They Found The Same Hidden Names In War And Peace.
A 1994 paper in a statistics journal reported names and dates hidden in Genesis at odds of fifty thousand to one. Five years later the same journal printed the answer: run the method on Tolstoy and it works as well.
A months-long reporting project. Documents cited below are held in The Vault and available to readers.
JERUSALEM — In 1994 a peer-reviewed statistics journal published a paper reporting that the names of rabbis and the dates of their births and deaths appear in the Hebrew text of Genesis, spaced at equal intervals, closer together than chance should allow.
The odds it reported were about fifty thousand to one.
In 1999 the same journal published the answer. Four mathematicians had run the method again, and one of the things they ran it on was War and Peace.
What the method does
Take a text, strip the spaces, and read every *n*th letter. Do it for every n. Words appear, because words always appear.
The 1994 experiment was more careful than that. It asked whether pairs of related terms — a rabbi's name and his date — turned up near each other in the grid more often than they should, and it computed a probability.
That is a real experiment. This desk wants to say so plainly, because the usual coverage treats the 1994 paper as obviously silly, and a journal's referees did not think it was.
Where the freedom was
Not in the mathematics. In the appendix.
A rabbi from that period can be named several ways: the name he was born with, the name he is known by, an honorific, an acronym of his title, or the title of his most famous book used for the man himself.
The 1994 authors had to choose one form for each of thirty-odd rabbis. Every choice they made was defensible.
The 1999 team's argument is that the number at the end depends on those choices far more than on the text — and that the space of defensible lists is enormous.
The Tolstoy test
So they tuned a list the same way and ran it on a Hebrew translation of War and Peace.
It found the rabbis there too, at strength comparable to Genesis.
Then they ran Genesis with a list chosen independently, by people not looking at the outcome. It found almost nothing.
The chart above is those four runs. The signal moves with the list. It does not move with the book.
The word our statistician would not let us use
Cheating.
Dr. Brennan-Oduya asked, before she would talk to us, that the word be kept out of the piece, and her explanation is at the foot of this story and is the reason to read it.
Her point is that every step in the process is honest. You try a list, the number is unremarkable, you reconsider a single name on genuine scholarly grounds, you run it again, the number improves. Nobody has lied. And the number at the end reports the odds of the last list you tried as though it were the only list you tried.
"That is not fraud," she told us. "It is the most human thing in science, and I have done it myself, and so has everybody who tells you they have not."
The fix is not integrity
This is the sentence this desk took away and it belongs to her.
Integrity does not help, because the people it happens to are behaving with integrity. The fix is writing the list down first, sealing it, and giving it to somebody else.
She noted, drily, that this newspaper does that with its own studies, and said it was the only reason she agreed to be interviewed.
The two readers
We interviewed two people who believe the codes are real. Both had read both papers. Both were more precise about the 1999 objections than most of the coverage this desk waded through.
One of them said something our correspondent has been turning over since.
"You are going to write that the second paper won. Fine. But notice what it took. Four mathematicians, five years, a full rerun, and a Russian novel. Nobody does that for an ordinary claim. They did it because the first paper was good enough to be worth destroying properly, and that is a compliment nobody will print."
We have printed it.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
Sources & Method
We read the appendix. The argument has been conducted for thirty years over the headline probability, and the thing that decides it is a list of name-forms in the back of the 1994 paper. Then we read the 1999 rerun, and asked a statistician with no stake in the dispute to explain what the second team actually varied.
Who we spoke to
- The 1994 paper, Peer-reviewed; published in a statistics journal. Read in full, including the appendix listing the name-forms used 2026 The appendix is the story. It is the part nobody reads.
- The 1999 paper in the same journal, Peer-reviewed, by four mathematicians. Read in full, including the supplementary experiments 2026 Free to read, as is the 1994 paper. Both have been public the whole time.
- Dr. Yaffa Brennan-Oduya, Statistician; works on multiple-comparison problems and has no stake in this dispute. Interviewed by video call three times; walked us through the appellation issue May–June 2026 Asked us not to use the word cheating anywhere in the piece, and explained why. We have not.
- Two readers who believe the codes are real, Named in our notes with consent. Interviewed at length; both have read both papers June 2026 Neither is a fool and both were more precise about the 1999 objections than most of the coverage.
Documents
- PX-2185 — Two papers in the same statistics journal, five years apart: the 1994 claim and the 1999 rerun accepted
What we could not confirm
- Nothing about the two papers. Both are peer-reviewed, both are free to read, and both are filed with this story.
- What the 1994 authors intended. The 1999 paper does not allege dishonesty and neither does this newspaper. The mechanism it describes works perfectly well on people trying their best.
- Whether any hidden structure exists in any text. This story is about one experiment and its rerun. It is not a general theorem about scripture and does not pretend to be.
How Others Covered This
The same events, as reported elsewhere on the same day. We list what each outlet had that we did not, as well as what we had that they did not — including where we come off worse. Why we print this.
-
ApertureThe Grid
An essay on letter-counting as a devotional practice, going back centuries, treated as a religious tradition rather than a claim.
Had that we did not
The history, which is long and serious and which our piece skates over in one sentence.
Left out
The statistics.
-
The Hollow PostTHE CODE THEY PROVED AND THEN BURIED
Reported the 1994 paper as a proof that was suppressed once its implications became clear.
Had that we did not
That a peer-reviewed statistics journal published it. It did, and said so on the cover.
Left out
That the same journal published the refutation, that the refutation is by four mathematicians who did the experiment again, and that it is free to read.
-
They Buriedthis newspaperThey Found The Same Hidden Names In War And Peace.
Reported both papers, five years apart, in the same journal, and what the second one actually did.
Had that we did not
The mechanism: the freedom was in the name-forms, and nobody had fixed them in advance.
Left out
Our headline gets its laugh from Tolstoy and buries the part that matters, which is that the original authors were not accused of dishonesty. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Dr. Yaffa Brennan-Oduya, statistician read the full draft; no deadline. Asked that one word be kept out of the piece and explained why at length. Printed unedited.
I asked your reporter not to write the word cheating and she has not, and I want to explain the request, because it is the only useful thing I have to say.
A rabbi from that period can be referred to in several ways. The name he was born with. The name he is known by. An honorific. An acronym of his title. The name of his most famous book, used for the man. All of these are legitimate and a scholar choosing among them is doing ordinary scholarship.
Now: you have a list of thirty-odd rabbis, several defensible name-forms each, and a procedure that returns a number. You try the list. The number is unremarkable. You reconsider one name — genuinely reconsider it, because on reflection the other form is more standard — and you run it again. The number improves.
Nobody has done anything wrong at any step. Each individual choice is defensible and would survive a viva. But you have walked through a space of thousands of possible lists in the direction of a better answer, and the number at the end does not know that. It reports the odds of the last list you tried, as though it were the only one you tried.
That is not fraud. It is the most human thing in science and I have done it myself and so has everybody who tells you they have not.
The fix is not integrity. Integrity does not help, because the people it happens to are behaving with integrity. The fix is writing the list down first, sealing it, and handing it to somebody else.
Your newspaper does that with its own studies, I notice. It is the only reason I agreed to talk to you.
Published unedited under our right-of-reply guarantee.
How was this story?
We publish the result, whatever it is. Reader verdicts appear on the front page and in our newsroom metrics.
Readers' Letters 0
Printed at once under the name you give and read by the desk afterwards; anything unfit is removed, with a note saying so, and nothing else is ever deleted — only corrected. Letters that changed something in the story carry a mark saying so, and there are 5 of those across the archive.