are related to each other, and ask whether the tales track the tree or track the map.
Plenty of them tracked the tree.
That's the first finding and it isn't nothing. It's the difference between inheriting a story from an ancestor population and picking it up from whoever was living next door. Getting there took a statistic that measures how strongly a trait clusters by ancestry, plus a second model to check whether the clustering was just neighbours being neighbours. They then ran ancestral state reconstruction, which is the step where you ask the model to say which nodes further up the tree probably had the tale, and let it run as deep as it would go.
One tale went all the way down. ATU 330, "The Smith and the Devil," reconstructed to Proto-Indo-European with 87% posterior probability. Posterior probability, in this setting, means roughly how much of the model's own guessing lands on that answer once you've fed it everything, so 87% is confident. It is not a carbon date. Four other tales cleared 50% and never reached the threshold the authors had set for themselves, so the authors declined to claim them, which I mention because that's exactly the part that evaporates when a study becomes a headline.
The tale itself is about four sentences long. A blacksmith makes a bargain with the devil, or with death, or with a demon depending on where you're standing, trading his soul for the power to weld any two things together. He then welds the devil to something immovable and walks away.
That's the plot. Faust is a late, fancy, badly overwritten descendant of it.
Under the steppe hypothesis for Indo-European origins, that node sits somewhere around 5,000 to 6,000 years ago, in the Bronze Age. Which is where the six-thousand-year-old-story headline came from, and it went absolutely everywhere, and it remains the most exciting sentence anyone has ever written about a blacksmith. It is also not quite what the paper says.
Where the dating gets thin
The authors put their own caveats in the paper. Which is the behaviour you want, and almost never what gets reported.
Phylogenetic signal degrades. Tales get dropped, borrowed, swapped and replaced, populations turn over, and the deeper you push the less the model knows. Their Proto-Indo-European node carried 77% posterior probability on its own, before you asked it a single question about smiths. And the entire exercise rests on the index being a decent census of what's out there.
Then there's the criticism from outside. It's real criticism, not the decorative kind.
The folklorist John Lindow raised the objection that does the most damage for the least effort: Proto-Indo-European may not have had a word for "smith." A story about a blacksmith requires blacksmiths. If the reconstructed vocabulary doesn't contain the job, the tale is younger than the node it's been pinned to, and the number comes down. Possibly a long way down.
Second, the index has a geographic bias. It's a catalogue of European and Near Eastern tale types, assembled by European scholars out of European collections, and what got collected reflects who was doing the collecting and where they could get to. Julien d'Huy and colleagues pressed this in PNAS in 2017 against a related study, and Tehrani and colleagues published a reply, and the exchange is worth reading if you enjoy watching two sets of people who are each right about something. Nobody wins. Nobody was going to.
Third, and this is the deep one: a tree is the wrong shape for a story. Species mostly can't hand genes sideways to unrelated species, which is what makes the tree work in biology. Stories do nothing but hand themselves sideways. Traders, soldiers, servants, marriages, raids, and eventually print. Mark Liberman made that point at Language Log and added a second one underneath it, which is that using how far apart two populations sit today as a proxy for how likely they were to swap stories assumes people stayed put, and staying put is the one thing human beings have never once reliably done.
There's a wobble underneath all of it that has nothing to do with folklore at all, too. Where you put Proto-Indo-European depends on which origin hypothesis you back, and that argument is still live among the people who study it for a living. The steppe version gives you 5,000 to 6,000 years. Back the Anatolian one instead and the node sits considerably deeper, which would make the tale older rather than younger, and the authors themselves pointed out that the deeper timeframe overlaps with the earliest archaeological evidence for smelting metal, so a smith could just about have existed to have the story told about him. The age of a fairy tale is partly hostage to an unsettled question in a completely different discipline, which is either delightful or maddening depending on what you wanted out of your afternoon.
So, how contested is it? The method is legitimate. The paper is careful and the correlations in it are real. The specific number is an estimate resting on an estimate, and the honest version is that "The Smith and the Devil" is very old, plausibly Bronze Age, and that 6,000 is the ceiling of a range which several serious people think is too generous.
The index that made any of this countable
None of it works without the catalogue underneath. The catalogue is a strange and faintly obsessive object.
Antti Aarne, a Finnish folklorist, published the first version in 1910. Stith Thompson translated and expanded it in America in 1928, then revised it again in 1961. Hans-Jörg Uther overhauled it in 2004, which is where the third letter comes from and why you'll see ATU rather than AT. What sits inside is roughly 2,400 numbered international tale types, sorted into ranges: animal tales at the front, Tales of Magic from 300 to 749, then religious tales, realistic tales and jokes.
The premise sounds unhinged right up until you look at what it caught. Somebody decided that "the story where a girl's stepmother sends her out to die and an animal helps her" is a countable object, assigned it a number, and then found it again. And again. And again. Collect stories across societies with no contact, no shared literature and no plausible route between them, and the same small set of structures keeps coming back with local furniture bolted on: different animal, different forest, same shape.
The index is not neutral and nobody should pretend it is. It's a European scholarly instrument with European gaps, and it flattens traditions that don't fit its slots by filing them under the nearest thing that does. Folklorists have been saying so for decades. They're right. It's also the only reason anyone could put the question to a computer at all, because you cannot count what nobody has numbered.
(Type 510A is Cinderella, which you'd expect. Type 510B is the one where the girl runs because her father has decided he wants to marry her, and it has been collected in enough places across enough centuries to have accumulated its own scholarly literature, and nobody read that one to you at bedtime for reasons I think we can all work out on our own, though somebody somewhere did read it to a child, repeatedly, in enough separate villages and separate centuries that a Finn in 1910 could give it a number and a German in 2004 could still find it sitting where he left it.)
The ones that didn't make it
Look at the 2016 result from the other end. Two hundred and seventy-five tale types went in. One came out of the deepest node with support the authors were willing to stand behind.
One.
Which means the overwhelming majority of stories people told each other didn't survive the trip, and the model can't see a single one of them, because a tale that died left nothing behind to reconstruct from. Everything in that index is a survivor. It's a census of winners. It tells you nothing about the losers except that there were a great many of them.
So the question worth carrying into the sixteen articles below isn't the one everybody asks first. "Did it happen" gets answered fast, and the answer is usually no, and the answer being no has never once slowed a story down. The more useful question is what the thing was doing for the person who bothered to repeat it, because repeating a story costs time and breath and a certain amount of your standing with the listener, and nobody spends all three for free. What's linked from here is mostly a set of answers to that second question, arranged by how much of each story survives contact with the written record, which varies more than you'd expect.
The cryptids, and what each one turns out to be made of
Start with Mothman. Cleanest case on the site. Four people in a car outside Point Pleasant, West Virginia, in November 1966, a grey thing with red eyes, and a bridge that came down thirteen months later and got welded onto the sightings afterward, largely by a book that arrived years after both. Two separate events. Everything that makes Mothman feel like a warning got attached later, by people who needed it to have been one.
The Loch Ness Monster is the opposite arrangement, where the famous evidence is a hoax that came apart in 1994 and the boring evidence is a survey of the DNA floating in the water. The kraken is the rare case where the monster turned out to be real and considerably smaller than advertised: a Norwegian bishop writing about something a mile and a half around, and an actual giant squid that nobody managed to photograph alive until 2004. The chupacabra shows up in Puerto Rico in 1995, and the goat-sucking part of it is 23 centuries older than that, which is a fact I'd have framed if it were mine.
Bigfoot and the yeti are the two where people keep testing things, and the results have gone in opposite directions. Bluff Creek, California, August 1958, sixteen-inch casts pulled out of the mud, and then a film in 1967 that nobody has authenticated and nobody has broken either. The yeti's Western career starts with a mistranslation in 1921, and every hair and tissue sample anyone has run has come back as bear. Every one.
The Jersey Devil is a family feud that got badly out of hand, involving the real Leeds family, the Quakers, Benjamin Franklin, and then one week in January 1909 when the newspapers assembled the modern monster more or less from scratch. The Michigan Dogman has the most thoroughly documented origin of anything on this page: a disc jockey wrote a song as an April Fools' joke in 1987 and the sightings came back at him afterward.
Two of the ten aren't the same category of thing, and the articles handle them accordingly. The wiindigoo is a living teaching in Anishinaabe and Cree tradition about greed and taking more than your share, tied to winter and famine, and the antlered deer skull you have seen a thousand times isn't traditional at all.
Nope. That one arrived later, through fiction and film.
And the skinwalker article deliberately gives you less than other sites will. Yee naaldlooshii is a Diné term, many Diné people consider the subject inappropriate to discuss with outsiders, and that piece says so plainly and then stops. That's the correct length for it. Not a gap somebody forgot to fill.
The legends, where the record is usually somebody's actual life
Werewolves is the one I'd hand a person first. No contest. There were real trials and real executions, the confessions were extracted under torture, and the full moon and the silver bullet were invented by Universal Pictures in the 1940s. The oldest part of that story has a body count. The part everyone knows is about eighty years old.
Ugh.
Slender Man went from a Photoshop thread in June 2009 to a legend cycle folklorists study seriously, and then to Waukesha, Wisconsin in 2014. That article handles the 2014 case carefully, because it involves a twelve-year-old who was stabbed and survived, and two twelve-year-olds who were seriously unwell.
The banshee announces a death and does not cause one, and standing behind her are the mná caointe, the keening women, who practised a paid professional craft and whom church synods banned over and over from 1660 onward. La Llorona is a living legend in Mexican and Chicano communities, still told, still frightening, and she has a documented job attached to her, which is keeping children away from fast water. Where her origins came from is a genuine scholarly argument with real disagreement in it, and the article walks through what's contested rather than picking the tidiest version and presenting it as settled. Kitsune are foxes serving as messengers of Inari Ōkami, and the single most common error to fix is that Inari is not a fox. Inari is a kami. The foxes work there. Baba Yaga travels in a mortar rather than on a broom, her name first appears in print in 1755, and the endlessly repeated claim that she's a demoted ancient goddess has nothing underneath it. No evidence at all.
How to read one of these without getting taken
Four habits. They'll carry you through almost anything on this subject.
Find the first date anybody wrote it down, then work out how much of what you thought you knew arrived after that date. The answer is usually most of it. The full moon, the deer skull, the silver bullet and the harbinger framing were all bolted on late, and every one of them is now load-bearing in the version you happen to be carrying around.
Ask who did the naming. An unsettling amount of the time it's a newspaper copy desk on deadline, and the name is the reason you've heard of it at all.
Ask whether the two halves were connected at the time. A creature appears and a disaster follows. The question is always whether anybody linked the two before the disaster or only afterward. Afterward is not a prophecy. It's a person going back through a bad year with a highlighter.
And ask whether it belongs to living people, because six of these sixteen do. Anishinaabe and Cree, Diné, Irish, Japanese, Slavic and Mexican traditions aren't a bestiary and the people holding them aren't gone. Where a tradition sets a limit on what outsiders get told, that limit is part of the tradition and not an obstacle sitting in front of it, and treating it as an obstacle is how most of the internet ended up publishing things it had no business publishing.
Anyway, the two kids on the concrete
They're not preserving anything. They'd be insulted by the suggestion. Rightly. One of them would tell you so at volume.
What they're doing is a thing that's fun and slightly rude, requires a partner, and gets faster until somebody fails. It has come through intact across generations, county lines and the entire arrival of television, and the reason is that it works this afternoon, for them, which is the only timescale a seven-year-old has ever operated on.
Nobody is maintaining it. It maintains itself by being useful to a seven-year-old with ten free minutes and one available friend.
The sixteen stories linked above are the same machine at longer range and much higher stakes. Nobody kept the wiindigoo teaching going out of antiquarian feeling. La Llorona didn't last because anybody found her charming, either. Somebody had children and an irrigation ditch running fast behind the house and needed those children frightened of it by Tuesday, and it worked, so she told it again the next year, and so did everybody else who'd heard it and had a ditch of their own to worry about, and that is the entire distribution mechanism, running on nothing but the fact that it kept working, for centuries, with no committee, no archive, no funding body and nobody at any point checking whether the woman in the story had ever existed.
A monster is what a fact looks like before anybody could check it, and the ones still standing are the ones that were holding something up.
