The flood story is built like a mirror

Picture someone walking into a snowfield and stopping. Then they turn around and walk out, placing each foot back in its own print. The flood story in Genesis is written like that.

It starts with Noah. Then God announces a flood, and Noah waits seven days. Forty days of rain follow, and the waters rise for 150 days. Then, at the exact middle, comes one line: "God remembered Noah."

From there the story walks back out. The waters fall for 150 days, and forty days pass. Noah waits seven days, God promises the flood will never return, and the story ends with Noah again.

Bible scholars call this shape a chiasm, after the Greek letter chi (X), whose two strokes cross in the middle. Gordon Wenham laid out this reading of the flood in 1978.¹1 G. J. Wenham, "The Coherence of the Flood Narrative", Vetus Testamentum 28 (1978), 336-348. His outline has 15 matching pairs around a centre at Genesis 8:1.

If he is right, the turning point of the most terrifying story in Genesis is not the storm. It is God's faithfulness to the family he saved.

It is a beautiful reading. It is also the kind of pattern an eager reader can find almost anywhere, so we asked a careful question: is the mirror really there in the words of the Hebrew, or only in the eye of the beholder?

A fair test for a mirror#

Think of a drawer of odd socks. If someone tells you they are really pairs, you can check: each sock should match its own partner better than it matches any other sock in the drawer.

We did the same with the flood. If Wenham is right, each part of the story on the way in should share more words with its own partner on the way out than with any other part. Rare words count for more than common ones, because two passages both mentioning "raven" tells you more than two passages both saying "and".

Then came the harder step. Neighbouring passages in any story share words, simply because they are about the same thing. So we took Wenham's layout and laid it over 1,000 other stretches of Hebrew storytelling, to see how often an ordinary story scores well by accident.

We wrote the whole test down, with its pass mark, before running it. That rule matters more than any other here: it stops anyone, us included, from adjusting the test until it says what they hoped.²2 Each test in this programme was registered in writing, with its pass mark, before its code was written.

What we found#

The flood's pairs beat every one of the 1,000 other stories. That is the strongest result this test can give.³3 Wenham's 15 pairs score z = 4.26. Across the 1,000 comparison stories the median is 0.60 and the 99th percentile 3.09, so the calibrated p is 1/1,001, the floor for 1,000 comparisons.

It is not just the famous day counts doing the work. Take out the 7, 40 and 150 days and the remaining pairs still stand well clear of ordinary stories.⁴4 Without the four day-count pairs the other eleven score z = 3.63 (p = 0.004).

And the mirror order matters. Pair the very same pieces front to back, first with first and last with last, and most of the effect disappears. The story was not just written about the same things twice. It was folded.⁵5 Paired in plain text order, the same elements score z = 1.09 (p = 0.072).

Not every pair is strong. Six of Wenham's partners share more with each other than with anything opposite them. A few are linked by theme rather than wording, such as the LORD shutting Noah in and Noah opening the window, which share only the word "which". A word test cannot see those links, so it is silent on them.

Kept faithfully for two thousand years#

The Samaritans have kept their own copies of the five books of Moses since their community parted from the Jewish one, around the second century BC. Their flood story differs from the standard Hebrew in 95 words. The mirror survives in their copy almost unchanged.⁶6 Samaritan Pentateuch: z = 4.22 against 4.26 in the Masoretic text. The broader test of nesting gives T = 0.055 in both lines (p 0.0013 and 0.0012).

A looser version of the pattern shows up even in the Septuagint, the Greek translation made by Jewish scholars before the time of Jesus. So the shape was in the story before those lines of copies went their separate ways, and the scribes of each line handed it on intact.⁷7 Septuagint: T = 0.039, p = 0.015. This broader test measures whether repeated words nest around the centre; it cannot single out Wenham's own pairs, because a story that keeps repeating "water", "day" and "ark" nests a little by itself.

Two famous mirrors that did not pass#

The same broad test was run on two other passages often read as mirrors. Neither passed, and that matters: a test that found mirrors everywhere would tell us nothing about the flood.

The tower of Babel (Genesis 11:1-9) is short and full of repeated words like "earth", "language" and "build". Those are exactly what a word test can see, and they simply do not fall in mirror order.⁸8 Tested at both proposed centres: 11:5 (Fokkelman 1975, Wenham 1987) and 11:6 (van Wolde 1994). Holm-adjusted p = 1.0 at each.

The opening of John's Gospel ("In the beginning was the Word") failed for a different reason. Its proposed mirror is built from ideas rather than shared words. That may be real, but this test cannot see it, and later tests that looked for synonyms, shared references and clause patterns found nothing either.⁹9 R. A. Culpepper, "The Pivot of John's Prologue", New Testament Studies 27 (1980), 1-31. The later tests are tracks 6t-r2 (synonyms) and 6u-r (the things referred to, and clause patterns).

Whose hand shaped it?#

Jewish and Christian tradition has long received Genesis as the work of Moses. Many scholars, following J. A. Emerton, read the flood story instead as two older accounts woven together by a later editor.¹⁰10 J. A. Emerton, Vetus Testamentum 37 (1987), and Vetus Testamentum 38 (1988), 1-21.

We tried to test that question too, and we have to report that we could not. We first checked whether our method could pull apart two writers we already know are different, Genesis and Chronicles, when they were cut into pieces as small as the flood's. It managed that only 126 times in 200. A tool that blunt cannot settle whether the flood has one voice or two, so we did not use it.¹¹11 Tracks 153 and 153r. A flood unit keeps about 10 words once names and dates are set aside, and the method needed far more. Both runs failed their own checks, so they answer the question neither way.

What we can say is this. The flood story, as Scripture gives it to us, was shaped on purpose around God remembering Noah, and that shape has been kept by every line of copies we checked. Whether the Spirit worked through one writer or through faithful hands that gathered older accounts, the story turns on the same truth: God does not forget his people.

Every test here was written down, with its pass mark, before it was run. The full reports, with their methods and every figure in this post, are to be published with Carestead Research's open data.