The opening letters

The inventory

Exactly 14 of the alphabet's 28 letters are ever used this way: half of it, lit below. 26 of the 29 suras are Meccan and 3 Medinan. In 20 places the letters are counted as a verse of their own; elsewhere they open a longer first verse.

The combinations

Runs

Four stretches of consecutive suras share an opening. The longest is the best known: seven suras in a row, 40 to 46, all beginning Ha Mim.

• ال Alif Lam: suras 10 to 15, 6 in a rowالرالرالرالمرالرالر
• طس Ta Sin: suras 26 to 28, 3 in a rowطسمطسطسم
• ال Alif Lam: suras 29 to 32, 4 in a rowالمالمالمالم
• حم Ha (ha') Mim: suras 40 to 46, 7 in a rowحمحمحمعسقحمحمحمحم

A run is counted on the first two letters, which is why Ash-Shura (42) belongs to the Ha Mim seven even though it alone continues, in a second verse, with Ayn Sin Qaf.

What follows them

The oldest observation about the letters is not statistical. It is that what comes immediately after them is, nearly always, the Book speaking about itself. That is checkable, so we checked it.

23 of 29 mention the Book, its signs, or its sending down in the very next verse-unit. Widening the window by a single verse makes it 24.

The 5 that still do not open elsewhere: Maryam on the mercy of a Lord to His servant, Al-Ankabut on a question about being tried, Ar-Rum on news of a battle, Al-Qalam on an oath by the pen. Ash-Shura deserves naming as a near miss rather than a miss: the verse after its letters does speak of revelation, but with the verb awha, which is not one of the five terms this rule searches for. Widening the term list would move it, which is exactly why the term list is printed above rather than left implicit.

The frequency claim, tested

The claim that circulates most widely online is that each sura's opening letters occur unusually often inside that sura. It is a testable claim, and it is the kind of claim that survives mainly because no one runs the control.

What the test does turn up is a handful of genuinely striking individual cases, which is not the same thing as a pattern:

And a matching handful run the other way. Ha ranks 23 of 28 for enrichment in Ad-Dukhan, a sura that opens with it, and Qaf ranks 25 in Ash-Shura, which also opens with it. Both lists are in the data; publishing only the first would be how this claim got its reputation.

Is it a code?

The question people arrive with. Another script, another language, a key that classifies the sura it opens. Three parts of it can be answered with the text and the measurements already on this site, and one part cannot be answered here at all.

The script is the same script

They are ordinary Arabic letters at ordinary Arabic codepoints, the same ones the rest of the book is written in. There is no second alphabet hiding in the file. What the vocalised text does add is a single mark, the maddah, and it is not sprinkled at random:

Across all 29 openings there is not one exception. And the split is not arbitrary: it is exactly the recitation rule. The 8 marked letters are the ones whose spelled names run to three letters and take the long elongation. The 6 unmarked are alif and the five whose names are two letters long. So the text does encode something about these letters, and what it encodes is how to say them.

Worth being exact about what that is and is not. These marks belong to the later vocalisation layer, not to the earliest skeletal script, which wrote these letters plain like everything else. They are a record of how the letters were recited, carried in the orthography. They are not a cipher, and nothing about them is concealed.

Does a shared opening mean a shared subject?

The strongest form of the idea: if Alif Lam Mim were a label for a kind of content, the six suras carrying it should resemble each other more than lettered suras do in general.

Does the opening echo the sura's rhyme?

A sound-level version of the same idea, tested against the prosody measured for the Rhythm: 9 of the 27 testable suras rhyme on a consonant that is one of their own opening letters, against 0.177 for the same openings scored against suras that never received them. On its face that is a p of 0.0374.

The hits, for anyone who wants to check: 3, 30, 31, 40, 41, 42, 45, 46, 68.

What this site cannot test

Whether the letters come from another language or an older scribal practice. Proposals exist and some are serious: abbreviations of words or names, a conjecture that they were initials of the scribes who held the early codices (its own author later left it), derivations from Syriac or Hebrew usage. None has manuscript evidence or an internal argument that has persuaded the field, and testing any of them needs comparative corpora this site does not have. They are reported here as positions people hold, which is the same footing as the section below. A page that measured what it could and then guessed at the rest would have wasted the measuring.

What the clustering found

The one genuinely startling result was not looked for. Every verse in the book was embedded by meaning and grouped by an algorithm that was never told these verses were unusual, never given the list, and never shown a sura boundary. It sorted the book into 120 groups, and one of them came out like this:

Both directions are exact. The group is the letter verses, and the letter verses are the group. It also caught the split in Ash-Shura, placing 42:2 with the rest without being told the two halves belonged together.

The honest reading of this is narrow, and worth saying plainly: it shows the letters are unlike everything else in the book, in a way a machine measuring meaning can detect. It does not show what they mean. A group of verses that share a form will cluster on that form; that is what clustering does. The finding is the sharpness of the boundary, not a decoding. You can see them in the Universe as a knot of stars sitting apart from the whole cloud.

What the commentators hold

Positions, not conclusions. These are held by people of knowledge and reported here as their positions; the site does not adjudicate between them.

• Their meaning belongs to God. The oldest and most widely held position, reported from several of the Companions: the letters are among what God alone knows, and the reciter says them as they were given.
• They are a challenge. The book is made of these same letters, which every Arab already possessed, and it is placed beside a standing invitation to produce its like. On this reading the letters point at the raw material and let the contrast speak.
• They are names. Of the sura, or of the Quran itself, or abbreviations of divine names. Two suras are known to this day by their letters alone: Ta Ha and Ya Sin.
• They arrest attention. A sound with no meaning, before a text with nothing but meaning, at the moment recitation begins.