Skip to content

Glossary

The words you will meet in Torah codes research and on this site, in plain English. Each one is defined the way the published papers use it.

Appellation
A name by which a person is known: a full name, title, acronym or nickname. The 1994 experiment used 1 to 11 appellations per rabbi. Which ones to include became the main point of dispute.
Best of k
The best result out of k tries. It looks about k times better than one try would, for no reason other than the number of tries. The trial counter reports it.
Bonferroni correction
A simple safeguard: if you test k things and keep the best, multiply its p-value by k. WRR did this with their four statistics, turning 4 in a million into 16 in a million.
Chance alone
Our label for the expected count: how many sequences a text with these letter frequencies should hold, on no assumption beyond that.
Cluster
A group of sequences of related words that stand close together, usually judged by the smallest rectangle or the shortest stretch of text that holds them.
Compactness
How tight a group of sequences is. Published measures include the area, diagonal or perimeter of a rectangle, and the length of a stretch of text.
Control
A text or rearrangement where the same search is run for comparison. Each control keeps some features of the text and breaks others. See the controls page.
Corrected distance
WRR's measure c(w, w′): the rank of a word pair's closeness among near-miss versions of the same pair. Small means unusually close. It runs from about 1/125 to 1.
Cylinder
The text wound around a tube, one long line in a spiral, so the right and left edges of a matrix meet. WRR measure distances on cylinders.
Defective spelling כתיב חסר
Writing a word without an optional vowel letter. The opposite is full spelling. Editions differ here, which is why letter counts differ.
Direction
Forward reads with the text, and backward reads against it. A backward sequence has a negative skip.
Domain of minimality
For a sequence: the longest stretch of text holding it and no other sequence of the same word with a smaller skip. A long domain marks a more noteworthy sequence in WRR's method.
Edition
A particular text of the Bible. Editions differ by a few letters. We use the Westminster Leningrad Codex. The 1994 paper used the Koren edition.
ELS
Equidistant letter sequence. A word spelled by letters at equal spacing, written (n, d, k) for its start, skip and length.
ELS random placement
A control by Robert Haralick, or ERP. It slides each found sequence to a random start, keeping its skip and letters, and measures the cluster again.
Expected count
The number of sequences that the letter frequencies alone would produce. For each skip: the places a term can start, times the product of its letters' shares, added up over the search.
Final letters
The five Hebrew letters with a separate word-end form: ך ם ן ף ץ. We count them as their regular forms.
Full spelling כתיב מלא
Writing a word with its optional vowel letters, such as ו or י.
Hebrew University committee
The group formed in 1996 to test a claim linking rabbis' names to their communities. Its report of 2004 says two tests failed to confirm the code, and two of its members dissented.
Ketiv כתיב
A word as it is written in the text. This is what we search.
Key word
A term a researcher looks for in a cluster experiment. Fair practice is to fix the list before searching.
Leningrad Codex
The oldest complete manuscript of the Hebrew Bible, copied about 1008 and now in St. Petersburg. Its scribes also drew carpet pages from tiny letters.
Look-elsewhere effect
The more places you look, the more likely an unusual result somewhere. Our trial counter exists because of it.
Matrix
The text laid out in rows of a chosen width, so sequences appear as straight lines.
MBBK
McKay, Bar-Natan, Bar-Hillel and Kalai, authors of the 1999 reply to WRR in Statistical Science.
Minimal skip
The smallest skip at which a word appears, across the whole text or a stretch of it. Every word that appears has one.
Monkey text
An older name for a randomized control text.
n/N
How we report a control: n runs out of N matched or beat the real result.
Null hypothesis
The default assumption that a test tries to reject. In WRR's test it says the true pairing of names and dates could fall at any rank among the scrambled ones.
Ordinal index
A number used only for ranking, not as a probability. WRR say P₁ and P₂ should be read this way.
P₁, P₂, P₃, P₄
WRR's four summaries of a whole list of pairs. P₁ counts pairs with c at or below 0.2. P₂ uses the product of all c values. P₃ and P₄ are the same with names beginning "Rabbi" removed.
Perturbation
In WRR's method, shifting the last three letters of a sequence by a small amount, to make near misses for comparison. There are 125 shift patterns.
Permutation test
A test that scrambles a pairing many times and ranks the true pairing among the scrambles. WRR scrambled which rabbi got which dates.
Poisson distribution
The standard model for a count of rare events. From an expected count, it gives the chance of seeing a given number or more.
Pre-registration
Fixing a test before you run it. In the laboratory, this is the Lock and search button.
Proximity
How close two sequences stand. WRR's measure is Ω, later corrected to c.
Qere קרי
A word as traditionally read where it differs from how it is written. We show it beside the verse. We do not search it.
Rank
Where a result falls among control results. A rank of 4 in a million means only 3 of the others were stronger.
Resolution
The finest figure a control can show. N runs cannot resolve anything finer than about 1 in N.
Resonance
Haralick's term for a row width at which a sequence looks compact, with small row and column steps.
Row-split
A display that divides each matrix row into 2 to 11 layers, showing a long skip in a tighter picture. It moves no letter.
Scope
The part of the text you search: the Torah, the Tanakh, Nevi'im, Ketuvim, a book, a parsha or a range of chapters.
Seed
The starting number for a shuffle. The same seed gives the same shuffle, so a control run can be repeated exactly.
Shortest window
The shortest stretch of text that holds one sequence of every term you searched.
Span
The distance from a sequence's first letter to its last: 1 + (k − 1) × |skip| letters.
Tanakh תנ״ך
The 24 books of the Hebrew Bible: Torah, Nevi'im (the Prophets) and Ketuvim (the Writings).
Toroidal display
A flat matrix whose edges wrap around, so lines leaving one side continue on the other.
Trial counter
Our record of what you have tried in a session: terms, skip ranges, places searched and widths.
Tuning
MBBK's word for an unknown process by which an experiment's choices end up favoring its result. They say they do not know how it happened, if it did.
Verification line
The small line on every export that gives the stream version and a short hash of the search, so anyone can rebuild the result.
WRR
Witztum, Rips and Rosenberg, authors of the 1994 Statistical Science paper. Also the name of its method.

Open the laboratory

Sources

Checked against its sources on 2026-10-31.