The history of Torah codes

The scribes who counted

The Talmud, in Kiddushin 30a, says the early sages were called soferim, a word linked to counting, because they counted every letter in the Torah. It records their results. The middle letter of the Torah is the ו of גחון in Leviticus 11:42. The middle of its words falls at דרש דרש in Leviticus 10:16. Its middle verse is Leviticus 13:33.

The passage then tells a small story. Rav Yosef asks whether that ו belongs to the first half or the second. The others suggest bringing a scroll and counting. Rav Yosef answers that the early scholars were experts in full and defective spellings, and that his generation is not. The same page gives 5,888 as the number of verses in the Torah.

Our text gives a good sense of how much those counts have moved. The Leningrad Torah has 5,853 verses. Its middle letter stands in Leviticus 8:29, inside the word החזה, and the ו of גחון is letter 157,239 of 304,850. That is 4,814 letters past the midpoint. None of this is a problem for the Talmud, which told us spellings had become uncertain. It is a useful reminder that letter counts are a fact about a particular copy.

The sages in this passage count letters. They say nothing about reading them at equal intervals.

Rabbeinu Bachya and the sequence at 42

The first clear written example we can find is from Rabbeinu Bachya ben Asher of Saragossa, in his commentary on the Torah, begun in 1291. On Genesis 1:2 he discusses בהר"ד, the traditional calendar sign for the first new moon of creation: the second day, five hours and 204 parts of an hour. Then he says this sign is marked in the opening passage of the Torah, with 42 letters from each of its four letters to the next.

That can be checked, and we checked it. Start at the first letter of Genesis, the ב of בראשית, and take every 42nd letter. You read ב, ה, ר, ד. The letters fall in Genesis 1:1, 1:2, 1:3 and 1:4. It works in the Leningrad text, as he describes it, more than 700 years later.

Rabbeinu Bachya himself thought a discerning reader would see this was no accident, and he tied it to a teaching about creation. For a measurement, here is the other side of the ledger. A four-letter string like this one is expected 0.51 times at that one skip in Genesis, so one appearance in the book is ordinary. That it starts on the very first letter is the striking part, and that is a kind of fact chance figures handle badly, since nobody fixed the rule in advance. The gallery entry has the full numbers.

Rabbi Weissmandl

Rabbi Michael Dov Weissmandl was born in 1903 and died in 1957. He was a scholar of manuscripts, a rescuer of Jews in Slovakia during the Shoah, and after the war he rebuilt the Nitra yeshiva in the United States, at Mount Kisco, New York. In the history of this subject he is remembered for finding sequences by hand, long before computers.

His best-known observation is תורה at intervals of 50 letters, starting from the first ת of Genesis and again from the first ת of Exodus. It is often told together with similar patterns at the openings of Leviticus, Numbers and Deuteronomy. His students published his writings as Torath Hemed in 1958, after his death, and much of what is told in his name reaches us through them. The 1994 paper credits him with first noticing the phenomenon and cites that book.

Our gallery checks each observation on the Leningrad text. All six entries reproduce at the positions given there. Where a popular retelling names a different starting letter, the entry says so.

Computers and the rabbis

By the 1980s computers could search for what Rabbi Weissmandl had sought by eye. The mathematician Eliyahu Rips and the physics-trained Doron Witztum began testing patterns systematically. According to McKay and his co-authors, Rips described an early experiment on 19 or 20 rabbis in a 1985 lecture, and a preprint in 1986 gave the first list of 34. The statistician Persi Diaconis asked for a standard test and a fresh sample, and a second list of 32 rabbis followed in 1987.

A shorter version went to the Proceedings of the National Academy of Sciences, where Robert Aumann handled it, and was turned down after long argument. Statistical Science took a revised paper giving only the second list's results.

1994: the paper in Statistical Science

"Equidistant Letter Sequences in the Book of Genesis" appeared in August 1994. It reported that names of rabbis stood closer to their dates in Genesis than in scrambled pairings, with a rank of 4 in a million, or 0.000016 after correction for four statistics. The same test on six control texts came out flat. The editor described the paper as a puzzle for readers. Our method page goes through it step by step.

1997: The Bible Code

Three years later the journalist Michael Drosnin published The Bible Code, a best-seller that made the subject famous around the world. It went far past the paper, into statements about future events, and its name for the subject stuck. The authors of the 1994 paper did not go with it. Rips publicly distanced himself from the book, and he and Witztum rejected its way of reading the future. The rest of that story is in our Bible code guide.

Critics answered quickly. Shlomo Sternberg attacked the subject in Bible Review that August. In the spring of 1998 Jewish Action, the magazine of the Orthodox Union, set out the debate in one issue. It printed Witztum's defense, a skeptical essay by the Caltech mathematician Barry Simon, and a reply by Rabbi Daniel Mechanic defending the use of the codes in outreach. Simon also circulated a statement signed by 45 mathematicians who called the evidence entirely unconvincing. The authors of the paper and of the book shared a 1997 Ig Nobel Prize, the award that pokes fun at odd research.

1999: the reply

Statistical Science printed "Solving the Bible Code Puzzle" in May 1999. McKay, Bar-Natan, Bar-Hillel and Kalai argued that the freedom in choosing names and date forms was enough to produce the result. They showed a list with a comparable score in a Hebrew War and Peace, reported that small changes to the method nearly always weakened the result, and found nothing in their own replications. They concluded that the puzzle had been solved. Witztum and others disputed this, and the exchange went on outside the journal. The critique page has the details.

2004: the committee

Harold Gans had reported a link between rabbis' names and their towns. A Hebrew University committee formed in 1996 to test it, chaired by Robert Aumann, with Hillel Furstenberg, Eliyahu Rips, Dror Bar-Natan and Isaak Lapides. Its two experiments came in at ranks of 463,406 and 578,075 out of a million. The 2004 report by Aumann and Furstenberg says both failed to confirm the code. Lapides and Witztum disagreed and published their reasons alongside it.

Since then

Proponents went on working. A group of papers appeared at a pattern recognition conference in 2006, including Haralick's framework and a cluster study by Rips and Levitt. In 2022 and 2026 Doron Witztum published new work in B.D.D., the Bar-Ilan journal of Torah and scholarship, arguing for a test based on shuffled texts. We have found no paper on the subject in a mainstream statistics journal since 1999.

Jewish views

Jewish opinion has never been of one mind. In the 1990s some outreach groups used the codes in their seminars, and Rabbi Mechanic reported that leading rabbis approved of this. Others objected. Barry Simon argued the science was shaky and that the codes are peripheral to Jewish values. He reports that Rabbi Simcha Wasserman, asked about them, said one would not say the Torah blessing over them. Menachem Cohen, a Bible scholar at Bar-Ilan University, was a sharp critic of the lists and of their use in outreach.

Torah Codex takes no side in that discussion. We are a measuring tool. If you bring the question of what these things mean to your own teachers and your own reading, you will be in good company on either side.