Has the Quran been perfectly preserved?

Summary

Two very different claims travel under the word "preserved," and almost every argument about the Quran's text is really an argument about which one is being defended.  Separate them and the evidence becomes clear.

• The Quran's consonantal skeleton descends from a single written archetype, and that archetype is early. Marijn van Putten showed this by tracking spellings that carry no meaning at all — whether ni'mat, "grace," is written with one final letter or another. The choice is scribal whim. Whim does not repeat itself across fourteen independent manuscripts. It does repeat, almost perfectly, which means every surviving copy of the standard text was copied from one written exemplar, not written down from dictation or from memory.¹¹
• That archetype is plausibly Uthman's, around 650. Van Putten's own words: "the manuscript evidence cannot preclude with absolute certainty an earlier redaction," but "a time much later than his reign is now a virtual impossibility."² This is a serious result, it is published in a top journal, it is open access, and it has not been refuted. Christians who claim the Quran was assembled in the eighth or ninth century are arguing against the evidence.
• The popular claim — "not one letter has changed" — is refuted by Sahih al-Bukhari and Sahih Muslim, before any manuscript, orientalist or Christian is consulted. Umar preached from the pulpit in Medina that the stoning verse "we did recite" is not in the Book.³ Aisha said a ruling of ten, then five, sucklings "was in the Quran and was being recited" when Muhammad died.⁴ Abu Musa al-Ash'ari told three hundred reciters at Basra that two whole surahs had been recited and forgotten, one of them the length of Surat Bara'a.⁵ Umar said of Ubayy ibn Ka'b, the best reciter among them, "we leave some of what he recites."⁶
• The tradition's own answer to all of this is a doctrine, not a denial. Naskh al-tilawa — abrogation of the recitation — says God removed those passages deliberately, on the authority of Quran 2:106.⁷ That is coherent theology. It is also a concession that the Quran recited today is not word for word the Quran recited in 632, which is exactly what the popular slogan denies.
• Uthman's committee edited, and then burned the alternatives. Bukhari 4987 says so in as many words, and the Arabic verb is yuhraqa, from the root for setting a thing on fire.⁸ The same report has Uthman instruct three Qurashi scribes to resolve any disagreement with Zayd ibn Thabit in favour of the Qurashi dialect. Bukhari 4988 says a verse of Surat al-Ahzab was missing from the sheets they were copying and had to be recovered from one man.⁹
• The recited text has never been single, and the tradition has never claimed it was. Ten readings are canonical, each with two named transmitters. Two are in ordinary print today: Hafs, used almost everywhere, and Warsh, used across Morocco, Algeria and West Africa. Counting Adrian Brockett's collation gives 138 individually listed differences between them in the first five surahs alone, four of which are in the consonantal skeleton itself.¹⁰ At Quran 3:146 one says many a prophet fought and the other says many a prophet was killed.
• One manuscript survives from outside the standard family: the erased lower layer of the Sanaa palimpsest. Its wording differs from the standard text in hundreds of places, its chapter order is different, and at Quran 2:196 it lists two options for pilgrimage expiation where the standard text lists three.¹¹ But the verses and their order inside the surahs are those of the standard text, with two exceptions Sadeghi names himself — Quran 20:31–32 stand in reverse order and Quran 9:85 is missing — and no doctrine changes anywhere in it.¹¹
• The radiocarbon evidence is much weaker than either side claims. One leaf of the Sanaa palimpsest was measured by four laboratories and came back 406–543, 441–636, 598–649 and 599–655 CE.¹² Radiocarbon dates the animal whose skin became the parchment, not the ink on it. For both the Birmingham and the Tübingen manuscripts, the specialists who studied the writing dated the text later than the parchment.¹³

Verdict:  the Quran's consonantal text is preserved about as well as any ancient text has ever been preserved, and the popular Muslim claim asks for something else entirely.  It asks for a claim about the recited text, and a claim about the twenty years between Muhammad's death and Uthman's committee, that the Muslim sources themselves refute.  The scholarly claim — a stable written archetype from the mid-seventh century, copied with extraordinary care ever since — largely survives, and this article says so without hedging.  The letter-perfect claim does not survive, and it was Muslim scholars, not Christian critics, who first wrote down the reports that sink it.

Where this article's own case is weak.  Four places, stated up front.  First, everything in the "material no longer in the text" section comes from the same hadith pool whose reliability is elsewhere disputed;  a Christian who uses Bukhari to convict must accept Bukhari when it acquits.  Second, naskh al-tilawa is a real doctrine with a real anchor in Quran 2:106, and calling it an evasion is not an argument against it.  Third, the manuscript evidence runs strongly for the Muslim side on the question that matters most to a historian — whether the text was stable — and Christian apologetics that leans on Daniel Brubaker's corrections or on the Sanaa variants is leaning on something Hythem Sidky's review took apart example by example.¹⁴  Fourth, the Christian New Testament has its own textual problems, some of them larger in absolute terms, and the comparison at the end of this article is drawn carefully because it is usually drawn badly by both sides.

The claim being tested

A Muslim who says the Quran is preserved usually has Quran 15:9 in mind.  In the Uthmani Arabic script served by quran.com, it reads إِنَّا نَحْنُ نَزَّلْنَا ٱلذِّكْرَ وَإِنَّا لَهُۥ لَحَـٰفِظُونَ.  Sahih International renders it:  "Indeed, it is We who sent down the message [i.e., the Qur'ān], and indeed, We will be its guardian."¹⁵

Two things must be said about that verse before it can be used, and both cut gently.

The Arabic word is al-dhikr, "the reminder."  It is not al-Qur'an.  The bracketed gloss "[i.e., the Qur'ān]" belongs to the Sahih International translators, not to the Arabic.  Muslim commentators have overwhelmingly read al-dhikr here as the Quran, and that reading is old and reasonable — but an argument may not lean on a translator's bracket as though the Arabic said it.

More importantly, the verse promises preservation without specifying the mechanism, the scope, or the terminus.  It is compatible with a consonantal text copied letter for letter from a single archetype.  It is equally compatible with the doctrine that God removed some recited passages on purpose.  Muslims on both sides of that question quote it, and both are entitled to.

So the verse cannot settle the argument.  What can be tested are four distinct propositions that popular argument runs together:

1. The letter-perfect claim. Not one letter of the Quran has changed since it was revealed. Every copy on earth is identical.
2. The archetype claim. The consonantal skeleton of the Quran — the bare letters, without vowel marks and often without the dots that distinguish one consonant from another — goes back to a single written master copy produced within about twenty years of Muhammad's death, and has been copied faithfully ever since.
3. The prophetic-original claim. That archetype is identical to what Muhammad recited.
4. The single-recitation claim. There is one Quran, recited one way, everywhere.

These are not the same claim, and they do not stand or fall together.  Where the Quran's material came from in the first place is a different question again, treated in the origins article.  Proposition 2 is well supported.  Proposition 1 is refuted by the Muslim sources.  Proposition 3 is not testable with the evidence anyone currently has, in either direction.  Proposition 4 has never been true and the tradition has never said it was.

This article uses the Muslim tradition's own sources wherever it can, for a plain reason:  a Muslim reader can check them, and does not have to accept a Western dating to follow the argument.  The hadith come from Sahih al-Bukhari (d. 256/870) and Sahih Muslim (d. 261/875), which no Sunni disputes;  where a report comes from a sunan collection with a modern grade attached, the grader is named.  The classical material comes from Ibn Abi Dawud (d. 316/928), al-Suyuti (d. 911/1505), Ibn al-Jazari (d. 833/1429) and al-Qurtubi (d. 671/1273).  Western scholarship enters where it weighs evidence the tradition cannot weigh — chiefly the manuscripts, which no medieval scholar could see.  How much weight the hadith pool can bear in general is its own subject, taken up in the article on hadith canon and criticism.

The strongest case for the Muslim claim

Before any of this is evaluated, the Muslim case deserves to be stated at full strength.  It is much stronger than most Christian apologetics admits, and three of its four pillars are simply correct.

First, the archetype.  In 2019 Marijn van Putten, a linguist at Leiden, published an argument in the Bulletin of the School of Oriental and African Studies that has changed the field.  Quranic orthography contains spellings that carry no linguistic information.  Feminine nouns in the construct state are usually written with a final ha', but about 22% of the time with a final ta'.  For ni'mah, "grace," the split across the Quran is nearly even:  twelve occurrences one way, eleven the other.¹  Nothing distinguishes them.  A scribe writing from dictation, or from memory, would land those spellings differently every time.

Van Putten checked all twenty-three occurrences against fourteen early manuscripts.  His result:  "there is an extremely strong correlation between the position of the idiosyncratic spellings in the Cairo edition and earlier Quranic manuscripts.  With only two locations, Q 16:144 and Q 37:57, showing any disagreement at all.  This great consistency cannot be attributable to chance."¹  His conclusion:  "there must have been a single written archetype from which all Quranic manuscripts of the Uthmanic text type are descended," and "from the very beginning of the text's standardization, every single manuscript that belongs to the Uthmanic text type must have been copied from a written exemplar.  It is exactly these kinds of orthographic idiosyncrasies that would be impossible to reproduce through a process of writing down from dictation."¹¹

He replicated the pattern with three further quirks.  Jaza'u in the construct state is normally spelled one way but appears differently in exactly five places — Quran 5:29, 5:33, 20:76, 39:34 and 42:40.  Al-mala'u is spelled one way twelve times and another way in exactly four — Quran 23:24, 27:29, 27:32 and 27:38.  "Every single early Quranic manuscript that I have examined is in agreement that those are to be spelled as such."¹

And his own statement of what this means for the traditional Muslim account:  "While it is of course impossible to prove that the standardization did not take place before 'Uthmān's reign on the basis of this data, the data is absolutely consistent with the traditional account."¹

Second, the stemma.  Michael Cook examined the consonantal differences that the tenth-century scholar al-Dani reports between the Uthmanic copies sent to Medina, Damascus, Kufa and Basra.  Those differences form a clean family tree with no cross-contamination:  Damascus shares readings with Medina but never with Kufa or Basra, and so on.  Medieval Muslim scholars had no concept of stemmatics and could not have manufactured that pattern.  Cook's conclusion, quoted by both Sidky and van Putten:  "this must count against any suggestion that the variants were invented. …We can accordingly infer that we have to do with genuine transmissions from an archetype."¹⁶  Yasin Dutton then found the predicted Syrian variants in real Syrian-type manuscripts.

Third, the collation results.  Adrian Brockett, who collated the two printed transmissions of the Quran in use today, reached a conclusion no Muslim would quarrel with:  "The simple fact is that none of the differences, whether vocal or graphic, between the transmission of Hafs and the transmission of Warsh has any great effect on the meaning."¹⁷  Arthur Jeffery, an orientalist with every professional motive to say otherwise, wrote in 1940 that "practically all the early Codices and fragments that have so far been carefully examined, show the same type of text, such variants as occur being almost always explainable as scribal errors."¹⁸

Fourth, the one manuscript from outside the family agrees on the hardest thing.  The erased under-text of the Sanaa palimpsest is the only known Quran manuscript that is not of the standard type.  Its wording differs in hundreds of places and its chapter order is different.  And yet:  "even a careful memorizer who reproduces the words exactly is prone to getting the order of the verses wrong;  yet C-1 has the same verses and the same order of verses as the standard Qur'ān."¹¹  Behnam Sadeghi's own verdict, in a secular journal, is more conservative than mainstream Muslim tradition itself:  "textual criticism suggests that the standard version is the most faithful representation, among the known codices, of the Qur'ān as recited by the Prophet."¹⁹

That is the Muslim case, and on the historian's question — was the text stable? — it wins.  What follows does not overturn it.  What follows shows that it is not the claim the dawah table is making.

What the tradition says happened

Scraps, a board and an inkpot

The tradition places the revelation over roughly twenty-two years, from about 610 to Muhammad's death in 11/632.  It does not describe a book being written.  It describes dictation onto whatever was to hand — Zayd ibn Thabit's list at Bukhari 4986 is al-'usub wa-l-likhaf wa-sudur al-rijal, palm stalks, thin white stones, and the chests of men, and the parallel at Bukhari 7191 adds "the pieces of leather and hides."²⁰

Bukhari 4990 has al-Bara' report that when Quran 4:95 came down, "The Prophet said, 'Call Zaid for me and let him bring the board, the inkpot and the scapula bone.'"  The same report says the blind 'Amr ibn Umm Maktum objected that he could not fight, and that the verse was re-revealed with the clause "except those who are disabled" inserted.²¹  Whatever one makes of the theology, the tradition's own account here is of a text edited in the Prophet's presence.

Abu Bakr's collection, after the battle of Yamama

The founding report is Sahih al-Bukhari 4986, in Book 66, "Virtues of the Qur'an," in a chapter headed bab jam' al-Qur'an, "the collection of the Qur'an."  Zayd ibn Thabit narrates:

Abu Bakr As-Siddiq sent for me when the people of Yamama had been killed …  Abu Bakr then said (to me), "'Umar has come to me and said:  'Casualties were heavy among the Qurra' of the Qur'an (i.e. those who knew the Qur'an by heart) on the day of the Battle of Yamama, and I am afraid that more heavy casualties may take place among the Qurra' on other battlefields, whereby a large part of the Qur'an may be lost.  Therefore I suggest, you (Abu Bakr) order that the Qur'an be collected.'  I said to 'Umar, 'How can you do something which Allah's Apostle did not do?' …"  Then Abu Bakr said (to me).  "You are a wise young man and we do not have any suspicion about you, and you used to write the Divine Inspiration for Allah's Messenger.  So you should search for (the fragmentary scripts of) the Qur'an and collect it in one book."²²

Three details in that report deserve attention, and they cut in different directions.

Zayd objects twice that he is being asked to do "something which Allah's Messenger did not do."  The tradition is candid that the collection was not a prophetic instruction.  It was an administrative response to battlefield deaths, and the reason given for it is that "a large part of the Qur'an may be lost."

The collection had a documentary standard, and it was strict.  Ibn Hajar (d. 852/1449) reports Abu Bakr's instruction as:  "Sit at the entrance to the [Prophet's] Mosque.  If anyone brings you a verse from the Book of Allah along with two witnesses, then record it."  M. M. al-Azami, the most substantial modern Muslim defender of the Quranic text, takes Ibn Hajar's second reading of "two witnesses" — "acceptance of only those materials which, according to the sworn testimony of two others, had been written in the Prophet's very presence."¹⁸  If that is right, Abu Bakr's committee accepted nothing on memory alone.  It is a real argument, not a slogan, and it is the strongest thing in the Muslim account of this stage.

But the same report ends with a documentary problem.  Zayd says:  hatta wajadtu akhira surat al-Tawba ma'a Abi Khuzayma al-Ansari, lam ajidha ma'a ahadin ghayrihi — "until I found the last of Surat at-Tawba with Abu Khuzayma al-Ansari;  I did not find it with anyone other than him."  One man, one copy.  The parallel at Bukhari 4989 says "the last two Verses of Surat at-Tauba," not the last verse.²³  Bukhari 7191 says the material was found "from Khuza'ima or Abi Khuza'ima" — the tradition is not sure which man it was.²⁰  On a two-witness rule this passage should not have made it in.  A Muslim reader may reasonably answer that Zayd, who had heard the Prophet recite it, was himself the second witness.  That answer works;  it also means the two-witness rule was not absolute.

Uthman's recension, about 30/650

Sahih al-Bukhari 4987, the next hadith in the same chapter, narrated by Anas ibn Malik:

Hudhaifa bin Al-Yaman came to 'Uthman at the time when the people of Sham and the people of Iraq were Waging war to conquer Arminya and Adharbijan.  Hudhaifa was afraid of their (the people of Sham and Iraq) differences in the recitation of the Qur'an, so he said to 'Uthman, "O chief of the Believers!  Save this nation before they differ about the Book (Qur'an) as Jews and the Christians did before."  So 'Uthman sent a message to Hafsa saying, "Send us the manuscripts of the Qur'an so that we may compile the Qur'anic materials in perfect copies and return the manuscripts to you." …  'Uthman said to the three Quraishi men, "In case you disagree with Zaid bin Thabit on any point in the Qur'an, then write it in the dialect of Quraish, the Qur'an was revealed in their tongue."  They did so, and when they had written many copies, 'Uthman returned the original manuscripts to Hafsa.  'Uthman sent to every Muslim province one copy of what they had copied, and ordered that all the other Qur'anic materials, whether written in fragmentary manuscripts or whole copies, be burnt.⁸

John Burton's independent academic rendering of the same Arabic, in The Collection of the Qur'an (Cambridge, 1977), agrees in every substantive detail:  Hudhayfa "had had an opportunity to observe regional differences over the Qur'an";  Uthman "commanded them to copy the sheets into several volumes";  and "when they had copied the sheets, 'Uthman sent a copy to each of the main centres of the empire with the command that all other Qur'an materials, whether in single sheet form, or in whole volumes, were to be burned."²⁴

The verb is not ambiguous.  The Arabic is wa-amara bi-ma siwahu mina l-Qur'ani fi kulli sahifatin aw mushafin an yuhraqa.  Yuhraqa is the passive of haraqa, root ح ر ق, the ordinary word for setting a thing on fire.  Ibn Hajar's summary term for the whole episode is al-tahriq, "the burning."  Al-Azami's own paraphrase is the honest one:  "Ibn Hajar says that despite most reports incorporating the word at-tahriq (burning), every possibility must be considered.  The fate of each fragment rested with the individual possessing it:  whether to erase, tear, or burn."¹⁸

How early is this report?  This matters, because a critic who treats it as a ninth-century legend has no business quoting it, and an apologist who treats it as an eyewitness minute has no business either.  Harald Motzki's analysis of the transmission chains puts the common link of the whole bundle at Ibn Shihab al-Zuhri (d. 124/741–742) — about ninety years after the event, and about a century before Bukhari wrote.  Van Putten, who prints the two independent Arabic transmissions from al-Zuhri side by side, calls the bundle "extraordinarily early" and concludes that "this report may very well retain a historical memory of the original language of composition of the Quran."²  So:  not contemporary documentation, but not a late invention either — a report with a named common link about ninety years downstream, and, as the manuscripts show, corroborated in its main outline.

Bukhari 4988 is the most awkward report in the file, and it sits immediately after 4987 in the same chapter:

Zaid bin Thabit added, "A verse from Surat Ahzab was missed by me when we copied the Qur'an and I used to hear Allah's Messenger reciting it.  So we searched for it and found it with Khuza'ima bin Thabit Al-Ansari.  (That Verse was):  'Among the Believers are men who have been true in their covenant with Allah.' (33.23)"⁹

Zayd's Arabic is faqadtu ayatan mina l-Ahzab hina nasakhna l-mushaf… fa-alhaqnaha fi suratiha fi l-mushaf — "I lost a verse of al-Ahzab when we copied the codex… and we attached it in its surah in the codex."  A verse was missing from Abu Bakr's sheets and was restored, at the Uthmanic stage, from one man's copy.

Two independent details are worth recording because they complicate both sides.  Jami' at-Tirmidhi 3104, graded Sahih by Darussalam — that is, al-Albani's twentieth-century judgement as adopted by that publisher, not a medieval grading — carries the same material from al-Zuhri, names the four scribes in a different order, adds a dispute over how to spell al-tabut that Uthman settled in favour of the Qurashi form, and does not contain the burning clause at all.²⁵  A Muslim cannot use Tirmidhi's silence to soften Bukhari, because Bukhari is the stronger collection.  But a critic who makes the bonfire the centre of the story should know that an early independent line of the same report omits it.  And Bukhari 4984 gives the whole episode without Hudhayfa and without the burning at all.²⁶

The objection nobody quotes

Ibn Abi Dawud's Kitab al-Masahif, the fourth-century collection devoted to exactly this subject, preserves a chapter on "Uthman's permitting recitation other than by his own codex."  In it the Egyptian delegation that came to confront Uthman lists its grievances.  The first one is:  naqamna annahu maha kitab Allah 'azza wa-jall — "We resent that he erased the Book of God Almighty."

Uthman's reply, in the same report:  "As for the Quran, it is from God;  I only forbade you because I feared division among you, so recite on whichever harf you wish."²⁷

That exchange is the whole controversy in miniature, preserved by a Muslim scholar for Muslim readers.  Uthman's motive, on his own reported testimony, was to stop a fight.  His critics' complaint, on their own reported testimony, was that he had erased scripture.  Arthur Jeffery's Western version of the same observation, written in 1937, adds the detail that matters:  "It is very significant that the Qurra' were violently opposed to 'Uthman because of this act, and there is evidence that for quite a while the Muslims in Kufa were divided into two factions, those who accepted the 'Uthmanic text, and those who stood by Ibn Mas'ud, who had refused to give up his Codex to be burned."²⁸

The companion codices

The men whose copies Uthman superseded were not marginal figures.  Bukhari itself establishes their standing.

Bukhari 4999:  the Prophet said, "Take (learn) the Qur'an from four:  'Abdullah bin Masud, Salim, Mu'adh and Ubai bin Ka'b."  Ibn Mas'ud is named first.²⁹  Bukhari 5000:  Ibn Mas'ud preached, "By Allah, I learnt over seventy Suras direct from Allah's Messenger.  By Allah, the companions of the Prophet came to know that I am one of those who know Allah's Book best of all of them, yet I am not the best of them."  The narrator Shaqiq adds:  "I sat in his religious gathering and I did not hear anybody opposing him."³⁰  Bukhari 5001:  at Hims a man told Ibn Mas'ud that Surat Yusuf "was not revealed in this way";  Ibn Mas'ud answered, "I recited it in this way before Allah's Messenger and he confirmed my recitation by saying, 'Well done!'"³¹

Bukhari also preserves two lists of the four men who had "collected the Qur'an" at the Prophet's death, and the lists do not match:  Bukhari 5003 names Ubayy, Mu'adh, Zayd and Abu Zayd;  Bukhari 5004 names Abu al-Darda', Mu'adh, Zayd and Abu Zayd.³²  Jeffery's inference is the obvious one:  "As no two of these lists agree with one another to any great extent one is driven to conclude that while it was known that such collections were made there was no accurate information, save with regard to a few names, as to who made them."²⁸

Ibn Mas'ud refused

The best-attested report of the refusal is at the end of Tirmidhi 3104 (Darussalam:  Sahih), on al-Zuhri's authority:

'Abdullah bin Mas'ud disliked Zaid bin Thabit copying the Musahif, and he said:  "O you Muslim people!  I am removed from recording the transcription of the Mushaf and it is overseen by a man, by Allah, when I accepted Islam he was but in the loins of a disbelieving man" — meaning Zaid bin Thabit — and it was regarding this that 'Abdullah bin Mas'ud said:  "O people of Al-'Iraq!  Keep the Musahif that are with you, and conceal them.  For indeed Allah said:  And whoever conceals something, he shall come with what he concealed on the Day of Judgement (3:161).  So meet Allah with the Musahif."²⁵

Al-Zuhri closes the report by noting that "some men amongst the most virtuous of the Companions of the Messenger of Allah disliked that view of Ibn Mas'ud."  Ibn Abi Dawud gives Ibn Mas'ud's opposition a chapter of its own, and his eventual acquiescence a second chapter.  In the first, Hudhayfa says publicly that he will urge the caliph to "drown" the rival readings, and Ibn Mas'ud answers:  "By God, if you do, God will drown you in something other than water."²⁷

What was in his codex.  Al-Suyuti states it flatly:  "In Ibn Mas'ud's codex there are a hundred and twelve surahs, because he did not write the mu'awwidhatan" — Quran 113 and 114.³³  Jeffery:  "It was well known in the early days of Islam that one peculiarity of Ibn Mas'ud's Codex was that it did not contain Suras I, CXIII and CXIV, i.e. the Fatiha, which is an opening prayer to the book, and the Mu'awwidhatani with which it ends."²⁸

And here the Muslim answer is strong.  Ibn al-Nadim, a Baghdad bookseller writing in the fourth/tenth century, says he had personally seen numerous codices ascribed to Ibn Mas'ud and had been unable to find any two in agreement with each other — including one from the second century that did contain al-Fatiha.¹⁸  That is a Muslim bibliographer, not an apologist, reporting that the "codex of Ibn Mas'ud" was already a moving target three hundred years before al-Suyuti wrote it up as a fact.  Two rival lists of his surah order survive, from Ibn al-Nadim and from al-Suyuti, and they do not agree with each other.²⁸

Ubayy and the two extra surahs

Al-Suyuti reports that Ubayy ibn Ka'b's codex had a hundred and sixteen surahs, "because he wrote at the end of it the two surahs al-Hafd and al-Khal'."³³  He then gives their Arabic.  In translation:

Surat al-Khal' — "O God, we seek Your help and Your forgiveness, and we praise You and do not disbelieve in You;  and we cast off and forsake whoever transgresses against You."

Surat al-Hafd — "O God, You alone we worship, and to You we pray and prostrate, and to You we hasten and hurry;  we hope for Your mercy and fear Your punishment;  Your punishment surely overtakes the unbelievers."

Any Muslim will recognise these immediately.  They are the qunut supplications recited in the witr prayer to this day.  Al-Suyuti records that Umar recited them in his qunut, with the basmala before each;  that Ibn Jurayj drew the inference that "the point of the basmala is that they were two surahs in the codex of some of the Companions";  and that al-Tabarani reports "with a sound chain" that Umayya ibn 'Abdillah led prayer in Khurasan reciting these two.³³

Even the count of the extra material is disputed inside the tradition, and al-Suyuti stages the dispute on the next page but one.  He reports a correction — "a group transmitted from Ubayy's codex that it was a hundred and sixteen surahs;  the correct figure is a hundred and fifteen, because Surat al-Fil and Surat li-Ilaf Quraysh were a single surah in it" — attributed to al-Sakhawi's Jamal al-qurra', from Ja'far al-Sadiq and Abu Nahik.  Then he rejects it in his own voice:  "I say:  it is refuted by what al-Hakim and al-Tabarani reported from the hadith of Umm Hani," in which the Prophet says God "sent down concerning them a surah of the Quran in which He mentioned nobody with them:  li-Ilaf Quraysh."³³  So the figure the article prints is 116, which is al-Suyuti's, and the 115 is a competing count he names and answers.

What Jeffery's Materials is, and what it is not

Arthur Jeffery's Materials for the History of the Text of the Qur'an:  The Old Codices (Leiden, 1937) is the book Christian polemics reaches for here, and it is almost always misused.  It prints a 324-page apparatus of variant readings organised by codex — fifteen "primary" codices, thirteen "secondary" ones, and a section of unnamed ones — with 88 pages given to Ibn Mas'ud alone and 65 to Ubayy.¹⁸

None of it is manuscript evidence.  Jeffery calls the collected variants "our sole witness to the types of text which 'Uthman's standard text superseded" precisely because of "the absence of any direct manuscript evidence."²⁸  His own list of sources is a list of commentaries, dictionaries and books on the readings — Abu Hayyan, al-Baydawi, Fakhr al-Din al-Razi, al-Farra', Ibn Manzur, Sibawayh, al-Tabari, al-Zamakhshari, and two dozen more.²⁸  Not one manuscript.

His own doubts are stronger than anything a Muslim critic has said about him:

The question arises, of course, as to the authenticity of the readings ascribed to these Old Codices.  In some cases it must be confessed there is a suspicion of readings later invented by the grammarians and theologians being fathered on these early authorities in order to gain the prestige of their name.  This suspicion is perhaps strongest in the case of distinctively Shi'a readings that are attributed to Ibn Mas'ud, and in readings attributed to the wives of the Prophet.²⁸

He adds that "on the whole, however, one may feel confident that the majority of readings quoted from any Reader really go back to early authority," but also that "Being uncanonical variants there was none of the meticulous care taken over their transmission such as we find for the canonical readings… Some of the variants in the form in which they have survived to us seem linguistically impossible."²⁸  And he concedes that what survives is filtered:  only "such readings as were useful for purposes of Tafsir and were considered to be sufficiently near orthodoxy to be allowed to survive."²⁸  That cuts both ways.  The surviving list understates the original divergence — and it is exactly the kind of material most likely to have been generated by exegesis in the first place.

Al-Azami's methodological critique is correct and should be conceded:  Jeffery's sources usually say a reader recited a thing, not that his codex contained it;  most entries carry no chain of transmission;  and much of the material runs through channels the Sunni tradition regarded as unreliable.¹⁸  Western scholarship has since moved the same way for different reasons.  Van Putten argues that the canonical readers were largely reading off the standard skeleton rather than transmitting an independent oral text, and offers a test case:  Hamza traces his reading back to Ibn Mas'ud, but Ibn Mas'ud's reported reading does not agree with the Uthmanic skeleton and Hamza's does.  "The most likely explanation for this discrepancy is that Ḥamzah purposely changed his reading in order for it to agree with the rasm."²

The honest summary:  the companion codices were real, and Sanaa 1 proves it (below).  But the lists of their readings are literary reports of what medieval scholars said early authorities recited.  They are not a collation of codices, because no codex of Ibn Mas'ud or Ubayy has ever been found.  Any argument that treats Jeffery's apparatus as a set of manuscript variants deserves to be dismissed, and will be.

Material the tradition says is no longer in the text

This is the section that decides the letter-perfect claim, and every report in it comes from collections Sunni Muslims accept without reservation.

The stoning verse

Sahih al-Bukhari 6830, in the Book of hudud — the fixed criminal penalties — gives Umar's sermon from the pulpit in Medina in full.  It contains two separate claims of lost wording:

Allah sent Muhammad with the Truth and revealed the Holy Book to him, and among what Allah revealed, was the Verse of the Rajam …  and we did recite this Verse and understood and memorized it.  Allah's Messenger did carry out the punishment of stoning and so did we after him.  I am afraid that after a long time has passed, somebody will say, "By Allah, we do not find the Verse of the Rajam in Allah's Book," and thus they will go astray by leaving an obligation which Allah has revealed. …  And then we used to recite among the Verses in Allah's Book:  "O people!  Do not claim to be the offspring of other than your fathers, as it is disbelief (unthankfulness) on your part that you claim to be the offspring of other than your real father."³

The shorter version is at Bukhari 6829.³⁴  Neither carries a grade line, because inclusion in Bukhari is itself the claim.  Umar is not a hostile witness.  He is the second caliph, testifying publicly, in the collection Muslims use for law, that two passages were once recited as Quran and are not in the Quran now.

The suckling verse

Sahih Muslim 1452a:

'A'isha reported that it had been revealed in the Holy Qur'an that ten clear sucklings make the marriage unlawful, then it was abrogated (and substituted) by five sucklings and Allah's Apostle died and it was before that time (found) in the Holy Qur'an (and recited by the Muslims).⁴

That is in Muslim.  It says a ruling of five sucklings was in the Quran and was being recited at the Prophet's death.  It is not in the Quran now.

The story most often retold in English is the weaker one.  Sunan Ibn Majah 1944 — graded Hasan by Darussalam, again al-Albani's twentieth-century judgement — has Aisha say:  "The Verse of stoning and of breastfeeding an adult ten times was revealed, and the paper was with me under my pillow.  When the Messenger of Allah died, we were preoccupied with his death, and a tame sheep came in and ate it."³⁵  The Arabic word for the animal is dajin, a domestic animal kept in the house;  it is usually retold as a goat.  Sunnah.com prints its own footnote on that hadith:  "These verses were abrogated in recitation but not ruling.  Other ahadith establish the number for fosterage to be 5."  The goat is a memorable detail and it is the weaker report.  Argue from Muslim 1452a instead.

Two whole surahs

Sahih Muslim 1050.  Abu Musa al-Ash'ari, addressing three hundred reciters at Basra:

We used to recite a surah which resembled in length and severity to (Surah) Bara'at.  I have, however, forgotten it with the exception of this which I remember out of it:  "If there were two valleys full of riches, for the son of Adam, he would long for a third valley, and nothing would fill the stomach of the son of Adam but dust."  And we used to recite a surah which resembled one of the surahs of Musabbihat, and I have forgotten it, but remember (this much) out of it:  "Oh people who believe, why do you say that which you do not practise" and "that is recorded in your necks as a witness (against you) and you would be asked about it on the Day of Resurrection."⁵

Surat Bara'a — Surat al-Tawba — has 129 verses.  A companion of the Prophet is reported in Sahih Muslim saying that a surah of that size had been recited and was forgotten, within living memory of the man who recited it.

A surah that shrank

Al-Suyuti, reporting from Malik, says that when the beginning of Surat Bara'a fell away, the basmala fell away with it, and adds:  "for it is established that it used to equal al-Baqara in length."³³  Surat al-Tawba has 129 verses.  Surat al-Baqara has 286.

A related and much-repeated claim — that Surat al-Ahzab once ran to 200 verses — is deliberately omitted from this article.  It is normally cited to Ibn Abi Dawud and to al-Suyuti's al-Durr al-manthur.  Neither locator could be verified in the editions consulted for this research, and an unverified locator is not evidence.  What is verified is the smaller and better-attested report:  a verse of al-Ahzab went missing and was recovered from one man (Bukhari 4988, above).

"We leave some of what he recites"

Sahih al-Bukhari 5005 is the hinge on which the whole question turns:

Narrated Ibn Abbas:  Umar said, "Ubayy was the best of us in the recitation (of the Qur'an) yet we leave some of what he recites.  Ubayy says, 'I have taken it from the mouth of Allah's Messenger and I will not leave it for anything.'  But Allah said None of Our Revelations do We abrogate or cause to be forgotten but We substitute something better or similar (Quran 2:106)."⁶

Read that carefully.  Umar concedes that Ubayy is the best reciter.  He concedes that Ubayy has it from the Prophet's own mouth.  He does not deny the material.  He appeals to a doctrine.

One honest caveat, because this article's own rule is not to rest a case on a single English rendering.  The Arabic of the clause is wa-inna la-nada'u min lahni Ubayy, and the word doing the work is lahn.  The Muhsin Khan translation on sunnah.com renders it "what he recites."  Lahn is not a plain word for "verses":  its range in classical Arabic runs from a manner or melody of speech, through a dialect, to an error of speech.  Nothing in this report puts a number or a wording on what was left out, and a Muslim reader is entitled to say that Umar may be talking about Ubayy's manner of reciting rather than about lost text.  What makes the passage still count is not the word by itself but the answer Umar gives:  he does not say Ubayy is mispronouncing, he cites Quran 2:106 on abrogation, which is an argument about material being replaced, not about accent.  That is why this report belongs beside the stoning verse and the suckling verse rather than standing alone.

The tradition's own answer: abrogation of the recitation

The doctrine Umar reaches for at Bukhari 5005 is naskh al-tilawa — abrogation of the recitation, as distinct from abrogation of the ruling.  Its anchor is Quran 2:106:  "We do not abrogate a verse or cause it to be forgotten except that We bring forth [one] better than it or similar to it."⁷

Applied across the reports above, it works like this:

This is a coherent theological answer, it is old, and it is not an ad hoc invention — it is already the answer Umar gives in the earliest report we have of the problem.  A Muslim is entitled to it.

But it has a price, and the price is exactly the popular claim.  Naskh al-tilawa concedes that the Quran recited today is not identical to the Quran recited in 632.  It relocates the difference from human error to divine intention.  That answers the charge of corruption.  It does not leave "not one letter has changed" standing.  A dawah speaker cannot use both.

It is worth adding, for fairness, that the most aggressive Western argument on this point cuts against the polemical use of the same evidence.  John Burton argued that the whole naskh al-tilawa doctrine, and the collection stories with it, were manufactured by the jurists — and concluded that the text we hold is therefore older than the stories about it:

The single vigorous Qur'an text that throughout the ages has successfully withstood the assaults of both the exegetes and the usulis, stoutly retaining its textual identity in the face of countless attempts to insinuate interpolations through exploitation of the alleged codex of this or that Companion, is none other than the unique text of the revelations whose existence all their tricks betoken, the text which has come down to us in the form in which it was organised and approved by the Prophet. …What we have today in our hands is the mushaf of Muhammad.²⁴

Burton's thesis is a minority position and the manuscript evidence has moved past it.  But it is a reminder that "the lost-verse reports are unhistorical" and "the Quran is unreliable" are not the same conclusion, and that a critic who wants both is arguing against himself.

The seven ahruf

Muslim tradition has an explanation for why early recitation varied:  the Quran was revealed in seven ahruf — seven "modes" or "letters."  The reports are in both Sahih collections.

Bukhari 4991:  "Gabriel recited the Qur'an to me in one way.  Then I requested him (to read it in another way), and continued asking him to recite it in other ways, and he recited it in several ways till he ultimately recited it in seven different ways."³⁶  The same report appears at Bukhari 3219.³⁷

Bukhari 4992 is the famous incident in which Umar hears Hisham ibn Hakim reciting Surat al-Furqan differently, drags him to the Prophet, and is told:  "Release him …  Recite, O Hisham!"  Both men recite;  both are told "It was revealed in this way";  and the Prophet concludes, "This Qur'an has been revealed to be recited in seven different ways, so recite of it whichever (way) is easier for you."³⁸  Sahih Muslim 818a carries the same incident.³⁹

Two further reports matter.  Sahih Muslim 819a carries al-Zuhri's own gloss:  "It has reached me that these seven styles are essentially one, not differing about what is permitted and what is forbidden."⁴⁰  And Tirmidhi 2944 (Darussalam:  Hasan) gives Gabriel's stated reason:  "O Jibra'il!  I have been sent to an illiterate nation among whom are the elderly woman, the old man, the boy and the girl, and the man who cannot read a book at all."⁴¹

So far, so clear:  the revelation was given in more than one wording, deliberately, as a concession to a population that could not read.

What is not clear is what "seven ahruf" means, and the tradition says so itself.  Al-Suyuti, writing his standard handbook of Quranic sciences in the ninth/fifteenth century, opens the section with:  ikhtulifa fi ma'na hadha l-hadith 'ala nahwi arba'ina qawlan — "There is disagreement about the meaning of this hadith amounting to about forty opinions."³³  He then lists them.  They range from "the meaning is unknown" (Ibn Sa'dan the grammarian) through "seven is an idiom for 'many'" through Ibn Qutayba's seven categories of variation to the most widely held view — seven different words for the same thing, aqbil, ta'al, halumma, 'ajjil, asri' — which al-Suyuti ascribes to Sufyan ibn 'Uyayna, al-Tabari, Ibn Wahb "and multitudes."³³

Al-Suyuti then reports, on Ibn Hajar's authority, that Ibn Hibban had counted thirty-five opinions, and prints all thirty-five.³³  Al-Azami's own count agrees with the larger figure:  "forty scholarly opinions exist as to the meaning of ahruf."¹⁸

Immediately after the thirty-five, al-Suyuti quotes al-Mursi's complaint, and it is the single most useful sentence in the whole subject:

Most of these views overlap;  I do not know their basis nor from whom they were transmitted, nor why each writer restricted the seven ahruf to what he mentioned, when all of it is present in the Quran. …Many of the common people have supposed that what is meant by them is the seven readings, and that is an ugly ignorance.³³

That is a classical Muslim scholar, quoted approvingly in the standard Muslim handbook, saying flatly that identifying the seven ahruf with the seven canonical readings is ignorant.  It disarms the commonest move on both sides at once:  the Muslim who says "the seven readings are the seven modes, so nothing was lost," and the critic who says "the seven modes were seven different Qurans."

Did the Uthmanic codices keep all seven?  Al-Suyuti reports both views and gives the majority answer:  a minority of jurists and theologians said yes;  but "the great mass of the scholars of the salaf and the khalaf and the imams of the Muslims" held that the codices contain only what their skeleton can bear of the seven, and that they embody the final review the Prophet made with Gabriel.  Ibn al-Jazari's verdict on that position:  "this is the one whose correctness is evident."  Al-Tabari's answer to the obvious objection is that reciting on seven modes was a permission, not an obligation, and the Companions agreed to give it up when division threatened.³³

Read that plainly and it is a concession, made by the mainstream of Muslim scholarship, that the Uthmanic codex does not contain everything the seven ahruf contained.  The tradition's own majority position is that some of what was revealed as permitted recitation was set aside for the sake of unity.  That is defensible.  It is not "not one letter has changed."

The ten readings and how the number was fixed

The system in use today is not seven modes but ten qira'at, readings, each named for an eighth-century reader and transmitted through two named pupils.

The first seven are Ibn Mujahid's;  the last three complete the ten.²  Jeffery's independent list gives the same names, with one or two of the death years differing by a year — 'Asim at 128 rather than 127, Hamza at 158 rather than 156.²⁸

Three features of that table are worth noticing before anything else is said about it.

The names repeat.  Khalaf appears twice — once as Hamza's transmitter, once as an independent reader.  Al-Duri appears twice, transmitting for both Abu 'Amr and al-Kisa'i.  These are the same men.  For a picture of ten independently mass-transmitted lines, that is a small but real embarrassment.

The number two is a later convenience.  Van Putten, reporting Shady Nasser's work, notes that Ibn Mujahid himself "reports no less than 49 immediate transmitters" for his seven readers.  The reduction to exactly two per reader "seems to have been first introduced by Abu al-Tayyib b. Ghalbun (d. 389/998), but really took off when Abu 'Amr al-Dani (d. 444/1052–1053) wrote his al-Taysir fi al-Qira'at al-Sab', and al-Shatibi (d. 590/1194) summarized it into a didactic poem."²  Fourteen transmissions is a fifth-century pedagogical arrangement, not a first-century fact.

The canon was closed with the help of a court.  Jeffery dates it to "the fixing of the text ne varietur by the Wazirs Ibn Muqla and Ibn 'Isa in 322 A.H. at the insistence and with the help of the savant Ibn Mujahid (d. 324), and the examples made of Ibn Miqsam (d. 362) and the unfortunate Ibn Shanabudh (d. 328) who persisted in making use of the old readings."²⁸  The trials are documented on the Muslim side too:  al-Azami quotes Ibn Shanabudh's recantation and records that he was given ten lashes, and that Ibn Miqsam was made to repent for holding that any reading which fitted the Uthmanic skeleton and Arabic grammar was valid without a transmitted chain.¹⁸

The three criteria.  Ibn al-Jazari (d. 833/1429) canonised the ten in al-Nashr fi al-qira'at al-'ashr, and stated the test in that book's introduction:

Every reading that (1) accords with Arabic even in one respect, and (2) accords with one of the Uthmanic codices even potentially, and (3) has a sound chain of transmission — that is the sound reading, which may not be rejected and which it is not permitted to deny;  rather it is one of the seven ahruf in which the Quran was revealed …  And whenever one of these three pillars fails, it is called weak, or anomalous, or false.⁴²

Two of those clauses do most of the work, and both are deliberately loose.  "Even in one respect" is defined by Ibn al-Jazari on the next page as any grammatical construction at all, including ones "which some or many of the grammarians rejected, but their rejection was not considered."⁴²  "Even potentially" means agreement with the written skeleton either explicitly or by inference — and his own example is malik against malik at Quran 1:4, "for it is written without an alif in all the codices;  the short reading fits it directly, and the long reading fits it by supposing the alif was omitted for brevity."⁴²

Notice that the second criterion is written in the plural.  "One of the Uthmanic codices" — because Ibn al-Jazari knows perfectly well that Uthman's copies differed among themselves.

Hafs and Warsh: what actually differs

Of the twenty transmissions, two are in ordinary printed use.  Hafs from 'Asim is what a Muslim in Cairo, Istanbul, Karachi or Jakarta will open.  Warsh from Nafi' is what a Muslim in Rabat, Algiers or Dakar will open.  Both are printed under official Islamic auspices;  neither is a fringe text.

Adrian Brockett collated the two, transmission against transmission, for his 1984 St Andrews doctorate.  His chapter 10 lists the differences for suras 1–5, sorted by type rather than by verse, and counting his entries gives 138 — 94 in the vocal form, where the consonantal skeleton is identical, and 44 in the graphic form, where the written letters themselves differ.  Brockett himself states no total, so that figure is a tally of his lists and not a number he prints.¹⁰  These are types at first occurrence:  Brockett says explicitly that when the same difference recurs elsewhere, only the first instance is listed.  So the number of actual occurrences is larger.

Here is a sample of real rows, Hafs first, Warsh second.

Quran 3:146 is the row worth dwelling on, because both readings are old and both carry theological weight.  Al-Qurtubi sets out the two and reports what hung on them.  Qutila — "many a prophet was killed" — is the reading of "Nafi', Ibn Jubayr, Abu 'Amr and Ya'qub, and it is the reading of Ibn 'Abbas, and Abu Hatim chose it."  On that reading al-Hasan al-Basri objected, ma qutila nabiyyun fi harbin qattu — "no prophet was ever killed in war" — and Sa'id ibn Jubayr said "we have not heard that a prophet was killed in battle."  Qatada and 'Ikrima held the opposite.⁴³  So the two printed Qurans of the Muslim world disagree about whether prophets die in battle, and classical scholars argued about it.

A widely repeated example that is not a Hafs/Warsh difference at all.  Quran 2:222, on when a husband may resume relations after menstruation, is constantly cited as a Hafs/Warsh case.  It is not one:  Warsh agrees with Hafs there, and Brockett's list for suras 1–5 does not contain it.  Al-Qurtubi gives the actual line-up:  Nafi', Abu 'Amr, Ibn Kathir, Ibn 'Amir and 'Asim in Hafs's transmission read yathurna;  Hamza, al-Kisa'i, and 'Asim in the transmission of Shu'ba and of al-Mufaddal read yattahharna;  the codices of Ubayy and Ibn Mas'ud had yatatahharna;  and the codex of Anas ibn Malik had a different sentence entirely — "Do not approach women in their menstruation and keep away from them until they purify themselves."⁴³

Three things follow from that one verse, and all three matter.

It is a live legal difference.  Yathurna means the bleeding has stopped.  Yattahharna means she has performed the ritual washing.  Whether a husband may resume relations at the end of the bleeding, or only after the ghusl, turns on whether one letter is doubled.  Al-Qurtubi reports that al-Tabari preferred the doubled reading.

The split runs between two transmitters of the same reader.  Hafs and Shu'ba both transmit from 'Asim of Kufa, and here they disagree.  Van Putten's comment is blunt:  "The differences between Hafs and Shu'bah, for example, are so numerous that they disagree with one another more often than two separate readers like Hamzah and al-Kisa'i."²

A canonical Sunni commentator prints a companion codex with a materially different sentence.  Whether Anas's wording is authentic is a separate question.  That al-Qurtubi records it is not in question at all.

How many differences in total between Hafs and Warsh?  No reliable published whole-Quran count could be found for this article, and rather than manufacture one by multiplying Brockett's 138 by six, this article declines to give a figure.  What can be said with locators is this.  Brockett found 138 listed differences in suras 1–5, plus seven categories of "consistent differences" that run throughout — chiefly the glottal stop, which Warsh has far less of, and vowel-fronting — of which he says "none of these differences has any effect on the meaning."¹⁰  Across all the canonical readings, not just these two, van Putten writes that differences in specific wording "still concern thousands of words," while changes in sound and form affect "the vast majority of the verses."²  Al-Azami reports that Ibn Mujahid counted "roughly one thousand multiple readings only" in the whole Quran, and explains in his endnote that these are due chiefly to varying placement of dots and marks.¹⁸  And one count that is precise:  van Putten, reporting Hythem Sidky's work, says "the canonical reading traditions of the Quran only disagree with each other 284 times on consonantal dotting," adding that "considering the thousands of times that the readers could have plausibly disagreed on the dotting but do not do so, the number is actually strikingly low."²

Do the readings touch the written skeleton?

Both popular answers are wrong.  The Muslim popular answer is "only vocalisation, never the letters."  The critical popular answer is "the text itself is different."  The truth is in between and the tradition states it openly.

Some differences live entirely in the dots.  Early Arabic script often left off the marks that distinguish one consonant from another, so the same skeleton could be read several ways.

Van Putten's reading of these is the serious challenge to the Muslim account of how preservation worked:  "In such cases, the most likely explanation as to why the readers disagree is not that they were transmitting an oral transmission, but rather that the readers were confronted with an ambiguous rasm and interpreted it in two ways that both made semantic sense."²

And some differences are in the letters themselves — because the Uthmanic copies differed from each other.  Ibn al-Jazari explains his own second criterion by listing them:  Ibn 'Amir's reading qalu ttakhadha llahu waladan in Surat al-Baqara without the waw, "for that is established in the Syrian codex";  Ibn Kathir's reading with an added min at the last such place in Surat Bara'a, "for that is established in the Meccan codex";  the reading of Surat al-Hadid with huwa deleted;  sari'u without the waw;  and minhuma munqalaban in the dual in Surat al-Kahf.  And then the crucial rider:  "had that not been so in any of the Uthmanic codices, reciting it would be anomalous, because it would contradict the agreed rasm."⁴²

Ibn Abi Dawud preserves the actual list, in a chapter headed "the differences between the provincial codices copied from the master copy," opening with al-Kisa'i's own enumeration.  The first dozen entries include:

²⁷

Now compare that list with the Hafs/Warsh table above.  Quran 2:132, 3:133 and 5:54 are three of the four skeletal differences Brockett found between his printed Hafs and Warsh copies in 1984 — the fourth, Quran 5:53, is of the same kind as 3:133, a waw present in one regional codex and absent in another.  They line up because Warsh transmits from Nafi' of Medina and Hafs from 'Asim of Kufa.  Fourteen centuries later, the printed books in Rabat and Cairo still carry the regional differences of the seventh-century codices, in the right regional pattern.  That is a genuinely impressive result and it belongs to the Muslim side of the ledger.

Brockett also documents that the classical commentators were untroubled by these.  On Quran 2:132 he quotes Ibn al-Jazari — "Nafi', Abu Ja'far and Ibn 'Amir read wa'awsa, which was how it was in the texts of the Medinese and the Syrians.  The rest of the 'ten' read wawassa which was how it was in their texts" — and al-Farra':  "wawassa… In the texts of the Medinese wa'awsa.  Both are correct and commonly heard."¹⁰

How many such skeletal differences are there?  Give the range, not a number.  Van Putten says "about 40 variants in the consonantal skeleton" in one place in his book and "about 60 locations" in another.²²  Al-Azami says "forty-four places in the Quran, a divergence of forty-five characters in a corpus nine thousand lines long."¹⁸  These are the same order of magnitude and the disagreement is about how to count.

One reading that never became canonical.  Al-Qurtubi, on Quran 33:6, records:  "in the codex of Ubayy ibn Ka'b:  'and his wives are their mothers, and he is a father to them.'  And Ibn 'Abbas read:  'from among themselves, and he is a father to them, and his wives are their mothers.'"⁴³  Four extra words, and directly relevant to Islamic law, since Quran 33:40 denies that Muhammad was "the father of any of your men."  That reading is shadhdh, anomalous:  it fails Ibn al-Jazari's second criterion.  The system is working exactly as designed.  But notice what the criterion is.  It is agreement with the Uthmanic skeleton — a test of conformity, not a test of antiquity.  A reading older than Uthman's committee would fail it just as surely as a reading invented in Baghdad.

The 1924 Cairo edition

Almost every printed Quran a reader is likely to meet descends from one edition, completed in Cairo on 7 Dhu al-Hijja 1342 / 10 July 1924.

It is not a typeset critical edition.  It is a lithographic facsimile of a manuscript in the hand of Muhammad "al-Haddad" ibn 'Ali ibn Khalaf al-Husayni, the Shaykh al-Maqari of the day.¹⁰  Work began around 1907 under 'Abbas II Hilmi, not under King Fu'ad, whose patronage came later.¹⁰  Its own colophon gives a pedagogical reason:  Quran copies for schools had been imported from abroad, many contained errors and had to be destroyed, and Egypt decided to produce its own text in the Uthmanic orthography.¹⁰

Three things about it are commonly misunderstood.

It reproduces one reading, not a reconstruction.  Gabriel Said Reynolds:  "The scholars who worked on that project did not seek to reconstruct the ancient form of the Qur'an, but rather to preserve one of the canonical qira'at… that of Hafs (d. 180/796) 'an 'Asim (d. 127/745).  But these qira'at are part of the history of the text, not its starting point."⁴⁴

Its basis was books, not manuscripts.  The committee worked from the medieval literature on Quranic orthography — al-Dani above all — rather than from old codices.  Brockett notes that it "claimed to have made a break with the continuous written Tradition and to have gone back to the original text of the caliph 'Uthman," by restoring spellings "according to the oral Tradition about the orthography of the Qur'an."¹⁰  Reynolds credits the thoroughness:  "they conducted this research with great thoroughness and attention to detail… Gotthelf Bergsträsser, for example, noted that in only a small number of cases is their reading contradicted by earlier sources on Hafs 'an 'Asim."⁴⁴  Van Putten's assessment is that the edition's orthography "is very conservative, and often is a fairly accurate representation of what we find in 7th century Quranic manuscripts," but that it over-uses one letter compared with the early manuscripts and innovates in places.²

It should not be called "the official Quran."  Brockett argues against the words "official," "standard" and "edition" alike:  it was official only in a limited Egyptian sense;  al-Azhar has since authorised other facsimiles with different conventions;  Saudi Arabia and Qatar print texts "differing considerably in orthography from the 1342 Cairo text… under official approval";  and "the term 'edition' implies editorial activity.  And that was scarcely even held to have been performed by 'Uthman, let alone by Egyptian scholars of only sixty years ago."¹⁰  Critics who call it "the official Quran" are overstating, and Brockett — not a Muslim apologist — is the one correcting them.

A note on the Nile story.  It is often said that the Egyptian government sank the older imported copies in the Nile.  The claim traces to one sentence in Reynolds:  "In response, the government destroyed a large number of such texts by sinking them in the Nile River and issued its own text."⁴⁴  His footnote cites Bergsträsser's 1932 article, which was not opened for this research, so the detail here is at one remove from its source.  Brockett, who treats the edition's history in far more detail and cites Bergsträsser repeatedly, does not mention the Nile at all.  Treat it as unconfirmed.

The manuscripts

Everything above comes from reports.  The manuscripts are a separate line of evidence, and they were unavailable to every scholar quoted so far except the moderns.

One archetype

Van Putten's archetype argument, set out above, does not depend on trusting any Muslim report or any Western reconstruction.  It depends only on the fact that some Quranic spellings carry no information, so their distribution should be random unless every copy came from one written parent.  Across fourteen early manuscripts it is not random.¹  His summary in the 2022 book states the position of the whole field:  "All early manuscripts known so far descend from a single text type referred to as the Uthmanic Text Type (ut), which may plausibly be connected with the standardization effort attributed to the third caliph ʕuṯmān b. ʕaffān (reign 23–35ah/644–656 ce).  While the manuscript evidence cannot preclude with absolute certainty an earlier redaction, a time much later than his reign is now a virtual impossibility."²

He is also careful about the limits, and an honest article must carry them.  The archetype is not identical to the Cairo edition:  a few of the Cairo edition's spelling quirks "seem to be a later innovation," and a few of the archetype's quirks were lost on the way.¹  And an archetypal spelling may simply be a mistake copied faithfully forever:  at Quran 21:88 and 12:110 a doubled letter is written singly, and van Putten allows that this may be "a writing error that has been propagated from the archetype."²

The live disagreement is fifty years wide.  Stephen Shoemaker accepts the single archetype and disputes its date:  "what van Putten's article appears to demonstrate is that all these early Qur'ans derive from a single written archetype;  however, in light of the best interpretation of the data that we have for the earliest Qur'anic manuscripts, that archetype appears much more likely to be the imperial Qur'an established and imposed by 'Abd al-Malik and al-Hajjaj."¹²  That is the whole distance between the traditional Muslim account, around 650, and the revisionist one, around 700.  Nicolai Sinai's verdict, quoted approvingly by Sidky, is that the traditional dating of 30/650 or earlier "ought to be our default view."¹⁴  This article follows Sinai, while noting that Shoemaker's objection is not frivolous:  it rests on the palaeographers repeatedly dating the writing later than the parchment.

Sanaa 1, the one manuscript from outside the family

In July 1973, restoration work on the Great Mosque of Sanaa opened a space between the roof and the false ceiling containing roughly fifteen thousand parchment fragments.  Among them was a Quran written over an erased Quran.⁴⁵

Behnam Sadeghi and Mohsen Goudarzi published a 129-page edition of the lower text in 2012.  It survives on 36 folios in Sanaa plus four that reached the auction market, and the erased writing is legible on 31 of the Sanaa folios plus the four strays.¹¹  The lower text runs continuously through whole surahs — all of sura 9, with gaps, preceded by part of sura 8 and followed by sura 19.

How different is it?  Neither Sadeghi paper gives a grand total;  they say a systematic analysis is postponed.¹¹  What exists are counted samples, and those are the numbers to use.

On the four auctioned folios, Sadeghi and Bergmann counted 35 differences in minor elements — particles, articles, conjunctions, prepositions, pronouns, prefixes and suffixes — made up of 12 omissions, 19 substitutions and 4 additions;  25 cases of major substitution, a different word or phrase, of which 18 sound like the standard reading and only 7 do not;  and, in pluses of a word or more, 14 in the standard text that Sanaa 1 lacks against 10 in Sanaa 1 that the standard text lacks.¹⁹  Sinai's independent tally over a different and wider set of folios found about eighteen pluses of one word or more in the lower layer against about six in the standard text, and about eighteen substitutions — a distribution running the opposite way from Sadeghi's sample, which Sinai himself flags.⁴⁶

For scale, Sadeghi and Goudarzi estimate that Sanaa 1 has "a lot more variants" than the reported codex of Ibn Mas'ud — "by a rough estimate perhaps twenty-five times as many."¹¹

Does anything legal change?  Yes, in small ways, and this article will not pretend otherwise.  Both of the clearest cases are in Quran 2:196, the pilgrimage verse.  The standard text reads "Do not shave your heads until the offering reaches its destination";  Sanaa 1 omits "your heads."  And the standard text lists the expiation as "fasting, or alms, or an offering";  Sanaa 1 has "fasting or an offering" — one of the three legal options is simply absent, and Sadeghi notes that neither version has a parallel elsewhere in the Quran, so this is not explicable as a copyist harmonising to a familiar phrase.¹¹

Two structural differences are also worth naming.  Quran 20:31–32, two three-word verses, appear in reverse order — and the literary sources report the same transposition in the codex of Ubayy.  And Quran 9:85 is missing entirely, sixteen words;  Sadeghi and Goudarzi treat this as eye-skip, since the end of verse 84 and the end of verse 85 carry an identical sequence, and note that it is a clear outlier against every other omission in the manuscript.¹¹

Does anything theological change?  No.  The substitutions that carry semantic weight are of this kind:  "on account of their wrongdoing" for "on account of their unbelief" at Quran 2:88;  "those who prosper" for "those who are guided" at Quran 9:18;  "the fire" for "hell" at Quran 9:73.⁴⁶  No variant adds or removes a doctrine.  A Christian apologist who claims the Sanaa palimpsest shows the Quran teaching something different is overstating, and can be refuted from the published edition in five minutes.

The chapter order is different.  The eleven surviving transitions give the runs 11 → 8 → 9 → 19;  12 → 18;  15 → 25;  20 → 21;  34 → 13;  39 → 40;  63 → 62 → 89 → 90.  Some match Ubayy's reported order, some Ibn Mas'ud's, some neither;  as a whole it is closer to Ubayy and Ibn Mas'ud than to Uthman.¹¹

And two facts cut hard the other way.  Sadeghi's own finding is that "C-1 has the same verses and the same order of verses as the standard Qur'ān," subject to the two exceptions named just above, and he draws the inference himself:  "even a careful memorizer who reproduces the words exactly is prone to getting the order of the verses wrong."  Verse order is exactly what a memory-based transmission would get wrong first, and here it is almost entirely right.¹¹  And the single famous anomaly of the standard Quran, the transition from sura 8 to sura 9 without the opening formula, is already present in Sanaa 1 — so it was not Uthman's committee's decision after all.¹¹

What Sanaa 1 settles.  John Wansbrough and John Burton had dismissed the companion codices as fictions — exegesis retrojected into scripture.  Sanaa 1 refutes that:  it is "direct documentary evidence for the reality of the non-'Uthmānic text types that are usually referred to as 'Companion codices.'"¹¹  But it is not Ibn Mas'ud's or Ubayy's:  "C-1 does not share the vast majority of its variants with these codices.  Nor are most of their variants found in C-1."¹¹

Asma Hilali's counter-thesis, and why the field rejected it.  Hilali argued in 2017 that neither layer was ever part of a complete codex — that the lower text is "probably a scribal exercise" produced in a teaching circle.  Sinai's answer is that her markers of "unfinished work" are exactly what a real first-century codex should look like, and that her argument rests on "anachronistic misconceptions about the standards of scribal and decorative accuracy, consistency, and uniformity that would have been deemed appropriate, in the first Islamic century, to a full quranic codex" — the decisive counter-evidence being the "unfailingly sequential arrangement of verses."⁴⁶  He also reports that where Sadeghi and Goudarzi read a variant, "Hilali's edition tends to have blanks," calls the discrepancy "puzzling," and says ultraviolet images confirmed them against her on three occasions.⁴⁶  Van Putten notes that her book does not reproduce her photographs, so "we have only her word against that of Sadeghi and Goudarzi."¹  Déroche, who examined the object, called the lower Quran "commissioned by someone at great cost, despite its imperfections" — not a practice sheet.¹²

Note what that means for both sides.  A Muslim cannot appeal to Hilali to make Sanaa 1 go away, because the specialists rejected her.  A Christian cannot appeal to Sanaa 1 to make the Quran's theology change, because it does not.

One further manuscript is worth a sentence.  Codex Mashhad, described by Morteza Karimi-Nia in 2019, carries more than ninety per cent of the Quran on 251 folios.  Its text is the standard Uthmanic text;  its surah arrangement is Ibn Mas'ud's, and it was later re-sewn toward the standard order with the seams still visible.⁴⁷  Ibn Mas'ud's chapter arrangement is physically attested.  His text is not.

Birmingham, and what radiocarbon can and cannot do

In July 2015 the University of Birmingham announced that two Quran leaves in its Mingana collection had been radiocarbon dated.  The press release says, verbatim:  "Radiocarbon analysis has dated the parchment on which the text is written to the period between AD 568 and 645 with 95.4% accuracy.  The test was carried out in a laboratory at the University of Oxford."⁴⁸

Three corrections belong beside that sentence.

The number has moved.  The raw measurement is 1456 ± 21 years before present, published in the Oxford laboratory's own datelist.⁴⁹  Recalibrated with the current IntCal20 curve, the range is 578–646 CE, which is the figure Corpus Coranicum now publishes.  The two Birmingham folios also belong to one codex with sixteen folios in the Bibliothèque nationale de France — eighteen leaves in all — a join Alba Fedeli made before the radiocarbon test was run.⁵⁰

The claim in the press release goes further than the evidence.  David Thomas was quoted saying the text "can, with a degree of confidence, be dated to less than two decades after Muhammad's death.  These portions must have been in a form that is very close to the form of the Qur'an read today, supporting the view that the text has undergone little or no alteration."⁴⁸  The move from a parchment date to a text date is precisely the step the specialists reject.

The specialist is Fedeli, whose doctoral thesis is the primary study of these leaves, and who is not a hostile witness.  Her palaeographic judgement is that the scribe "was copying the text from an exemplar, and in accomplishing such a task, he expressed his mastery, e.g. in planning the page layout."  She continues:

The mechanism of copying from an exemplar implies consequently that the work could not have been executed very early, as the written exemplar requires a period of time for producing the exemplar and also the establishment of a mechanism of copying from an authoritative text.  Moreover, the regular and coherent presence of a blank line between two sūras seems to be interpreted as a sign of a later practice, as it was introduced and established during the so-called second maṣāḥif project accomplished in the period between 84-85 AH (703-705 CE), whose main initiator was al-Ḥajjāj (d. 95/713) as interpreted by Hamdan.¹³

And then she says it outright:  such a conclusion from the script and layout "disagrees with the results of the carbon 14 analysis, which reveals a high probability of an earlier date."¹³  She lists other mismatches of the same kind:  one Sanaa manuscript dated by carbon to 657–690 has "clear palaeographical features from 710-715 CE";  a group of giant Qurans that Déroche dated to the later eighth century came back 640–765;  a manuscript whose own endowment note reads 907 CE was carbon dated to 716–891.¹³

The radiocarbon results do not cohere with each other either.  Folio 13 of the Sanaa palimpsest was measured by four separate laboratories.  Recalibrated to IntCal20, the four answers are:

¹²

The Lyon result is a century before Muhammad was born.  Shoemaker's comment is fair:  "these assays have yielded significantly different results for the very same parchment leaf!  Such results hardly settle the matter but instead only confirm that, again, something clearly is not working correctly with this method."¹²  Déroche's reaction to the impossibly early Lyon figures was that "they cannot be accepted."¹²

Neither extreme survives this.  "Radiocarbon proves the Quran comes from Muhammad's lifetime" is wrong:  it dates an animal.  "Radiocarbon is worthless" is also wrong:  the measurements do cluster in the seventh century, and Sidky points out that the field's confidence in the method is considerably higher than critics of it allow.¹⁴  The honest position is Fedeli's:  the parchment is old, the writing on it is probably a generation or two younger, and neither fact rescues or wrecks anyone's theory.

One result is worth quoting exactly, because it is the one everyone cites.  The Arizona measurement on the auctioned Sanaa folio gave a radiocarbon age of 1,407 ± 36 years BP, calibrating to 614–656 CE at one sigma and 578–669 CE at two sigma.  Sadeghi computed the one-sided probabilities and concluded:  "The probability that the parchment is older than AD 646 is 75.1 %, or a three-to-one likelihood.  It is highly probable therefore, that the Ṣanʿāʾ 1 manuscript was produced no more than 15 years after the death of the Prophet Muḥammad."¹⁹  Note what that sentence is about:  the parchment, not the writing.

A claim to stop repeating.  It is often said that the two layers of the Sanaa palimpsest are separated by up to fifty years.  Nothing measures this.  No radiocarbon date exists for the upper text at all;  it is dated on script alone, to "sometime during the seventh or the first half of the eighth century."¹¹

Corrections in early manuscripts

Daniel Brubaker, a Christian scholar with a 2014 Rice dissertation on the subject, published Corrections in Early Qur'an Manuscripts:  Twenty Examples in 2019.  The descriptive fact he documents is true and uncontested:  early Quran manuscripts contain physical corrections — erasures, overwriting, insertions.  That is genuinely useful against the claim that not one dot has ever changed.

The inference he draws from them does not survive review.  Hythem Sidky's sixteen-page assessment in Al-'Usur al-Wusta walks through all twenty examples.  Its core objections:

The reasoning is backwards in time.  Brubaker treats corrections that move a manuscript toward the 1924 Cairo edition as evidence of a text evolving toward a standard.  Sidky:  "this is a teleological argument for an end goal that did not exist at the time," and "corrections apparently in the direction of conformity to the Cairo edition are not evidence of a changing standard, but evidence of the existence of a standard in the first place!"¹⁴

Most of the twenty are ordinary scribal blunders.  Sidky assigns seven of them to eye-skip, several of which produce ungrammatical Arabic without the correction — which is decisive, because no scribe deliberately introduces a solecism to conform to a rival reading.  On one example Brubaker notes that the inserted word "is not grammatically necessary";  Sidky replies that "its ungrammaticality is the obvious reason Brubaker 'found no mention of an issue at this spot in the qirāʾāt literature.'"¹⁴

There are factual errors.  Brubaker states that the earliest Hijazi manuscripts were written without diacritical marks;  Sidky:  "In reality, the very earliest Hijazi manuscripts contain occasional diacritical marks."¹⁴

Sidky's verdict:  "If the objective of Brubaker's book is to demonstrate the humanity of the scribes involved in transmitting the Quranic text, it certainly succeeds. …Where it falls short, however, is in its methodology and analysis. …In fact, the vast majority of examples in the book are best explained through simple scribal error.  The main thesis, namely, that the flexibility of the Quranic text persisted centuries beyond its standardization, remains unproven."¹⁴

Sidky is not writing a hit piece;  he closes by praising the presentation and asking for more descriptive work of the same kind.  And he concedes one genuinely interesting case:  a page in the Museum of Islamic Art, Doha, with eight points of divergence from the standard skeleton and five later corrections which, even after correction, still does not conform to the standard text.¹⁴  That is the kind of evidence a serious argument would be built on.

Guidance for anyone using Brubaker:  cite him for the fact of corrections, cite Sidky alongside him, and do not cite him for the claim that the text was still fluid centuries after standardisation.  A reader who checks will otherwise conclude the homework was not done.

What Keith Small actually concluded

Keith Small's Textual Criticism and Qur'an Manuscripts (2011) is the standard Christian-side treatment and is more restrained than its usual quotation.  His concluding judgement, in full:

While bearing testimony to the careful preservation of one particular consonantal text, the history of the transmission of the text of the Qur'ān is at least as much a testament to the destruction of Qur'ān material as it is to its preservation.  It is also testimony to the fact that there never was one original text of the Qur'ān.⁵¹

Quoting only the second sentence misrepresents him, and it is regularly done.  His defensible argument is the narrow one from the previous page:  the Uthmanic consonantal skeleton "has been very well preserved from the seventh/first century," but the vocalisation has not — "the earliest extant manuscripts have different diacritical point patterns and no vocalization marks," so "since it cannot be demonstrated that there was one version going back to Muḥammad, it also cannot be demonstrated that seven or ten recitations went back to him."⁵¹  The Seven and Ten systems cannot be documented in complete form before the fourth/tenth century.  Small is cited approvingly by Sinai on verse-division systems, so he is inside the academic conversation rather than outside it.

"Christians footnote, Muslims burn"

The joke runs:  when Christians find a variant they add a footnote;  when Muslims find a variant they light a bonfire.  It is memorable, it is half true, and repeating it as usually told costs more than it gains.

What is true.  Uthman did order the burning.  It is in Bukhari, in the Arabic, with an unambiguous verb, and it is the Sunni tradition's own account of how the text was standardised.  A Christian does not need a hostile source for this and should not use one.  Uthman's own critics, in Ibn Abi Dawud's collection, said he "erased the Book of God."  The evidential consequence is real:  where a tradition destroys its non-conforming copies, later scholars lose the ability to check the surviving one against anything.  If someone in Kufa in 660 changed a word, no surviving evidence could show it.

What is not true.  Four things.

Uthman did not tour the cities burning Qurans.  Bukhari 4987 describes copies being sent out from Medina with an order that the rest be burnt.  There was no tour.

"Variant Qurans" overstates what the reports describe.  The problem the reports name is ikhtilaf fi l-qira'a, differences in recitation, among troops from Syria and Iraq.  Some companion-codex differences were substantial;  the crisis as reported was recitational.

The destruction was not total, and everyone knew it.  Ibn Mas'ud publicly refused and told a province to hide their copies.  Muslim scholars then went on cataloguing companion readings for a thousand years, in books written for that purpose:  Ibn Abi Dawud's Kitab al-Masahif, al-Dani's works, Ibn al-Jazari's al-Nashr, the whole shawadhdh literature of anomalous readings.  Muslim scholarship produced an enormous written apparatus of variants.  It is where most of the evidence in this article comes from.  A joke whose punchline is that Muslims do not footnote variants is refuted by the footnotes in this article.

And the motive was not concealment.  On the tradition's own telling, Hudhayfa came back from the Armenian front alarmed that soldiers from different provinces were reciting differently, and asked Uthman to act "before they differ about the Book (Qur'an) as Jews and the Christians did before."  Whatever one thinks of the decision, the reported reason for it was to stop a quarrel, not to hide evidence.  A Christian who cannot state an opponent's stated motive fairly should not expect to be believed about anything harder.

The honest version of the point survives without the joke, and it is this:  the two traditions bought different things.  Islam bought uniformity and gave up verifiability.  Christianity kept verifiability and gave up uniformity.  Neither purchase was free.

Comparing with the New Testament, carefully

The comparison is unavoidable, because both sides make it, and it is usually made badly.  The two traditions have different transmission models with different failure modes, and a raw variant count means almost nothing across them.

Read across that table and the shape of the argument changes.  The Quran's variant count is low because its rivals were destroyed and its copies descend from one exemplar.  The New Testament's variant count is high because nobody destroyed anything and thousands of independent copies survive.  A low count is evidence of a successful standardisation, not by itself evidence of an accurate one.  A high count is evidence of many witnesses, not by itself evidence of corruption;  the half-million figure measures how often scribes copied, not how badly.

The Christian side has its own concessions to make here, and the companion article on the New Testament makes them:  the favourite apologetic numbers are inflated, there are no first-century New Testament papyri, and the first hundred years of copying cannot be directly checked.  See the New Testament manuscript evidence for the audit, and the article on tahrif for the mirror-image Muslim charge that the Bible was corrupted.

The fair statement is symmetrical.  Both traditions can reconstruct an early archetype with high confidence.  Neither can document the first two decades.  For the Quran the undocumented window is roughly 610–650;  for the New Testament it is roughly 50–125.  Anyone who says his own tradition's window is fine and the other's is fatal is not doing history.

The scorecard

Four questions, answered separately, because the answers differ sharply and running them together is what produces bad arguments on both sides.

Did the text change between Muhammad and Uthman?

Yes, on the tradition's own reports.  A verse of al-Ahzab was missing from Abu Bakr's sheets and had to be recovered from one man.  The end of Surat al-Tawba was found with one man only, and the tradition cannot agree whether it was one verse or two, or which man.  Umar testified that the stoning verse was recited and is absent, and that a second passage about false lineage was likewise recited and is absent.  Aisha said the five-sucklings ruling was in the Quran and being recited when the Prophet died.  Abu Musa said two whole surahs had been recited and forgotten.  Umar said of the best reciter among them, "we leave some of what he recites."  Uthman's committee was told to resolve disagreements in favour of the Qurashi dialect, which is an editorial instruction whatever else it is.

One item on that list has since been weakened by a manuscript, and the article says so rather than leaving it to a reader to find.  Sanaa 1 carries the last verses of Surat al-Tawba in their expected place, and Sadeghi draws the obvious conclusion:  "Since they are also found in the ʿUthmānic Qurʾān, and since it is not reported that any Companion codex was without them, these verses must have belonged to the prototype from which the C-1 and ʿUthmānic text types emerged.  Therefore, one should not read too much into the report."  The Khuzayma report may still be evidence about the process — that the committee had to hunt for material — but it is not evidence that anything was missing from the text.¹¹

None of that is a critic's reconstruction.  All of it is in Bukhari and Muslim, in the chapters on the virtues of the Quran and on criminal penalties, where Muslims read it for law.  The tradition's answer is naskh al-tilawa, and it is a real answer — but it is an answer that concedes the change and relocates its cause.

Did the text change between Uthman and today?

Very little at the level of the written skeleton, and the manuscripts show this independently of the tradition.  All early manuscripts so far examined descend from one written archetype.  The regional differences among Uthman's own copies — forty to sixty places, depending how you count — are still visible in the printed books of today, in the right regional pattern.  Jeffery, who had every professional motive to say otherwise, wrote in 1940 that the early codices "show the same type of text, such variants as occur being almost always explainable as scribal errors."  This must be said without hedging, because it is true.

Is the consonantal text stable?

Yes.  This is the best-attested proposition in the field and the Muslim side should be given it without argument.  Brockett's explanation of why it is stable is worth quoting in full, because it credits both halves of the Muslim account:

Thus, if the Qur'an had been transmitted only orally for the first century, sizeable variations between texts such as are seen in the hadith and pre-Islamic poetry would be found, and if it had been transmitted only in writing, sizeable variations such as in the different transmissions of the original document of the Constitution of Medina would be found.  But neither is the case with the Qur'an.  There must have been a parallel written transmission limiting variation in the oral transmission to the graphic form, side by side with a parallel oral transmission preserving the written transmission from corruption.¹⁷

Brockett even reprints, approvingly, William Muir's 1858 verdict:  "There is probably in the world no other work which has remained twelve centuries with so pure a text."  Read Muir's whole sentence before using it, because it is narrower than the version that circulates.  What Muir says has been handed down unaltered is "the recension of ʿUthman," and his twelve centuries are counted from Uthman, not from Muhammad.  The claim is about the stability of the standard text after 650, and it says nothing at all about what happened between 610 and 650.  Brockett notices the same thing and instructs the reader to substitute:  "For Muir's 'recension of ʿUthman', however, now read 'the Qur'an'."¹⁷

Is the recited text single?

No, and the tradition has never claimed it was.  Ten readings are canonical, with twenty transmissions.  Two are in ordinary print, differing in at least 138 listed places in the first five surahs, four of them in the letters themselves.  Two transmitters of the same reader disagree more often than two separate readers do, and at Quran 2:222 their disagreement carries a legal ruling.  Al-Suyuti records forty opinions on what the seven modes even were, and a classical scholar's flat verdict that identifying them with the seven readings is "an ugly ignorance."

Van Putten's linguistic assessment of the readings is severe and should be represented accurately rather than softened:  "none of these readings represent anything like natural language, or in fact any kind of language described by the Arab grammarians… The readings in many cases embrace artificial features for a certain artistic effect, which suggests a conscious attempt of the readers to beautify their recitation with unusual and exotic features."²

The strongest Muslim responses

The most substantial modern Muslim defence in English is M. M. al-Azami (1930s–2017), Professor Emeritus at King Saud University, Cambridge PhD 1966, King Faisal Prize 1980.  His The History of the Qur'anic Text from Revelation to Compilation answers most of this article directly, and four of his moves deserve a serious reply.

"The bonfire proves consent, not concealment"

Al-Azami argues that the destruction was uncontroversial at the time:  "Musʿab bin Saʿd asserts that the people were pleased with 'Uthman's decision;  at the very least no one voiced any objections.  Other reports confirm this unanimous approval, including 'Ali bin Abi Talib who says, 'By Allah, he only did what he did with these fragments in clear view of us all.'"  And then:  "This scene alone, of the Prophet's Companions in their thousands eyeing the bonfire as old Qur'anic fragments were tossed in, is a powerful testimony that they all assented to the purity of the Mushaf's text."¹⁸¹⁸

The reply.  The consent claim is not sustainable on the tradition's own reports.  Ibn Mas'ud publicly refused and told Iraq to hide their codices — that is in Tirmidhi, graded sound by Darussalam.  The Egyptian delegation's first grievance against Uthman, in Ibn Abi Dawud, is that he "erased the Book of God."  Jeffery reports Kufa divided into two factions over it.  Al-Azami's own book records the Ibn Mas'ud material.  "No one voiced any objections" is the one thing the sources do not say.

"Jeffery's variants are not manuscript evidence"

Al-Azami's methodological point is that Jeffery's sources say a reader recited something, not that his codex contained it;  that most entries carry no chain;  and that much of the material runs through transmitters the Sunni critics distrusted.¹⁸

The reply is agreement.  This is correct, and this article has said so.  Western scholarship has since moved the same way for different reasons.  Anyone building a case on Jeffery's apparatus as though it were a manuscript collation is building on sand.

"A manuscript that disagrees is by definition not Quran"

Al-Azami's decisive move:

If any researcher now brings forth a parchment with wording that is significantly different from 'Uthman's Mushaf, one whose consonantal backbone does not settle into the Mushaf's text, it cannot be accepted as Qur'an.  This is because it cannot share 'equal status' with the Mushaf, which was compiled from first-hand documents backed by personal testimony.  One is of known pedigree and the other is a complete unknown. …It may be pre-'Uthmanic or post-'Uthmanic or a curio or a fake, but it embodies none of the certainties of the 'Uthmanic text and therefore cannot be regarded as Qur'an.¹⁸

The reply.  This is coherent, and on its own premises it is defensible.  It should be named for what it is:  a rule that makes the Uthmanic text unfalsifiable by manuscript discovery.  Any find that agrees confirms preservation;  any find that disagrees is by definition not Quran.  A Muslim may fairly answer that this is the right rule for a community defining its own scripture — canon is a communal decision, not a text-critical one, and Christians make the same kind of decision about the Gospel of Thomas.  But it is not an evidential argument and must not be offered as one.  Sanaa 1 is the test case:  it exists, its wording differs, and the rule declares it irrelevant in advance.

"The Munich institute found no variants"

The strongest form of the preservation claim in al-Azami's book quotes a 1977 lecture by Muhammad Hamidullah:

Shortly before the Second World War, a preliminary and tentative report was published that there are of course copying mistakes in the manuscripts of the Qur'an, but no variants.  During the war, American bombs fell on this Institute, and all was destroyed, director, personnel, library and all…  But this much is proved – that there are no variants in the Qur'an in copies dating from the first to the present century.¹⁸

The reply must be careful, because al-Azami himself supplies it.  His own endnote identifies the source as a Tokyo lecture of April 1977, not a report of the Munich collation.¹⁸  And on the same page al-Azami records that the archive was not destroyed:  "the bombs that fell on that April night in 1944 did not destroy the archive.  The building was demolished and word was spread that the collection was indeed wiped out as well, but in secret the cache of photos survived under the care of the library's former curator.  Only in 2008 … was this widely publicized."  Those photographs are the groundwork of the Corpus Coranicum project in Berlin.  So the strongest "no variants" claim in the most serious Muslim book on the subject rests on a lecture describing a preliminary report on an archive that turned out to have survived.  That is worth saying plainly, and it is not a cheap shot:  al-Azami reports the survival honestly himself, which is to his credit.

Where al-Azami is right and should be credited

Two places.  His critique of Jeffery is sound.  And his account of the regional codices — forty-four places, forty-five characters in a corpus nine thousand lines long, none of the differences yielding a contradiction in meaning, with each surviving early manuscript falling on one side or the other of every one of them — is careful, specific, and independently corroborated by the manuscript work of Cook, Sidky and van Putten.¹⁸

A note on Mustafa Shah

The leading British academic writing on the readings from within a broadly traditional frame is Mustafa Shah of SOAS, whose Journal of Qur'anic Studies articles and Oxford Handbook of Qur'anic Studies chapter argue that the grammarians' engagement with readings long predates Ibn Mujahid.  None of his work could be obtained for this article — the publisher's site refused every request made.  His argument is therefore not represented here, and a reader who wants the strongest traditional-side academic case on the readings should go to him directly rather than trust this article's silence.

Common objections and responses

"Quran 15:9 promises God will preserve it, so it is preserved"

The verse promises preservation.  It does not say of what, by what means, or to what standard.  It is quoted by Muslims who believe the consonantal text is letter-perfect and by Muslims who accept that God abrogated some recited passages.  Both readings fit the Arabic.  A promise is only as specific as its wording, and this one leaves the scope open.  Note also that the noun is al-dhikr, "the reminder";  the identification with the Quran is the commentators', and it is reasonable, but it is not in the Arabic of the verse.

"The variants are only in the vowels, and Arabic vowels are not written anyway"

Not only.  Four of the differences between the two Qurans currently in print — at Quran 2:132, 3:133, 5:53 and 5:54 — are in the written letters, not the vowels.  Ibn al-Jazari wrote his second criterion in the plural ("one of the Uthmanic codices") precisely because Uthman's own copies differed from one another in forty to sixty places.  Ibn Abi Dawud devotes a chapter to listing those differences.  This is not a Western discovery;  it is standard classical Muslim scholarship.

That said, the objection is mostly right, and the article should not pretend otherwise.  The overwhelming majority of differences among the readings are in vowels and dots, and van Putten's count of 284 dotting disagreements across all the canonical readings is, as he says, "strikingly low."

"The lost verses were abrogated by God, so nothing was lost"

This is the tradition's own answer and it is theologically coherent.  It cannot be refuted by evidence, because it makes a claim about intent rather than about events.  But it settles the historical question in the article's favour:  it concedes that passages once recited as Quran are no longer in the Quran.  What it denies is that this was an accident.  A Muslim is entitled to that position.  A Muslim is not entitled to that position and the slogan that not one letter has changed.

"Uthman only standardised the recitation; he did not change the text"

This is close to right and it is the best version of the Muslim case at this point.  The reports do frame the crisis as one of recitation.  Uthman's own reported reply to the Egyptians is that he acted because he "feared division."  But two things in the same reports resist the softest reading.  The committee was instructed to resolve disagreements in favour of one dialect — that is a decision about wording, not about pronunciation.  And Bukhari 4988 says a verse was missing from the sheets and had to be found.  Standardising recitation does not require recovering a lost verse.

"The Sanaa palimpsest proves the Quran was rewritten"

It does not, and Christians should stop saying it does.  It has the same verses in the same order inside the surahs, with the two exceptions Sadeghi names — Quran 20:31 and 20:32 reversed, and Quran 9:85 absent.  Its doctrinal content is the same.  Its differences are overwhelmingly particles, pronouns and synonyms.  Its one whole missing verse has a mechanical explanation.  Sadeghi's own conclusion, in a secular journal, is that the standard text is the better witness to what Muhammad recited.  What Sanaa 1 does prove is narrower and still significant:  a real non-Uthmanic text type existed, in a real and costly codex, and it was erased.

"Radiocarbon proves the Quran is from Muhammad's lifetime"

Radiocarbon dates the death of the animal whose skin became the parchment.  It does not date ink.  Parchment could be stored, and in one case a leaf gave four incompatible answers from four laboratories.  Where palaeographers have examined the writing on carbon-dated leaves — Birmingham, Tübingen — they have consistently concluded that the writing is later than the skin.

"The Quran has one text; the Bible has thousands of variant manuscripts"

Both halves are true and the inference is backwards.  The Quran has one printed text mainly because of a twentieth-century Egyptian editorial decision resting on a seventh-century destruction order.  That is a fact about policy.  Many surviving independent manuscripts are what make it possible to detect and correct a change;  destroying the non-conforming copies removes that check.  If a word had been altered in Kufa in 660, no surviving evidence could reveal it.  The Christian tradition pays for its verifiability with about forty doubtful verses that every modern Bible flags.  That is the trade, and it should be stated as a trade.

"There are 2,270 variants in the Topkapi manuscript"

This figure circulates widely.  It comes from Tayyar Altikulaç's 2007 facsimile study, published by an Islamic institution, and the differences are reported to be orthographic — spelling, not words.  The figure could not be verified in Altikulaç's own text for this article, so it is not used here.  Anyone quoting it should read the source first.

"Brubaker documented thousands of textual changes"

He documented corrections — erasures, insertions and overwriting.  On Sidky's analysis, the vast majority of his published examples are ordinary scribal slips being fixed, several of them producing ungrammatical Arabic if left uncorrected.  The fact of corrections is real and worth knowing.  The inference that the text was still fluid centuries after standardisation is not established.

What evidence would change the conclusion?

What would strengthen the Muslim claim.

• A dated documentary Quran, or a securely dated Quranic inscription of any length, from before 650. The parchment dates cannot supply this; a dated colophon or a dated building inscription could.
• Radiocarbon work that resolved the laboratory disagreements — the same folio, blind, at several laboratories, with agreement — combined with palaeographic dating that agreed with it. At present the two methods contradict each other systematically.
• A demonstration that the canonical readings preserve information the written skeleton cannot supply, in places where the skeleton is unambiguous. That would answer van Putten's argument that the readers were largely reading off the page, and it is the load-bearing beam of the Muslim account of how the oral tradition worked.
• Publication of the Berlin Documenta Coranica reconstruction of the Sanaa lower text, if it showed the variants to be fewer or more mechanical than Sadeghi and Goudarzi reported.

What would weaken it.

• A second non-Uthmanic manuscript, from a different place, agreeing with Sanaa 1 against the standard text. One outlier can be an anomaly. Two would be a text type.
• A Quranic manuscript containing any of the passages the tradition reports as once-recited-and-now-absent — the stoning verse above all. None has ever been found, which is itself a datum in the Muslim side's favour.
• Evidence that the archetype belongs to 'Abd al-Malik's reign rather than Uthman's. Shoemaker argues this and the palaeographic dates lean his way. If it were established, the documented history of the Quran would begin around 700 rather than 650, and the gap between Muhammad and the first documented text would double.
• A published, verified whole-Quran collation of Hafs against Warsh — or better, of all twenty transmissions — showing meaning-affecting differences at a rate far higher than Brockett's sample suggests.

What would not change anything.  More corrections in early manuscripts;  more radiocarbon dates on parchment;  more lists of companion readings out of medieval commentaries.  Those three lines of evidence have been worked hard and have yielded what they are going to yield.

Sources

Every hadith below was fetched as bytes from sunnah.com and the tags stripped locally, in the research dossiers this article is built from.  Every Arabic primary work was fetched as bytes from shamela.ws or api.alquran.cloud.  Every book and article was downloaded as a PDF and read with a text extractor.  Nothing here was established through a web-summarising tool.

Quran

Hadith.  Sahih al-Bukhari (d. 256/870) and Sahih Muslim (d. 261/875) carry no grade line on sunnah.com, because inclusion in either collection is itself the claim being made for the report.  Gradings printed for Jami' at-Tirmidhi and Sunan Ibn Majah are Darussalam's, reproducing Muhammad Nasiruddin al-Albani's twentieth-century judgements, not medieval ones. Arabic primary works.  Page numbers were read from the [ص: N] printed-page markers in the Shamela text, not inferred from an offset. Books and articles.