An atlas of recurring story
For thousands of years, humans who never met told the same stories.
A child of miraculous birth. A descent into the land of the dead. A serpent coiled around something precious. These patterns surface in scriptures and songs separated by oceans and millennia — from a Babylonian tablet of ca. 1,700 BCE to oral traditions first written down in living memory.
Coincidence, contact, or something structural in the human mind? This atlas is an attempt to ask that question carefully — with a fixed corpus, evidence-grounded extraction, and tests that give chance every opportunity to win.
The corpus
The library of strangers
The atlas reads 338 public-domain texts from 106 traditions: Gilgamesh and the Book of the Dead, the Upaniṣads and the Eddas, the Popol Vuh and the Kalevala, Hopi and Zuni and Inuit tellings, Sufi poetry and Gnostic gospels. Every tradition keeps its own voice; nothing is paraphrased into sameness before the comparison begins.
From these pages, 60,521 passage-level extraction records were drawn, each one anchored to quoted evidence in the source text — 290,891 candidate motifs in the raw, later gathered into 64 canonical families.
338 texts
106 traditions
60,521 evidence-anchored records
64 motif families
The method
Three readers, and a jury of chance
Seeing patterns is easy; the human mind was built for it. So the atlas does not trust a single reader. 6 independent model readers extracted motifs from the same passages, and their answers were compared — agreeing on the canonical family 55% of the time on average across reader pairs.
Every motif claim must point at quoted lines in the source. And every pattern the atlas reports is put before a jury of chance: shuffled null models that ask, if the motifs were dealt at random, how often would this arrangement appear on its own? Patterns that chance can produce are discarded — including some of the atlas's own early favorites.
The full apparatus — agreement matrices, permutation counts, per-lineage tables — is laid out in the Lab.
6 independent readers
55% mean family-level agreement
200 permutations per null test
First finding
Not many webs — one
When families of motifs are joined wherever they recur together across unrelated traditions, the expectation was archipelagos: a hero-cluster here, a creation-cluster there. Instead the 346 conserved edges close into a single connected constellation of 28 families. Descent touches initiation touches sacred knowledge touches the flood. There is one web.
The jury of chance was asked. Across 200 shuffles, random dealing produced 325 conserved edges on average — never 346 (z = 6.81, p = 0.00498).
Second finding · preregistered · the crown
Two worlds that never met, one architecture
The sharpest test the atlas has run. Split the corpus into the connected Old World and the lineages that never traded a word with it — the Dreamtime, the Arctic, the Amazon, the Americas. Let each world sort its own motifs into families blind, from nothing but their likeness to one another — no reference to the atlas's families, or to the other world. Then ask whether the two independently-drawn webs of story share a shape.
At the preregistered resolution the isolated world's bonds re-form in the connected world's web 52.4% of the time, where chance manages 36.6% — and the observed rate beat all 500/500 shuffles, at every resolution tried. The two webs are, by their degree structure, the same kind of shape (KS p = 0.227). The preregistered verdict: STRONG
Full method, the 24 named cross-world bonds, and the preregistration are on the Crown page. An earlier mapping-based version of this test is preserved in the Lab with its leakage flag — superseded by this design, and kept on the shelf, because the atlas keeps its corrections next to its findings.
52.4% of bonds re-form across worlds
36.6% expected by chance
500/500 permutations beaten
×1.43 lift over the null
Third finding · a falsification
The grammar that mostly wasn't
Not every beautiful hypothesis survives its jury. Campbell's claim was that stories don't just share ingredients — they share an order. The atlas asked the texts: 72 strong before-and-after orderings emerged, and under within-text permutation 61 of them dissolved — chance writes those orderings on its own. As a universal grammar, the hypothesis failed.
But 11 lines of order held. The strongest: sacred knowledge → descent — the wondrous child comes before the sacred fire, 91.4% consistent across 186 texts, an order chance matched in only 1% of shuffles. The atlas keeps its falsifications on the same shelf as its findings; the full ledger of what dissolved and what held is in the Lab.
Two further folios
The shared, and the singular
Two ledgers follow from the same indexes. The most shared stories of humankind ranks the motif families carried by the greatest number of traditions — spread, not volume — and marks the families whose partnerships re-formed among peoples who had no contact. Its mirror, what each culture alone remembers, inverts the question: for every tradition, the motifs that it alone — so far — attests.
The descent
The one web is walkable.
From the One to its 28 web-forming families (of 64 in all), down through sub-families to single motifs, each resting on quoted passages in the books themselves. Every claim above is a path you can follow to its evidence.