Flood Legends Around the World

"In Africa, including Egypt, native legends of a great flood are conspicuously absent; indeed no single clear case of one has yet been reported." — Sir James Frazer, 1918¹

"Stories of a similar kind originally obtained at a missionary station, or from some godly traveller, get, in course of time, so mixed up and metamorphosed by heathen ideas, that they look exceedingly like native traditions." — Robert Moffat, Missionary Labours and Scenes in Southern Africa, on discovering that his informant's "ancient" flood tradition came from a missionary named Schmelen²

"Did you know that all of these tribes had historical traditions describing the great flood recorded in Genesis? … Almost every tribe of which we have real historical data has a memory of the great flood." — Nick Liguori, Answers in Genesis, 2021³

Summary

Flood stories really are everywhere.  Isaak's public compilation holds 262 of them from nine world regions;  Nick Liguori's 2021 collection holds over 300 from the Americas alone.  That much is not in dispute, and it is a fact that wants explaining.

What is in dispute is what the stories say.  Apologetics writing has quoted the same set of percentages for ninety years — 95% of flood legends describe a worldwide flood, 88% save a favoured family, 70% use a boat, 67% take animals aboard, 35% send out birds.  Those numbers are real, and this article traces them to their source, which nobody quoting them names:  a hand-drawn chart on page 169 of Byron Nelson's The Deluge Story in Stone (1931).  Nelson tells you, in the same appendix, that the chart is "based on the legends recorded in this book" — a sample he chose himself, from legends he selected because they resembled Genesis.

So the percentages were recomputed here from a corpus nobody selected for resemblance.  The script, the corpus, and every row of output are published beside this article, and they run in a second.  The results:

Five findings, and they do not all run the same way.

1. The motifs really do travel in packages.  Deal the eleven elements out at random, each at its own observed rate, and the spread of "how many elements per story" you get is nothing like the real one.  The real corpus has 37 stories with none of the eleven where chance predicts 14, and five stories with nine or more where chance predicts none at all.  Observed variance is 2.2× the independence null.  Something is being transmitted.  This is the strongest thing the data say for the traditional argument, and it is rarely stated this precisely by the people making it.

2. But the package is not the Genesis package.  The elements split into two clusters that barely talk to each other.  A boat-rescue cluster — warning, vessel, animals, birds, family — holds together (pairwise correlations 0.19 to 0.36).  A mountain-refuge cluster — flood covers the world, survivors climb to high ground, repopulate — holds together separately, and is uncorrelated with the boat (mountain-and-vessel correlation is 0.00).  The world's most common flood story is people fleeing uphill, not people building a ship.  Genesis is the boat type, and the boat type is the minority.

3. Nobody has an uncontaminated legend that matches Genesis in full.  Five of the 262 carry all five diagnostic elements at once, and one of those five turns out to be a coding artifact — the Tinneh bird is a duck diving for mud, not a scout.  That leaves four:  Hebrew, Assyrian, Masai, Tarascan of Michoacán.  Two are the Mesopotamian source cluster.  The other two are on the list Frazer drew up in 1918 of stories he judged to show "clear traces of late missionary or at all events Christian influence."  Not one full-strength Genesis parallel comes from a group with no missionary in its history.  And the bird-scout motif itself, outside the Near East, is largely a post-conquest Mesoamerican cluster — seven of its fifteen strict occurrences are Mexican traditions recorded through Spanish colonial channels.

4. Nelson's own decay test fails.  His fourth argument was a prediction:  a story degrades with retelling, so the farther from Ararat, the less Genesis-like the legend.  Regressing motif score on distance from Ararat across 262 legends gives r = −0.10 — the right sign, but tiny.  Drop the nine Near Eastern entries, which are the source of the story rather than a test of it, and the correlation is r = 0.000, exactly.  Outside Mesopotamia, distance from Ararat predicts nothing.  Africa is 5,000 km from Ararat and has the corpus's weakest flood traditions;  the Pacific Islands are 15,000 km away and have some of the strongest.

5. The one thing everyone gets wrong about Africa is wrong in both directions.  Frazer's flat "no single clear case" is too strong — Hermann Baumann showed in 1936 that flood stories are spread in pockets across sub-Saharan Africa, and modern surveys agree.  But the corpus still puts Africa at the bottom on every Genesis-diagnostic element:  6% mention divine judgment against the corpus's 30%, 12% a vessel against 34%.  African flood stories exist;  they are mostly not this story.  A calabash bursts, a pot shatters, a snake is offended.

What secular scholarship now thinks, and why it is not what a skeptic assumes.  The standard critical position is not that flood myths are independent local inventions.  Julien d'Huy's phylogenetic analyses of Yuri Berezkin's motif database recover a real tree with real ancestry, and reconstruct the core of the flood tradition to Southeast Asia, spreading with human migration.  He states plainly that "most of the myths of the flood around the world do not appear to be related to the Christian or Muslim traditions."  The mainstream and the apologist agree that these stories descend from something.  They disagree about what, and where, and when.

Where the traditional argument is at its best.  Patrick Nunn and Nicholas Reid published 21 Australian Aboriginal stories in 2016 that describe named stretches of coast being drowned, and matched each against the postglacial sea-level curve.  Every one requires a memory at least 7,250 years old.  Oral tradition can carry a real geophysical event across hundreds of generations;  that is now a measured result, not a hope.  This is a genuine win for the principle behind the flood-legend argument.  It is also a problem for the chronology that uses it, since those same memories are older than a young-earth Flood and, in several cases, older than a young-earth creation.

Verdict.  The bare fact of worldwide flood traditions is real evidence for something, and the clustering statistics show it is not noise.  But the specific argument — that hundreds of independent peoples preserve a recognisable memory of Noah — does not survive contact with a corpus its advocates did not assemble.  The percentages are wrong, the diagnostic package is missing from every uncontaminated case, and the movement's own falsifiable prediction, distance-decay from Ararat, comes back flat.  Anyone who wants to use this argument should use the weaker and truer version:  humanity remembers water.

Every source is catalogued with what it says and which way it cuts in a companion page:  the literature.

The argument, stated fairly

Put at its strongest, the argument runs:

1. Flood traditions are found on every inhabited continent, among peoples who had no contact with one another.
2. They agree on specific, arbitrary details — a warned survivor, a vessel, animals aboard, birds sent to test the water, a mountain landing.
3. Independent invention will not produce that agreement. Neither will missionary diffusion, because the traditions were often recorded before missionaries arrived, and because the details vary in ways a copied text would not.
4. Therefore the traditions preserve a real event, and Genesis, being the fullest and least corrupted version, is the original.

Every step of that is checkable, which is why it is worth checking.  Nobody should mistake the following for hostility to the conclusion.  Two of the four steps survive.

Where the percentages come from

The numbers are quoted almost everywhere:  Apologetics Press, Answers in Genesis, Institute for Creation Research, and a thousand sermons.  Apologetics Press gives the fullest form and, to its credit, a citation:⁴

"In 95 percent of the more than two hundred flood legends, the flood was worldwide;  in 88 percent, a certain family was favored;  in 70 percent, survival was by means of a boat;  in 67 percent, animals were also saved;  in 66 percent, the flood was due to the wickedness of man;  in 66 percent, the survivors had been forewarned;  in 57 percent, they ended up on a mountain;  in 35 percent, birds were sent out from the boat;  and in 9 percent, exactly eight people were spared."

That is quoted from James Perloff, Tornado in a Junkyard (1999), p. 168.  Perloff's chart is adapted from Byron C. Nelson, The Deluge Story in Stone (Augsburg, 1931), Appendix I, Figure 38, p. 169.⁵

Nelson's book is a lending-only scan, but its text is searchable, and Appendix I says exactly what the chart is:

"The accompanying chart (Fig. 38, page 169), which is based on the legends recorded in this book, shows graphically the points of similarity between all the Deluge tales."⁵

And what the book contains is stated in the preface:

"Flood traditions exist and have existed in large numbers outside the Hebrew nation, and as these unmistakably substantiate the Biblical account, the writer has thought it well to offer the reader a fair sample (by no means all) of these traditions."⁵

So the denominator is a sample Nelson picked, from a book whose stated purpose is that the traditions "unmistakably substantiate the Biblical account."  Selecting cases on the outcome you are measuring inflates every percentage you then compute.  This is not a subtle statistical point and it is not an accusation of dishonesty — Nelson says what he did, in print, on the page the chart sits on.  It is simply that nobody quoting the numbers ninety years later has read that far.

One more sign of the sample's size:  Nelson asks, on the same page, "How, for example, is it to be explained that in seven widely scattered regions (see chart) some animal, usually a bird, is sent out to discern the state of dryness of the earth?"⁵  Seven regions is the whole bird-scout column of the chart.  A percentage of 35% with a numerator of seven implies a denominator of twenty.  The "more than two hundred flood legends" of the modern quotation is not Nelson's denominator at all;  it is a later number stapled onto Nelson's proportions.

Recounting, on a corpus nobody chose

To test the percentages you need a corpus assembled by somebody with no stake in the answer.  Mark Isaak's Flood Stories from Around the World is the obvious candidate:  262 summarised stories in nine regions, compiled 1996–2002, with a stated collecting rule — "collect every flood story I find" — and full source citations.⁶  Isaak is a critic of creationism, which cuts the other way, but his selection rule is the opposite of Nelson's:  he includes borderline cases rather than excluding them, and he keeps stories in which no one survives, no boat appears, and no god is involved.

Eleven Genesis elements were coded on each story by keyword rules.  The script is code-motifs.ts, the corpus is isaak-corpus.json, and the per-story output is motifs.tsv.  Three things about the method, stated up front because they matter:

• The rules are conservative by design. A summary that does not mention birds is scored no-birds, even if the underlying story has them. Every percentage below is therefore a floor, not a point estimate.
• The tool was calibrated before it was trusted. Genesis itself is in the corpus and scores 11 of 11 — the known-good case passes. The Norse Bergelmir story, which is a flood of a slain giant's blood with no warning, no judgment and no birds, scores 2 — the known-bad case fails. Atrahasis correctly scores zero on "human wickedness," because in Atrahasis the gods send the flood over noise, not sin.
• Isaak sometimes merges several tellings from one culture into one entry, and this hits the top of the table hardest. A merged entry scores the union of its tellings, so it looks more Genesis-like than any one telling is. The Hawaii entry merges four stories, one of which says flatly "There was no ship involved." The Greek entry merges six — Apollodorus, Lucian, Hellanicus, the Megarian tale, Ogyges and Plato — and only Lucian's loads animals. The Assyrian entry is Gilgamesh XI plus a Sharur fragment; the Tarascan entry is two separate Michoacán tellings. Every one of the highest-scoring rows in the table is a merged entry, which is not a coincidence: merging is what produces high scores.
• One coding rule conflates two motifs, and the error runs against this article's own conclusion. The bird rule fires when any bird is sent out over the water. Six of its 21 hits are earth-diver stories, in which a duck or beaver dives to fetch mud from which the world is rebuilt — Berezkin's motif C6C, structurally unrelated to Noah's raven and dove, which is his C5A. Batak, Tinneh, Cree, Chippewa, Montagnais and Chitimacha are all earth-diver. Counted strictly, the Genesis-style bird scout appears in 15 of 262 stories, 5.7%, not 8%.

Results across all 262:

And by region, as a percentage of that region's stories:

The Near East leads on almost everything, which is what you expect when the region is the place the story was written down rather than an independent witness to it.  The interesting rows are the others.

The table

Every one of Isaak's 262 stories, coded.  Column keys:  Wor flood is worldwide;  Sin sent for human wickedness;  Wrn advance warning;  Ves a built vessel;  Fam a family survives;  Ani animals taken aboard;  Brd bird or animal scout;  Mtn mountain landing or refuge;  Sac sacrifice after;  Pro promise, covenant or rainbow;  Rep survivors repopulate.  n is the row total out of 11.

Sorted within each region by score, so the top of each block is that region's closest approach to Genesis.  Codings are machine-generated from summaries, so treat a single cell as indicative and the columns as sound.

Near East (9 stories)

Europe (11 stories)

Africa (17 stories)

Asia (55 stories)

Australia (17 stories)

Pacific Islands (17 stories)

North America (81 stories)

Central America (24 stories)

South America (31 stories)

What the table shows

The motifs travel in packages

If the eleven elements had arrived independently in each culture — some invented locally, some borrowed from a neighbour, with no common package — then the number of elements per story should look like eleven independent coin flips, each weighted at its own observed rate.  It does not.

Both tails are far too fat.  There are 2.7 times as many stories with nothing in common with Genesis as chance allows, and five stories with nine or more elements where chance allows none.  Observed variance is 3.95 against a null of 1.80, a ratio of 2.20.

This is real support for the traditional argument, and it should be said plainly.  Flood stories are not a random scatter of independently invented tales that happen to involve water.  Whatever else is true, elements of these stories are being transmitted together.

But the package is not the Genesis package

The next question is which elements travel together.  Correlating each pair of columns across the 262 stories:

Two blocks, not one.

The boat-rescue type.  Warning, vessel, animals, birds, family, and sacrifice cohere:  warning-with-vessel .36, warning-with-birds .35, vessel-with-birds .32, animals-with-birds .29.  Someone is told the water is coming, builds something, loads it, and later sends a bird.  This is Genesis, and it is also Gilgamesh, Atrahasis, Deucalion, Manu, and the Mesoamerican Tezpi.

The mountain-refuge type.  A flood covers the world, people climb to high ground, the survivors repopulate.  Flood-with-mountain .23, mountain-with-repopulate .24.  And crucially:  mountain-with-vessel is 0.00.  Knowing that a story has a mountain tells you exactly nothing about whether it has a boat.

That is a structural fact, not an interpretation.  The world's flood traditions are not one story degrading at different rates;  they are at least two stories with different geographies.  The mountain-refuge type is the commoner one — 103 of 262 stories put someone on high ground, against 89 who put someone in a vessel — and it is the one Genesis does not tell.  Noah does not climb a mountain.  He lands on one.

Nobody has a clean full match

Five stories out of 262 carry all five diagnostic elements at once — worldwide flood, advance warning, a vessel, animals aboard, and a bird sent out.  Here they are, all of them:

Nothing on that list is an independent witness.  Two are the Near Eastern tradition itself.  Three were flagged as contaminated more than a century ago, by a man with no interest in defending Genesis.

Reclassify the Tinneh entry, whose "bird" is a duck diving for mud rather than a scout sent to test the water, and the list is four:  two Mesopotamian texts and two flagged tellings.

Where the bird actually lives.  Strip the earth-diver cases and fifteen stories in the whole corpus send a bird out to check the water.  Two are Hebrew and Assyrian.  Two are the East African pair Frazer judged biblical — Masai and Rukwa.  Four are scattered, and one of those four is a false positive:  the Fijian entry is about a lost bird being searched for, not a bird sent to test the water.  That leaves Altaic, Tlingit, and an unnamed Cascade Mountains tradition recorded by Myron Eells in the 1870s.  The other seven are all Mexican — Tarascan, Michoacán, Cora, Zapotec, Totonac, Tzeltal, Popoluca.  Central America has the highest bird-scout rate of any region outside the Near East, 29%, and Central America is where Spanish friars documented indigenous tradition earliest and most intensively.  Three of those seven are on Frazer's list, and the Popoluca telling does not need a list:  in it, "Christ sent down the hawk and hummingbird and finally came himself."  A motif whose non-Mesopotamian distribution is "colonial Mexico, plus three" is not a global memory.

Loosen the criterion to four of five and eleven more appear:  Greek, Transylvanian Gypsy, Altaic, Sumerian, Cree, Chippewa, Hareskin, Zapotec, Tzeltal, Popoluca, and the Rukwa region of Tanzania.  The Cree and the Rukwa story are also on Frazer's list.  The four Mesoamerican entries were all recorded by or through Spanish clergy after 1521.  The Greek and Sumerian belong to the same Eastern Mediterranean and Mesopotamian literary world as Genesis.  The pattern holds:  as the match to Genesis gets better, the provenance gets worse.

The distinctive elements

The point of a table like the one above is that it flattens everything.  What it cannot show is how strange most of these stories are — and the strangeness matters, because it is what a diffusion-from-Genesis account has to explain away.

Read down the "unique" column and the diffusion hypothesis gets harder, not easier.  A story that had been copied from missionaries would not acquire a self-growing soil, an arrow-rigged vulture, a rain of resin, or a rebellion of household crockery.  Byron Nelson made exactly this point in 1931 and he was right to make it.⁵

Read across the "vessel" and "survivors" columns and the other problem appears.  In the Greek story no animals are saved.  In the oldest Indian version Manu survives alone and has to have a wife manufactured out of dairy products afterwards.  In the Popol Vuh nobody survives at all.  In the Chinese story the flood is not survived but drained.  These are not degraded copies of Genesis;  they are different stories that share water.

Missionary contamination: the honest state of the question

Documented cases exist

This is not a hypothetical.  The best-documented case is worth quoting at length, from the missionary who blundered into it and then published his own mistake.  Robert Moffat, travelling in Namaqualand around 1820, went looking for a flood tradition and found a man who had one:

"Discovering that he possessed some knowledge on the subject, and being an utter stranger to any of the party, and to all appearance a child of the desert, I very promptly took my pen and wrote, thinking myself a lucky discoverer.  I was perfectly astonished at some of his first sentences, and, afraid lest I should lose one word, I appointed two interpreters:  but by the time I reached the end of the story, I began to suspect.  It bore the impress of the Bible.  On questioning him as to the source of his information, he positively asserted that he had received it from his forefathers, and that he never saw or heard of a missionary.  I secretly instituted inquiries into his history, but could elicit nothing."²

Moffat filed the paper away unresolved.  Weeks later, on the road south, he asked for a guide to Bethany — Schmelen's mission station — and a bystander pointed at his informant:  "There is a man that knows the road to Bethany, for I have seen him there."

"The mystery of the tradition was in a moment unravelled, and the man decamped, on my seeing that the forefather who told him the story, was our missionary Schmelen."²

Everything a field worker would use to authenticate a tradition was present.  The informant was a stranger to the party, appeared to have no European contact, gave the story spontaneously, asserted native descent when challenged, and survived a background check.  He was still wrong, and only an accident caught it.

Frazer records a second, more delicate case from the Rukwa region of what is now Tanzania.  The informant "professed to have had it from his grandfather, and stoutly asserted that it was a genuine old tradition of the country and not borrowed from foreigners."  His account has God telling two men to enter a ship with seeds and animals male and female, and a dove sent out that returns, then a hawk that does not.¹  Sincere assertion of native origin is not evidence of native origin, because the informant is usually two or three generations downstream of the borrowing and has no way of knowing.

Nelson answered this objection with an anecdote he thought settled it — a New Hebrides islander interrupting a missionary's reading of Genesis to say "Stop, that is almost the same as what we have been told."⁵  Read again, the anecdote is a first-contact record of a story spreading, not proof it predated contact:  the islander then concludes that the missionary's version "must be more correct than ours," which is precisely the mechanism by which a local tradition gets overwritten.

The correlation, and why it proves less than it looks

Frazer listed ten peoples whose flood stories he judged to carry "clear traces of late missionary or at all events Christian influence":  the Kamars of central India, the Minahassans of Celebes, Hawaiians, Macusi of Guyana, the Michoacán Indians, the Cora, Cree, Tinneh, Eskimo, and Masai.¹  Matching those names against the coded corpus gives eighteen entries.  They average 3.17 of 11 elements, against 2.45 for the other 244;  on the five diagnostic elements, 2.00 against 1.08 — nearly double.

That is a large effect.  It is also partly circular, and the article will not pretend otherwise.  Frazer's grounds for flagging were sometimes provenance ("Moffat traced this one to a missionary") and sometimes resemblance ("the stories are plainly mere variations of the Biblical narrative").  Where the ground was resemblance, using the flag to predict resemblance proves nothing.  A clean test would need contamination flags assigned on provenance alone — date of first recording, mission presence, the recorder's own account of who told him — and nobody has built that dataset.  It is the single most useful thing anyone could do for this question.

What is not circular is the observation in the previous section:  the five full-strength matches are two Mesopotamian texts and three flagged tellings, with no fourth category.  That holds whatever Frazer's reasons were, because it is a statement about the composition of a list.

Where the contamination argument overreaches

Two things are commonly said against the flood-legend argument that are simply not true.

"They were all recorded by missionaries."  Nelson's third objection is sound:  Catlin, Bancroft, Schoolcraft, Dorsey, Curtis, Kroeber and Boas were ethnographers, not evangelists, and Lucian of Samosata — the source for the Syrian Deucalion — was an open scoffer at religion who recorded the story with a sneer.⁵  The corpus is not a missionary artifact.

"Most flood myths are Christian in origin."  The specialists say the opposite.  Julien d'Huy, whose phylogenetic work is the most quantitative treatment of the question, writes that "most of the myths of the flood around the world do not appear to be related to the Christian or Muslim traditions."⁷  The mainstream position is that these traditions are old and genuinely inherited.  The disagreement is about from what.

Testing Nelson's own prediction

Nelson's fourth objection was the only one he stated in a form that can be checked:

"the more ancient a tradition is, and the more close geographically to that central place in Asia where the Bible says the ark landed, the more reasonable and sane is the legend, the more generally like the Biblical account, and vice versa.  It is well known that a story loses in accuracy each time it is told."⁵

That is a decay model, and it makes a quantitative prediction:  motif score should fall with distance from Ararat.  Assigning each of the nine regions a centroid and computing great-circle distance from Ararat (39.7° N, 44.3° E):

Across all 262 stories, r = −0.103, slope −0.057 elements per 1,000 km.  The sign is Nelson's, but the size is not:  travelling the 14,900 km from Ararat to Tierra del Fuego costs 0.85 of one element out of eleven.

Then remove the Near East.  Those nine entries are Genesis, Gilgamesh, Atrahasis, the Sumerian flood story and their neighbours — the tradition itself, not an independent sample of it.  With them gone:

• all 253 remaining stories: r = 0.000, slope 0.000
• excluding Europe as well (242 stories): r = +0.024
• New World only (136 stories): r = −0.011

The gradient is not weak.  It is absent.  Every bit of the apparent decay is the Near Eastern block sitting at the origin of the axis.  Once you are outside the region where this story was written down, how far you are from Ararat tells you nothing whatever about how much your flood legend resembles Genesis.

The two clearest counterexamples are visible in the table.  Africa is nearer to Ararat than anywhere but Europe and the Near East, and has the weakest flood traditions in the corpus.  The Pacific Islands are the farthest place on earth from Ararat and score above the global mean.  A decay-from-a-centre model predicts the reverse of both.

The Africa problem, in both directions

Frazer was categorical:

"In Africa, including Egypt, native legends of a great flood are conspicuously absent;  indeed no single clear case of one has yet been reported."¹

He had gone looking.  Petrie told him nothing survives in Egyptian literature.  John Roscoe, after twenty-five years among the peoples of the Uganda Protectorate, told him he had found no native flood legend at all.  Livingstone, recording a lake-formation story in Angola, called it "remarkable as the only one I have met with in this country."¹

Frazer was too strong, and the correction comes from mainstream folklorists, not apologists.  Hermann Baumann's 1936 survey found flood stories "basically spread, in pockets, all over sub-Saharan Africa, with some variants," and Michael Witzel notes that Baumann was already criticising "the then (as today) prevailing opinion that the flood myth is hardly found in Africa."⁸  Jean-Loïc Le Quellec's 2021 book makes the same correction with more data.⁷  Isaak's corpus carries seventeen African entries.  Anyone repeating Frazer's flat denial in 2026 is a century out of date.

But the shape of the African material is the real point, and it does not help the Genesis case.  On every Genesis-diagnostic element Africa is last or nearly last:  6% mention divine judgment against a corpus average of 30%;  12% have a vessel against 34%;  6% have a mountain against 39%.  Mean score 1.41 of 11, the lowest of the nine regions.  What Africa has instead is a distinct family of stories — the ocean kept in a clay pot that shatters, the "devouring calabash," a wounded lightning-spirit whose death stops the rain.  These are flood stories.  They are not this flood story.

That matters because Africa is where a Babel-dispersal model has the least excuse.  On that model the sons of Noah were in Mesopotamia four thousand years ago and Africa is next door.  The peoples with the freshest memory should be the near ones.  They have the faintest.

What the Chinese actually remember

China is the test case apologetics writing most often gets wrong, and it is worth doing carefully because the Chinese record is long, literate, and continuous.

Frazer's observation still stands as stated:  neither Chinese nor Japanese literature preserves "any native legends of a great flood of the sort we are here considering, that is, of a universal inundation in which the whole or the greater part of the human race is said to have perished."¹

China has a great flood, and it is one of the founding stories of Chinese civilisation — but read what it is.  The Supreme Sovereign sends water as punishment for human misbehaviour.  The flood lasts twenty-two years.  People survive by living in mountain caves and trees.  The hero Gun steals xirang, self-growing soil, from heaven to dam the waters, and is executed for the theft;  his son Yu emerges from his corpse as a horned dragon and finishes the job by dredging channels to the sea.⁶

There is no ark, no favoured family, no pairs of animals, no bird, no covenant.  The flood is defeated by hydraulic engineering, and the story's point is the founding of the Xia dynasty and the legitimacy of public waterworks.  Of the eleven elements it scores judgment, mountain refuge, and nothing else.

Interestingly, there may be a real event behind it.  In 2016 Wu Qinglong and colleagues published evidence in Science for a landslide-dam outburst flood on the Yellow River around 1920 BC, which they argued supports "the historicity of China's Great Flood and the Xia dynasty."⁹  The dating is contested.⁹  But a dated, local, catastrophic river flood is exactly the kind of thing that generates a national flood tradition, and it is not Noah's.

The character 船 does not mean "eight mouths in a boat"

The claim appears on Answers in Genesis, Creation Ministries International, Apologetics Press and in a shelf of books:  the Chinese character for ship, 船 chuán, is built from 舟 "boat" plus 八 "eight" plus 口 "mouth" — a vessel for eight people, a memory of Noah's family.⁴¹⁰

The refutation is a Chinese primary source, written about AD 100, by a man who had never heard of Genesis.  Xu Shen's Shuowen Jiezi, the first systematic Chinese dictionary, analyses the character in the boat section:

船：舟也。从舟，鉛省聲。

"Chuán:  a boat.  From 舟 [boat], with 鉛 [qiān] abbreviated as the phonetic."¹¹

The right-hand element is a phonetic — a sound cue — not a numeral and a mouth.  This is the ordinary structure of the overwhelming majority of Chinese characters:  one part for meaning, one for sound.  The related readings confirm it:  the Jiyun gives 船 an alternate reading 余專切, "pronounced like 沿 yán," which is the same phonetic series.¹¹

So the earliest Chinese analysis of the character, produced 1,500 years before any Christian missionary reached China, says it is boat-plus-sound.  You have to overrule Xu Shen to get Noah out of it.

CMI's published defence does not engage the Shuowen at all.  It argues that the right-hand element "is just a stylized way of writing 'eight'," on the authority of "a Chinese gentleman educated in Taiwan," and that 口 can mean people because "eight mouths to feed" means eight people in English.¹⁰  Neither is an argument about the character's history.  The strongest version of the creationist claim would be that the Shuowen preserves a rationalised later analysis and the oracle-bone form shows something else — but 船 is not among the securely attested oracle-bone graphs, and the Chinese Text Project's script gallery for it shows seal and bronze forms only.¹¹

This argument should be dropped.  It is not close, the primary source is free and online, and continuing to run it costs credibility that the better arguments in this article need.

Where the argument is strongest

Oral tradition really can carry a real event for millennia

The most important recent work on this subject was not done by either side of the creation debate.

Patrick Nunn, a geographer, and Nicholas Reid, a linguist, collected Australian Aboriginal stories that describe specific stretches of coastline — named islands, named reefs, named bays — as having once been dry land where people hunted.  For each of 21 locations they computed the water depth that would have to be exposed for the story's details to be literally true, and compared it against the Australian postglacial sea-level curve.¹²

Every one of the 21 requires a memory older than 7,000 years.  The minimum ages run from 7,250 cal BP (Brue Reef, Oyster Harbour) to 11,730 (Cape Chatham);  maxima reach 13,310.  Their conclusion is that these traditions "have endured longer than the notional limits (500–800 years) for orally transmitted knowledge," and that they "provide empirical corroboration of postglacial sea-level rise."¹²

This is a genuine and underused point in favour of the principle behind the flood-legend argument.  The stock objection — "oral tradition cannot preserve anything real past a few centuries" — is now falsified by measurement.  A memory of a real inundation can survive a hundred generations.

And it comes with a bill.  The same paper dates those memories to between 7,250 and 13,310 years ago.  On the Ussher chronology the Flood is about 4,375 years ago and the creation about 6,030.  A young-earth reader who leans on Nunn and Reid to establish that flood memories are durable is leaning on a paper whose central measurement will not fit inside his timeline — several of the stories are dated older than his creation.  There is no way to take the durability result and leave the dating result;  they are the same calculation.

Nunn and Reid also draw a distinction worth keeping.  Their stories describe a permanent drowning — the sea came and stayed.  The world's flood myths, they note, "involve a flood/tsunami that eventually recedes, restoring the Earth to the condition it was in before," and appear "to have diffused from a few sources."¹²  The Australian coastal-inundation memories and the global flood-myth corpus are, on their reading, two different things.

What the specialists actually reconstruct

The critical literature is more interesting than the popular version of it.  Yuri Berezkin's Analytical Catalogue of World Mythology codes over 2,200 motifs across more than 900 peoples, and Julien d'Huy has run phylogenetic analyses on its flood section three separate ways.⁷

The findings:

• The flood tradition has real ancestry. The trees have signal; the retention index of 0.52 in the first study indicates partial package transmission — the same conclusion the recount here reached independently from the overdispersion test.
• The reconstructed centre is Southeast Asia, not Mesopotamia, with spread into Eurasia and the Americas along the routes of human migration. The tree resolves a first migration to South America and a second, later, northern one into North America — a structure recovered from myth motifs that matches the structure recovered from genetics.
• The bird-scout motif, Berezkin's C5A, is a real and ancient global motif in its own right — "birds or humans later changed to birds are sent to explore the earth." It is present in Australia. Noah's raven and dove are one instance of a very widespread narrative device, not a fingerprint.
• Africa is the outlier in the tree as well as the count, which is one of d'Huy's reasons for putting the origin east of it.

Anyone arguing this question should know that the serious secular position is not "these are unrelated stories."  It is "these are related stories whose relationships have been mapped, and the map does not run through Ararat."

Which real flood, then?

If some of these traditions remember something, several candidate events are on the table.  None of them is a global flood, and the honest thing is to say which of them the evidence actually supports.

The Mesopotamian flood layers.  Leonard Woolley sank shafts at Ur in 1929 and found up to eleven feet of clean, water-laid silt above Ubaid pottery.  He announced he had found Noah's flood.  He had not.  Samuel Noah Kramer's summary of the state of the question is blunt:  "his Flood deposits at Ur go back to at least 3500 B.C., and this is far too early for the Ziusudra Deluge."  Flood layers at Kish and at Shuruppak — Ziusudra's own city — sit above Jemdet Nasr material, around 3000 BC, and do not correlate with Ur's.  Max Mallowan's 1964 review, Noah's Flood Reconsidered, concluded that the Mesopotamian tradition rests on "an actual catastrophic but by no means universal disaster" around 3000 BC, leaving traces at Kish and Shuruppak.¹³  Different cities, different centuries, different silts.  That is what a river valley with endemic flooding looks like.  It is not a single global event, and the excavators' own stratigraphy is what killed the idea.

Postglacial sea-level rise.  Sea level rose roughly 120 m from the last glacial maximum, drowning continental shelves worldwide over several thousand years.  Nunn and Reid's Australian data show this event is remembered.  It would have been slow — centimetres a year at peak — but it was universal, it was inexorable, and it took land that people lived on.  If anything is the shared substrate under the world's flood traditions, this is the best-evidenced candidate.

Local catastrophes.  Frazer's account of Baghdad in 1849, when the Tigris rose 22½ feet and the city survived as an island in an inland sea until a month later, is worth reading in full.¹  Twelve thousand of seventy thousand inhabitants died of fever from the standing water.  Sir Everard Im Thurn's caution about Guiana is the same point from the other end:  an Indian's account of "the highest flood which covered all the small world known to him" is easily heard, by a European "theologically prejudiced in favour of Noah," as an account of a universal deluge.¹

The Black Sea.  Ryan and Pitman's 1997 proposal that the Mediterranean broke catastrophically into a low-lying Black Sea basin around 7,600 years ago is often cited as the flood behind the flood.  It has not held up well;  Giosan and colleagues concluded in 2009 that the reconnection was gradual and that the shelf was already partly flooded.¹⁴  It should be cited, if at all, as a contested hypothesis and not as a result.

Objections and responses

"You used a hostile compilation."  Isaak's collecting rule is stated and permissive:  every flood story he found, including borderline ones.  Nelson's rule is stated and selective:  a sample of traditions that "unmistakably substantiate the Biblical account."  Of the two, the permissive rule is the correct denominator for asking what fraction of flood legends share a feature.  If someone builds a larger unselected corpus — Liguori's 300-plus American traditions would be a fine start, if released as data rather than prose — the recount should be redone on it, and this article will change if the numbers do.

"Your keyword coder undercounts."  Certainly it does, and by design;  every rule errs toward scoring zero.  That is why the article calls the percentages floors.  But undercounting cannot rescue the comparison, because it applies equally to every column and every region — and the shape results, which is where the argument actually turns, do not depend on the levels at all.  The two-cluster structure, the zero mountain-vessel correlation, the flat distance gradient and the composition of the five full matches would all survive a uniform doubling of every rate.

"Frazer is a century old."  He is, and where he is wrong this article says so — his Africa denial in particular.  He is used here for three things:  a primary corpus of flood traditions with citations to the ethnographers who collected them, a contamination list drawn up in 1918 for reasons unrelated to any modern argument, and documented cases of contamination in the field.  For the modern state of the question the article uses Baumann via Witzel, Berezkin, d'Huy, Le Quellec, and Nunn and Reid.

"Similarity in the details is too specific for chance."  Agreed — and the overdispersion test above quantifies it.  The question is not whether transmission happened but what was transmitted.  The answer the data give is a boat-rescue narrative package and a separate mountain-refuge package, both older and more widely spread than Genesis, with the specifically Genesis-shaped full combination appearing only in Mesopotamia and in tellings recorded after contact.

"Missionaries could not have reached all these places early enough."  For the Americas this is simply not so.  Every American tradition in either corpus was written down after European contact, and in most cases long after:  Liguori's own earliest United States source is John Josselyn's 1663 voyage, by which time New England had been evangelised for a generation, and most of his citations are nineteenth- and twentieth-century ethnography.  For Australia, the Pacific, and interior Asia the objection has real force, and those are the places where the flood traditions are least like Genesis, which is exactly what the contamination hypothesis predicts.

"What about the Iroquois who commemorated the 4,000th anniversary of the flood?"  Liguori reports an old Iroquois man in about 1911 describing his people's commemoration of the flood's 4,000-year anniversary on Mount Marcy, and observes that 4,000 years after 2348 BC lands in AD 1653.³  The arithmetic is right and the datum destroys the argument it is offered for.  An absolute year-count is not the kind of thing oral tradition carries;  it is the kind of thing you get from Ussher's published chronology.  A tradition that knows the date of the Flood in Ussher's reckoning has demonstrably been in contact with Ussher's reckoning.  The provenance compounds it:  the source is a 1912 newspaper column by Henry W. Shoemaker, whose folklore collecting is the standard academic example of what Richard Dorson named "fakelore," and whose methods are the subject of a full critical study by Simon Bronner.¹⁵¹⁶  Nothing should be built on it.

"Then you have conceded the flood."  No.  The article concedes that flood traditions are widely shared, that their elements are transmitted rather than independently invented, and that oral tradition can carry a real event for millennia.  It denies that the shared package is the Genesis package, that the surviving parallels are independent, and that the distribution looks like radiation from Ararat.  Those are separable claims and the evidence separates them.

Verdict

Against the argument as usually made.  The percentages in circulation are a hand-drawn 1931 chart of a self-selected sample, repeated for ninety years without anyone opening the book.  Recounted on an unselected corpus they fall by a factor of two to four.  The single most Genesis-specific element, the bird sent out, appears in 8% of flood legends — 5.7% once earth-diver stories are separated out, as they should be — not 35%.  Every legend in the corpus that carries the full Genesis package is either Mesopotamian or was flagged as Christian-influenced a century ago.  The movement's own testable prediction — decay with distance from Ararat — measures exactly zero outside the Near East.  Africa, next door to Ararat, is the region with the weakest flood traditions on earth.

For the argument's underlying instinct.  The stories are not independent inventions;  the clustering statistics rule that out at better than 2:1 against the independence null.  Something is being transmitted, and the specialists who have mapped it agree it is old — older, on d'Huy's reconstruction, than any historical religion.  Oral tradition demonstrably carries real geophysical events across seven to thirteen thousand years, which is a measured fact and a good one.  And the sheer strangeness of the individual traditions is a genuine problem for anyone who wants to explain the whole corpus as missionary echo.

What should be retired:  the 95%/88%/70% table, in every form;  the claim that the traditions "were recorded long before any missionaries arrived," which is false for the entire Western Hemisphere;  the Chinese character 船;  and the Iroquois 4,000-year anniversary.

What can be kept:  that flood traditions are near-universal;  that they share a transmitted narrative package;  that Genesis is not derivative of Gilgamesh in any simple textual sense, which is argued at length in Noah's Ark:  The Logistics;  and that the durability of oral memory is now an established result rather than a hope.

The honest form of the argument is smaller than the popular form and considerably harder to answer.  Humanity remembers water.  Whether it remembers this water, the flood legends by themselves cannot settle.

Method and data

Everything computed above re-runs from three files beside this article:

• isaak-corpus.json — Isaak's 262 stories, extracted from his HTML with region, culture, summary text and source citation.
• code-motifs.ts — the coder. Every rule is a readable regular expression with a comment saying what it is trying to catch. deno run --allow-read code-motifs.ts.
• motifs.tsv — one row per story, eleven 0/1 columns, plus totals.

Anyone who thinks a rule is wrong can change it and see what moves.  That is the point of publishing it.

Sources