The Seven Ahruf of the Qur'an: Anatomy of a Puzzle

Summary. — A hadith reported by al-Bukhârî and Muslim states that the Qur’an “was sent down according to seven aḥruf.” Twelve centuries of exegesis have failed to establish what that word denotes: Ibn Ḥibbân already counted thirty-five opinions, all of which he judged plausible. This article lays out the file — the body of hadiths, the major interpretations and their dead ends, the actual typology of transmitted variants, the material history of the text in the light of the manuscripts, and the theological knot. It will emerge that the most widespread error today, the identification of the seven aḥruf with Ibn Mujâhid’s seven canonical readings, was denounced as early as the fifth/eleventh century by the specialists themselves, and that the tradition produced on this question a critique more radical than that of many orientalists.

The textual file

The most frequently cited account features ʿUmar b. al-Khaṭṭâb hearing Hishâm b. Ḥakîm b. Ḥizâm recite Sûrat al-Furqân in a manner that was not his own. ʿUmar seizes him by the cloak, brings him before the Prophet, and is told that both recite correctly, for “this Qur’an was sent down according to seven aḥruf: recite whichever of them is easy for you”¹. Its authenticity is not in question; its object is.

One point deserves emphasis from the outset: the hadith defines nothing, it authorises. Its narrative function is to defuse a conflict between two reciters, not to set out a doctrine of variation. The whole difficulty stems from this — a text has been asked to settle a question it does not raise.

A second family of narratives, reported notably from Ubayy b. Kaʿb, tells of a negotiation. Jibrîl comes to the Prophet and proposes that he recite the Qur’an according to a single ḥarf; Mîkâʾîl, seated on his left, urges him to ask for more; and so they ascend, by degrees, to seven — “all of them healing and sufficient” (kulluhâ shâfin kâfin)². One variant shifts the emphasis towards mercy: the Prophet invokes God’s forgiveness for his community, “for my community is not capable of this.” Another specifies the motive — “I have been sent to an unlettered community: among them are the young boy, the servant, the decrepit old man, the old woman” — to which Jibrîl replies: “then let them recite the Qur’an according to seven aḥruf”³.

This is a doctrinal point too rarely noted: on its own logic, the hadith grounds plurality not in semantic richness but in human weakness. The seven aḥruf are there a rukhṣa, a dispensation granted to unequal reciters.

The corpus itself does not agree on the number. A report from Samura b. Jundub speaks of three aḥruf; the dominant version gives seven, transmitted from Ubayy b. Kaʿb, ʿUmar, Abû Hurayra, Muʿâdh b. Jabal, Samura and Ibn ʿAbbâs; a report attributed to ʿAlî b. Abî Ṭâlib counts ten⁴. The last is revealing, for it is no longer speaking of the same thing at all: “the Qur’an was sent down according to ten aḥruf: bearer of good tidings and warner, abrogating and abrogated, exhortation and parable, univocal and ambiguous, lawful and unlawful.” These are not manners of reciting but categories of content. The same word, ḥarf, here designates something else entirely — which alone would suffice to show that the term was already, for the transmitters, an empty slot.

Several versions finally add a limit: one may recite as one wishes, “so long as one does not end a verse of punishment with a mention of mercy, nor a verse of mercy with a mention of punishment”⁵. A variant from Ubayy even specifies the nature of the permitted substitutions: samîʿan ʿalîman, ʿazîzan ḥakîman — that is, the verse-final clausulae formed from divine names. A version from Abû Bakra adds, by way of example: “as when you say taʿâl, aqbil, halumma, idhhab, asriʿ, aʿjil.”

This clause is perhaps the most concrete indication in the corpus: it suggests that the variation bore upon form — interchangeable clausulae, synonyms — and not upon substance, whose integrity is precisely what the clause protects.

What is a ḥarf? An inventory of answers

One must begin with an admission the tradition makes itself: the meaning of the word ḥarf in this hadith has never been established. Ibn Ḥibbân (d. 354/965) counted thirty-five opinions of scholars and philologists, adding that “these opinions resemble one another and are all plausible” — a tally al-Suyûṭî takes up in al-Itqân⁶.

Such profusion is not a sign of doctrinal richness: it is the symptom of a lost datum. From the third/ninth century onwards, scholars reason about a term whose living usage escapes them. And Ibn Ḥibbân’s judgement that all thirty-five opinions are “plausible” is an admission: no criterion allowed one to decide between them.

The seven dialects

The most widespread interpretation makes the seven aḥruf seven dialects (lughât) of Arab tribes. A report attributed to Ibn ʿAbbâs holds that five of the seven fell to the ʿajz Hawâzin — Thaqîf, Banû Saʿd b. Bakr, Banû Jusham and Banû Naṣr b. Muʿâwiya. Ibn ʿAbd al-Barr judges this report unestablished in its transmission⁷.

The thesis runs up against a massive internal objection, which the classical scholars saw: the quarrel between ʿUmar and Hishâm sets two Qurashîs against each other. If the seven aḥruf were seven tribal dialects, two men of the same tribe would have had no reason to diverge. Ibn al-Jazarî indeed explicitly rejects the identification of the aḥruf with seven lughât⁸. Moreover, the lists of tribes proposed vary from author to author — Quraysh, Hudhayl, Thaqîf, Hawâzin, Kinâna, Tamîm, Yaman, in unstable combinations — which betrays an a posteriori reconstruction rather than a memory.

The licence of synonymy

A second line understands the aḥruf as the licence to employ a synonym. This is the ninth opinion recorded in al-Itqân: “seven ways of saying, with concordant meanings, through different expressions — such as aqbil, taʿâl, halumma, ʿajjil, asriʿ.” Al-Qurṭubî makes it the principal position in the introduction to his commentary, and ranges al-Ṭabarî and al-Ṭaḥâwî under it; the latter grounds the licence in the difficulty dialectal groups had in moving from one form of speech to another, divergence in wording being permitted “so long as the meaning agrees”⁹.

The file contains a celebrated case, reported by Ibn Jinnî (d. 392/1002) concerning Qur’an 9:57. Al-Aʿmash relates that he heard Anas b. Mâlik recite yajmizûn where the received text has yajmaḥûn. He is met with the objection: “what is yajmizûn? It is yajmaḥûn!” Anas replies: “yajmaḥûn, yajmizûn and yashtaddûn are one and the same thing”¹⁰.

Two clarifications are needed, which the secondary literature often glosses over. First, the objection is phrased in the impersonal passive (qîla lahu, “it was said to him”): nothing warrants attributing it to al-Aʿmash, who is merely the reporter of the scene. Second — and this matters more — Ibn Jinnî does not endorse the reading ascribed to him. He writes that the apparent sense (ẓâhir) of the anecdote is that the ancients “read one word in place of its equivalent without any prior reading having established it, purely on the strength of agreement in meaning”; he immediately acknowledges that “this is a place where an opponent would find grounds for attack”; and he then neutralises the difficulty by positing that the three words were in fact already transmitted readings, and that the hadith of the seven aḥruf covers them all. His invocation of the aḥruf is therefore exculpatory, not demonstrative.

An internal tension nevertheless remains. Commenting on a second anecdote from Anas — on Qur’an 73:6, aqwamu qîlan / aṣwabu qîlan — Ibn Jinnî goes further: “this suggests that these people attended to the meanings and held to them; once they had grasped and secured them, they allowed themselves some latitude in the expressions that convey them.” And he goes on to adduce non-Qur’anic poetic parallels, placing the phenomenon within a general linguistic series rather than within the doctrine of the aḥruf¹¹. The two passages do not say the same thing; they must be cited together.

Another case is better attested still: Ibn Masʿûd reciting fa-mḍû ilâ dhikri Llâh instead of fa-sʿaw ilâ dhikri Llâh (62:9), and justifying his reading by reasoning — “if it were fa-sʿaw, I would run until my cloak fell off.” Ibn ʿUmar further testifies that ʿUmar recited fa-mḍû until his death¹².

If this reading of the aḥruf is the right one, a consequence follows, and it is a heavy one: the seven aḥruf would not be seven revealed texts but a latitude granted to the reciter. What ʿUthmân’s recension suppressed would then be not six revelations but a freedom. This is, moreover, the position of al-Bâqillânî and al-Ṭaḥâwî as al-Suyûṭî reports them: the licence of substitution was an early rukhṣa, subsequently abrogated.

The registers of meaning

A third family identifies the seven aḥruf with seven categories of content: “command and prohibition, exhortation and warning, disputation, narratives and parables” (Abû Qilâba, mursal, in al-Ṭabarî); or, in Ibn Masʿûd’s version, “prohibition and command, lawful and unlawful, univocal and ambiguous, parables.”

These categories cannot be the seven aḥruf, and the corpus shows this by implication. The hadith of Abû Bakra juxtaposes in a single sentence the registers not to be confused (punishment / mercy) and the permitted synonyms (taʿâl, aqbil, halumma). In other words, the categories of meaning are precisely what must not vary when the form varies. To take them for the seven aḥruf is to confuse the rule with its object.

The outer and the inner

There remains a line apart, which pulls the hadith towards interiority: “the Qur’an was sent down according to seven aḥruf; every verse has an outer and an inner” (li-kulli âyatin minhâ ẓahrun wa-baṭn), with this continuation in certain versions: “every ḥarf has a limit, and every limit a point of ascent” (wa-li-kulli ḥaddin maṭlaʿ).

The critical file on this hadith is instructive in itself. Ibn Ḥibbân retains it in his Ṣaḥîḥ — but with âya (“every verse“) and not ḥarf. Shuʿayb al-Arnaʾûṭ judges it “of ḥasan chain, if Abû Isḥâq is indeed al-Hamdânî as the author indicates; and weak if it is Ibrâhîm b. Muslim al-Hajarî” — a conditional judgement too often cited without its condition, and one the same al-Arnaʾûṭ contradicts elsewhere by holding the hadith to rest on two weak chains. Al-Suyûṭî declares it ṣaḥîḥ, al-Albânî ḍaʿîf; Ibn Ḥazm declares it mursal, “which affords no proof”¹³.

What is striking in the afterlife of this dictum is its progressive detachment. Al-Ghazâlî, in the book on the etiquette of Qur’an recitation in the Iḥyâʾ, cites “the Qur’an has an outer, an inner, a limit and a point of ascent” — but without the matrix of the seven aḥruf, which he does not mention¹⁴. The esoteric motif thus separates from its qirâʾât frame, and it is in this detached form that the Sufi tradition inherits it. Is this a matter of form and substance, or of exoteric and esoteric? The two readings do not bear on the same object — and the history of transmission has in fact separated them.

A typology of attested variations

Rather than speculate on the meaning of ḥarf, one may inventory what actually varies in the transmitted readings. The following classification, in eight types, moves from the most innocuous to the weightiest.

Different vocalisation, single meaning. sadd / sudd; udhun / udhn. Purely phonetic variation.
Different vocalisation, different meaning. At 85:21-22, depending on the final vowel, maḥfûẓ qualifies the Tablet (“a preserved Tablet”) or the Qur’an itself (“a glorious Qur’an… preserved”). The difference is syntactic, and theologically notable.
Different diacritical points, identical rasm. taʿqilûn / yaʿqilûn; at 6:57, yaquṣṣu l-ḥaqq (“He sets forth the truth”) / yaqḍî l-ḥaqq (“He decides according to the truth”); at 10:22, yusayyirukum / yanshurukum (“He causes you to travel” / “He disperses you”).
Different rasm, identical meaning. qâla / qul, long â not always being written with an alif in early script; at 83:31, fakihîn / fâkihîn; at 18:74, zakiyya / zâkiya.
Different rasm and meaning. kânû hum ashadda minhum quwwatan / minkum — “mightier than them” or “mightier than you.”
Word order altered. At 3:195, fa-yaqtulûna wa-yuqtalûn / fa-yuqtalûna wa-yaqtulûn — “they kill and are killed” / “they are killed and kill.” At 50:19, wa-jâʾat sakratu l-mawti bi-l-ḥaqq / sakratu l-ḥaqqi bi-l-mawt — “the agony of death came with the truth” / “the agony of truth came with death”; al-Ṭabarî reports this second reading from Abû Bakr al-Ṣiddîq himself.
Change bearing on a single letter. At 2:132, waṣṣâ / awṣâ; at 36:35, ʿamilathu aydîhim / ʿamilat without the pronoun — “what their hands have made” or “what their hands have not made.”
Change bearing on a word. At 73:6, aqwamu qîlan / aṣwabu qîlan; at 101:5, ka-l-ʿihni l-manfûsh / ka-l-ṣûfi l-manfûsh, the reading of Ibn Masʿûd; at 18:79, yaʾkhudhu kulla safînatin ghaṣban with the addition of ṣâliḥatin (“every sound vessel”), a reading attributed to Ibn Masʿûd, Ubayy, Ibn ʿAbbâs and Saʿîd b. Jubayr. Here the variant is no longer a synonym: it is an addition that changes the import of the Khiḍr narrative, restricting the seizure to vessels still fit for service.

The third type calls for a remark: divergence of this kind indicates that some preserved the text without having memorised it. There could therefore exist a conforming, memorised version and a deciphered one. This is exactly the point on which the modern discussion will turn.

Two major cases deserve separate treatment. The dhakar and the unthâ (92:1-3): the canonical version has wa-mâ khalaqa l-dhakara wa-l-unthâ (“and by that which He created, the male and the female”), but Ibn Masʿûd — and, according to Abû l-Dardâʾ, the Prophet himself — recited wa-l-dhakari wa-l-unthâ, without mâ khalaqa. The episode is remarkable for its conclusion: Abû l-Dardâʾ explicitly refuses to follow the received reading — “by God, I shall not follow them”¹⁵.

The ṣalât al-wusṭâ (2:238): the muṣḥaf of ʿÂʾisha, like that of Ḥafṣa, had ḥâfiẓû ʿalâ l-ṣalawâti wa-l-ṣalâti l-wusṭâ wa-ṣalâti l-ʿaṣr — the gloss “and the afternoon prayer” incorporated into the text. The testimonies agree, and Ibn Ḥajar notes some twenty lines of transmission all carrying wa-ṣalâti l-ʿaṣr. Al-Barâʾ b. ʿÂzib indeed presents the matter as an abrogation: it was recited thus, then God abrogated it¹⁶.

This last case is doubly instructive. It furnishes a well-attested variant; and it shows that the tradition itself hesitates over its classification — ḥarf? interpolated explanatory gloss? verse abrogated in its recitation? The three categories overlap, and it is here that the file of the seven aḥruf communicates with that of naskh.

The material history of the text

The Arabic script of the earliest decades notes the consonants imperfectly and the short vowels not at all. The count is precise: of twenty-eight letters, twenty-two fall into nine classes of homographs, only six having a distinct shape. The consonantal skeleton therefore comprises only fifteen distinct forms in initial and medial position, eighteen in isolated and final position.

The chronology of pointing: a necessary correction

One claim circulates widely, including in francophone reference notices: the first appearance of diacritical marks is said to be “an inscription of 677.” It must be corrected on two counts.

It is identifiable: this is the inscription of Muʿâwiya’s dam near Ṭâʾif (Sadd Saysad), dated to the year 58 AH / 677-678¹⁷. But it is no longer the earliest. Three attestations precede it: the Nabataeo-Arabic inscription of Raqush at Madâʾin Ṣâliḥ (267 CE); the bilingual papyrus PERF 558 from Egypt (22 AH / 643); and the inscription of Zuhayr in north-western Arabia (24 AH / 644-645), which according to its editors displays a fully constituted system of diacritical marks¹⁸. As for Raqush, it establishes that pointing is not an Islamic invention.

The claim about 677 is therefore not an error: it is an outdated datum, and its obsolescence can be dated — it holds for the state of the epigraphic file before the publication of the Zuhayr inscription in 2008.

The other landmarks stand as follows. Short vowels are said to have first been marked by coloured dots, a system attributed to Abû l-Aswad al-Duʾalî (d. 69/688). The systematic consonantal pointing of the muṣḥaf is attributed to Naṣr b. ʿÂṣim al-Laythî (d. 89/707-708) — nicknamed Naṣr al-ḥurûf — and to Yaḥyâ b. Yaʿmar, under al-Ḥajjâj. The present system is attributed to al-Khalîl b. Aḥmad al-Farâhîdî (d. 175/791); it spreads progressively from the end of the ninth century¹⁹.

These traditional attributions are today contested. The earliest ḥijâzî manuscripts already carry diacritics, sporadic but real, and non-Qur’anic Arabic documents of the same period display the same sparing usage. Pointing is therefore not a datable invention ascribable to one man: it is a continuous scribal practice, of which the accounts concerning Abû l-Aswad and al-Ḥajjâj constitute a retrospective aetiology²⁰.

The philological thesis and its counterpoint

François Déroche defends a position that reverses the traditional perspective: many variants would go back not to an original plural recitation but to the deciphering of ambiguous manuscripts. The plurality of readings would in part be an effect of writing, not a datum of revelation²¹. The thesis has a genealogy: Ignaz Goldziher had formulated it as early as 1920, making the ambiguity of the rasm the primary source of the variants²².

Research of the past twenty years has considerably complicated this picture, in two opposing directions.

On one side, the single written archetype. Marijn van Putten starts from a minute observation: the phrase niʿmat allâh may be written with open tâʾ or with tâʾ marbûṭa, and the spelling adopted varies from one occurrence to another. Yet across fourteen early manuscripts, the spelling is constant for each given occurrence. Such convergence of orthographic idiosyncrasies is explicable only by copying from a single written archetype. The corollary is clear: Qur’anic manuscripts descend, from the first century onwards, from a written exemplar — which massively supports the traditional account of an ʿUthmânic recension²³.

On the other, irreducible orality. Hythem Sidky catalogues 292 points of consonantal pointing disagreement among the ten canonical readers and submits them to principal component analysis. His result tells against Goldziher and qualifies Déroche: the structure of the disagreements reveals an inherited oral tradition going back to the ʿUthmânic recension, and not a series of independent decipherings of a mute text²⁴.

The two results converge on one point: the rasm is stable at an early date, and the qirâʾât are not the inventions of perplexed scribes. What varies lies elsewhere — in spelling detail, in the order of the sûras, in pointing and vocalisation.

What the manuscripts show

The Codex Parisino-petropolitanus comprises 98 folios from an original codex estimated at 210-220. It is the work of five copyists working in all likelihood in parallel — a division of labour that presupposes a written exemplar to copy from. Déroche dates it to the third quarter of the first/seventh century²⁵.

The Birmingham manuscript has yielded a radiocarbon dating of 568-645 CE at 95.4 per cent confidence, and belongs to the same codex as sixteen Paris folios. Van Putten has shown that its orthographic idiosyncrasies make it clearly a descendant of the ʿUthmânic text type, which rules out a pre-ʿUthmânic origin despite the early date of the parchment. The Tübingen manuscript gives 649-675 at the same confidence.

The Codex Mashhad is the most eloquent: 251 folios, more than 90 per cent of the text, ʿUthmânic rasm but Ibn Masʿûd’s order of sûras — an order subsequently rearranged by cutting and re-pasting so as to conform to the canonical sequence²⁶. The dissociation between stability of the rasm and freedom of arrangement is observable there on one and the same object.

The Ṣanʿâʾ palimpsest occupies a place apart: its lower layer is, according to its editors, the sole surviving witness to a non-ʿUthmânic textual tradition. Its variants partly overlap with those attributed to Ibn Masʿûd and Ubayy, without being identifiable with either: the order of sûras there places al-Tawba after al-Anfâl, the reverse of the order ascribed to Ibn Masʿûd²⁷. Asma Hilali has seen in it not a muṣḥaf but a scribal exercise; the thesis has been widely rejected, notably by Éléonore Cellard’s codicological reconstruction, which concludes to a codex of professional manufacture²⁸.

A methodological remark is in order here: radiocarbon dating dates the death of the animal from which the parchment comes, not the writing. The Ṣanʿâʾ file furnishes the experimental demonstration, certain folios yielding intervals earlier than the preaching itself.

It is from this witness that Déroche formulates his strongest hypothesis: the sûras would already have been largely constituted before the Prophet’s death, the vulgate and the lower layer of Ṣanʿâʾ depending on collections in which the prophetic teaching had been set down; whereas the sequence of the sûras would not have been stabilised. The Codex Mashhad points the same way.

The “second maṣâḥif project”

Omar Hamdan argues that an undertaking conducted at Wâsiṭ in 84-85 AH / 703-704, under al-Ḥajjâj b. Yûsuf and with caliphal patronage, played a decisive part in consolidating the text: diacritical pointing, counting of letters and verses, divisions of the text, authoritative distribution of exemplars, public reading from the codex in the mosques²⁹.

Three clarifications are needed, for this file is frequently distorted. First, Hamdan does not argue that al-Ḥajjâj altered the text: he explicitly refutes that reading, notes that the principal aim was the political image of the Umayyads, and stresses that al-Ḥajjâj failed to standardise the oral tradition. Second, the thesis has been seriously criticised: Nicolai Sinai charges it with presupposing a unified editorial project on the model of a late narrative, without considering that this narrative may itself be a retrospective ordering; Alain George notes in it a confusion between vocalisation and diacritics, and recalls that diacritics are attested at least half a century earlier³⁰.

Third — and this is the most interesting point — a partial agreement emerges on precise ground. Van Putten, while setting aside the idea of a recasting of the rasm under al-Ḥajjâj, observes that the alif is precisely the letter for which fidelity to the idiosyncrasies of a single archetype is not observed, and suggests that a process of this kind may indeed have taken place; Sinai, for his part, defends the traditional dating of the rasm “excepting certain orthographical features.” The three positions thus converge on a reform limited to orthography, the consonantal skeleton remaining ʿUthmânic.

One final methodological point. At 24:10, one manuscript carries a markedly divergent reading in which one might be tempted to see a lost ḥarf. But verse 24:21 carries precisely the formula in question: the hypothesis of contamination by homoeoteleuton — the most banal of copyist’s errors — is here the most economical. Not every variant is a ḥarf, and a critical history of the text must be able to distinguish the transmitted variant from the material accident.

From aḥruf to qirâʾât: a confusion and its denunciation

Al-Ṭabarî’s position is unambiguous: Muslims today recite only according to the single ḥarf chosen for them by their imam — ʿUthmân — to the exclusion of the other six³¹. It is the most coherent thesis, and the most costly. Coherent, because it explains the ʿUthmânic unification without detour. Costly, because it concedes that six-sevenths of what was revealed is no longer recited. Ibn al-Jazarî would defend the contrary position: the seven aḥruf are contained within what the ʿUthmânic rasm carries.

In 324/936, Abû Bakr Ibn Mujâhid published his Kitâb al-Sabʿa fî l-qirâʾât and retained seven readers: Nâfiʿ (Medina, d. 169/785), Ibn Kathîr (Mecca, d. 120/738), Abû ʿAmr b. al-ʿAlâʾ (Basra, d. 154/770), Ibn ʿÂmir (Damascus, d. 118/736), ʿÂṣim (Kûfa, d. 127/745), Ḥamza (Kûfa, d. 156/773) and al-Kisâʾî (Kûfa, d. 189/804). The selection is massively reductive: at least some fifty readings were then in circulation³².

And it is backed by power: the two trials (miḥna) of Ibn Miqsam (322/934) and Ibn Shannabûdh (323/935), under the vizier Ibn Muqla, chronologically frame the undertaking. They deserve to be read together, for they condemn symmetrical and opposite excesses. Ibn Miqsam held that any reading linguistically valid and compatible with the rasm is licit — that is, the rasm as an open matrix, without isnâd. Ibn Shannabûdh recited publicly, in prayer, variants of Ibn Masʿûd and Ubayy that diverged from the rasm — that is, isnâd against the written text; he was stripped and flogged, having pleaded to no effect that each of his readings rested on impeccable chains. Canonisation therefore requires the conjunction of rasm and isnâd: neither suffices alone.

The medieval denunciation of the “seven equals seven” equation

Here the file holds its greatest surprise: the identification of the seven readings with the seven aḥruf was denounced by the specialists themselves, and in the sharpest terms.

Al-Mahdawî (d. 430/1039) coined a neologism that has remained famous: “he who sevened these seven (musabbiʿ hâʾulâʾi l-sabʿa) did what he ought not to have done; he confused the common people to the point of making them ignorant of what it was not permitted for them to be ignorant of, and he made every shallow mind believe that these are the ones the prophetic hadith mentions.” Makkî b. Abî Ṭâlib (d. 437/1045): “whoever imagines that the reading of each of these readers is one of the seven aḥruf stipulated by the Prophet commits a blunder.” Abû Shâma (d. 665/1267) judges the notion “contrary to the consensus of the people of knowledge.” Ibn Taymiyya: “there is no disagreement among scholars worthy of consideration that the seven aḥruf mentioned by the Prophet are not the readings of the seven celebrated readers”³³.

Ibn al-Jazarî (d. 833/1429), finally, devotes an entire section to the question and does not mince his words. He states that he has expanded the chapter “because word has reached us that certain people, devoid of knowledge, claim that the sound readings are those of these seven, or that the seven aḥruf to which the Prophet alluded are the reading of these seven.” And he adds this diagnosis:

Worse still: it has prevailed among many of the ignorant that the sound readings are those found in the Shâṭibiyya and the Taysîr, and that these are what the Prophet’s words designate: “the Qur’an was sent down according to seven aḥruf”³⁴.

In other words: the canon came to be identified with its teaching manuals. This is a phenomenon of reification that will recur, six centuries later, with the Cairo edition.

The verdict of modern scholarship

Christopher Melchert joins the tradition on this point, and goes further: the identification of the seven aḥruf with the seven readings appears to him contrary to reason and unsupported by the Islamic tradition. He further disputes that Ibn Miqsam’s trial had anything to do with the Seven, the accounts bearing throughout on the interpretation of the consonantal outline and departure from received tradition. He notes, finally, that Ibn Mujâhid was himself not scrupulous about chains, the reading of Nâfiʿ being described as his personal synthesis of five Medinan readings³⁵.

One question remains open, and must be left so: it is not known whether Ibn Mujâhid himself ever invoked the hadith of the seven aḥruf to justify his selection³⁶. Déroche writes, cautiously, that one cannot rule out that the choice aimed at conflating the two realities, and recalls that medieval scholars reproached him for it. Al-Mahdawî’s regret — he wished Ibn Mujâhid had chosen some other number — suggests an unpremeditated effect rather than a design.

What is certain, by contrast, is that the equivalence is arithmetically untenable: the canonical number is not fixed. Ibn al-Jazarî raised it to ten by adding Abû Jaʿfar (Medina, d. 130/747), Yaʿqûb al-Ḥaḍramî (Basra, d. 205/820) and Khalaf al-ʿÂshir (Baghdad, d. 229/843). And al-Dimyâṭî (d. 1117/1705) counts fourteen³⁷.

What the system owes to its codifiers

Three historical clarifications guard against common anachronisms.

The “seven readers × two transmitters” system is not Ibn Mujâhid’s. He documents, on the contrary, multiple transmitters per reader. The reduction to two râwî is the work of the Andalusian tradition: it takes shape with Ibn Ghalbûn (d. 389/999), is systematised by al-Dânî (d. 444/1053) in al-Taysîr, then crystallised by al-Shâṭibî (d. 590/1194). That is some two hundred and fifty years after the selection of the seven, and applied retroactively³⁸.

The three circles are not concentric. The Seven date from 324/936, the Ten from the first half of the ninth/fifteenth century, the Fourteen from the twelfth/eighteenth. These are three distinct gestures, separated by centuries and performed in three unrelated political settings: Abbasid Baghdad, the late Mamluk world, Ottoman Egypt.

The boundary of the canon separates transmissions, not men. Al-Dûrî is a transmitter of two of the Seven — directly from al-Kisâʾî, indirectly from Abû ʿAmr — and his intermediary to Abû ʿAmr, al-Yazîdî, is himself one of the four readers classed as shâdhdha. Khalaf is at once a transmitter of Ḥamza and the tenth reader. The canon cuts across chains, not persons.

The competing recensions

Déroche sets the indeterminacy of the word ḥarf against a perfectly clear historical situation: that in which ʿUthmân’s opponents reproached him with having reduced to one a Qur’an that was multiple. The hadith of the seven aḥruf asserted, in all its versions, that variation formed an integral part of transmission; but what variation? The medieval scholars did not elucidate it.

Two facts mark the persistence of competing recensions well after the recension. Mâlik b. Anas refused the validity of prayer performed with Ibn Masʿûd’s text, and appealed to the Abbasid authorities to block its circulation — which presupposes a circulation to block. Saʿîd b. Jubayr recited during Ramadan one day Ibn Masʿûd’s version, the next day ʿUthmân’s. This second testimony is the more telling: it implies that the two versions coexisted without either cancelling the other, in the devotional practice of a leading Successor.

Déroche suggests that orality may have played a greater part in this version, whose transmitters perhaps accepted recourse to synonyms — which would account for the differences observed among its transmissions. In the tenth century, al-Nadîm could still declare that he had seen several maṣâḥif presented as Ibn Masʿûd’s version, and differing among themselves.

Bergsträsser and Pretzl held that the expression there was clearer than in the ʿUthmânic recension, and the vocabulary more current. The argument deserves to be turned around: in textual criticism, the principle of the lectio difficilior would have it that a clearer text is later, difficulty tending to be smoothed away in the course of copying. Their judgement, which appears to favour Ibn Masʿûd’s codex, would therefore tell rather, on this precise point, against its priority.

One material trace deserves mention. Ibn Masʿûd’s reading at 2:137 — fa-in âmanû bi-mâ âmantum bihi instead of bi-mithli mâ âmantum bihi — is attested both by the exegetes and by at least one early manuscript. Al-Mâturîdî reports from Ibn ʿAbbâs a theological justification: not to read bi-mithli, “for God has no like”; al-Qurṭubî notes the same reading in Ibn Masʿûd’s ḥarf. Déroche draws a chronological observation from this: at the beginning of the eighth century it was still possible to incorporate a reading from another version into an otherwise canonical exemplar, even as the authorities were striving to prevent the production of divergent copies.

Déroche draws a conclusion that may serve as a synthesis: the Muslim tradition itself knows of various traces of variation in the wording of the text. Some are held, at least in part, to form part of the revelation. Others were excluded over time — thus the possibility of employing synonyms — but their memory was preserved in a specialised literature.

This is perhaps the most accurate formulation of the puzzle: the seven aḥruf are not a lost secret; they are a preserved memory. The literature of the qirâʾât, of the maṣâḥif and of naskh is precisely where the tradition filed away what it had ceased to practise.

The theological knot

The files of the seven aḥruf and of abrogation communicate, and the boundary between them is porous. The ṣalât al-wusṭâ is presented now as a variant reading, now as a verse abrogated in its recitation. The hadith of Abû Mûsâ al-Ashʿarî — who declares that he recited a sûra “which we used to compare in length and severity to Barâʾa,” of which he now recalls only one verse — belongs to the same complex³⁹. The two notions perform the same operation — accounting for material that was Qur’anic and no longer is — but they impute it to different agents: a caliph’s decision in one case, a divine act in the other. The choice between the two classifications is not philological: it is theological.

Verse 15:9 — innâ naḥnu nazzalnâ l-dhikra wa-innâ lahu la-ḥâfiẓûn — promises the preservation of the Reminder. How is this to be squared with the disappearance of six aḥruf? Two classical answers. The first restricts the scope of the promise: it bears on the ḥarf retained, alone destined to endure. The second integrates the suppression into the divine act itself, on the basis of 87:6-7: “We shall make you recite, and you will not forget, save what God wills.” The exception is read as bearing precisely on what God wills to be forgotten. What disappears therefore does not disappear against preservation: it disappears through it.

The paradox of tawâtur

The most remarkable point in the theological file has gone almost unnoticed outside specialist circles. At the head of al-Nashr, Ibn al-Jazarî sets out the three conditions for a valid reading:

Every reading that accords with Arabic, even in but one respect, that accords with one of the ʿUthmânic codices, even approximately, and whose chain of transmission is sound — that is the ṣaḥîḥ reading, which it is not permitted to reject nor licit to repudiate; it belongs rather to the seven aḥruf according to which the Qur’an was sent down. […] And whenever one of these three pillars is wanting, it is termed weak, anomalous or void — whether it comes from the Seven or from those greater than they⁴⁰.

The third condition is the soundness of the chain (ṣiḥḥat al-sanad), not tawâtur — transmission by a number of reporters precluding collusion in falsehood. Ibn al-Jazarî explains himself, and the admission is direct: “I formerly inclined to that opinion, then its unsoundness became apparent to me.”

His argument is reductionist, not apologetic, and that is what gives it its value. First: if tawâtur were established, the other two conditions would become superfluous, since a mutawâtir variant contradicting the rasm would have to be accepted — which wrecks the three-condition system. Second: were tawâtur required for every variant, “a great part of the established variants of these seven imams themselves would fall away.” The requirement of tawâtur would destroy the very canon it claims to found.

He was not the first to see it. Makkî b. Abî Ṭâlib had already written that the criterion is the conjunction of the three elements, “even if seventy thousand reporters had transmitted it” — an explicit rejection of counting as a criterion. And Abû Shâma conceded that tawâtur is not realised for every variant, proposing a fallback criterion of notoriety.

The paradox is this: the man who abandons tawâtur as a condition is the man from whom tawâtur becomes dogma. A century later al-Suyûṭî incorporates the Jazarian criterion but subordinates it: he makes it the definition of the category mashhûr, ranged beneath mutawâtir in a six-tier typology. What was for Ibn al-Jazarî the sovereign criterion becomes for him a second-rank category. This is a neutralisation, not a reception. Shady Nasser formulates the outcome in modern terms: the transmission of the readings does not satisfy the conditions of tawâtur as the uṣûliyyûn define them, whence a lasting tension between the transmission of the written text and that of its oral recitation⁴¹.

The Imâmî position

The Twelver tradition offers a useful counterpoint, and a more nuanced one than is often said. Al-Kulaynî reports that al-Fuḍayl b. Yasâr asked Jaʿfar al-Ṣâdiq: “people say that the Qur’an was sent down according to seven aḥruf.” The reply: “They have lied, the enemies of God. It was sent down according to a single ḥarf, from the One“⁴². The play on oneness — a single ḥarf, come from the One — carries the whole argument. A neighbouring hadith, from Zurâra on the authority of al-Bâqir, does not mention the seven aḥruf but imputes divergence to transmission: “the Qur’an is one, sent down from the One; but divergence comes from the reporters.”

A piquant detail: al-Majlisî grades the polemical hadith ḥasan, and the conciliatory one ḍaʿîf. In both traditions, the grading readily follows the doctrinal stake.

The doctrinal position is old and stable: al-Ṭûsî writes that “what is known of the doctrine of our companions, and what is widespread in their traditions, is that the Qur’an was sent down according to a single ḥarf, upon a single prophet” — while noting at once that they agree on the licitness of reciting according to the readings in use. Al-Ṭabrisî and, in the contemporary period, al-Khûʾî take up the position, the latter objecting notably that the hadith “does not refer to any intelligible meaning”⁴³. But the Imâmî corpus is not monolithic: al-Ṣadûq devotes in al-Khiṣâl a chapter entitled “the Qur’an was sent down according to seven aḥruf,” in which he reports traditions that affirm them. The consensus of the ḥarf wâḥid was therefore built doctrinally, by marginalising these reports — exactly as the Sunnî tradition built its own by marginalising the reports of three or ten aḥruf.

A rukhṣa become dogma

The central shift in this whole history must be named. What the hadith presents as a dispensation granted to an unlettered community was transformed by the later tradition into a permanent property of the revealed text, and then into proof of its inexhaustible richness.

The movement extends into the contemporary period. The Cairo edition, published on 10 July 1924 / 7 Dhû l-Ḥijja 1342, fixes a text according to the single transmission of Ḥafṣ from ʿÂṣim. It proceeds not from early manuscripts but from the medieval normative treatises — al-Dânî for the rasm, Abû Dâwûd Sulaymân b. Najâḥ and al-Tanasî for the ḍabṭ — a method already applied by al-Mukhallalâtî in his edition of 1308/1890. Its claim to validity rests, as Gabriel Said Reynolds writes, “not on antiquity, but on canonicity”: the task was to choose a text, not to recover one⁴⁴.

Two details complete the picture. The edition’s appendix describes as “errors” the variants of the school copies it came to replace — and a great many of those copies were destroyed by sinking them in the Nile, a gesture Reynolds explicitly likens to those ascribed to ʿUthmân and al-Ḥajjâj. As for the committee’s president, Muḥammad b. ʿAlî al-Ḥusaynî al-Ḥaddâd, shaykh al-maqâriʾ of Egypt, he is the author of a refutation of Ṭâhâ Ḥusayn in which he maintains that the present reading of the Qur’an is exactly that of the Prophet and cannot be referred to the dialects of the Arab tribes. The theological thesis and the editorial act are here one and the same gesture.

One last point calls for correction, since it circulates everywhere: the figure of 95 per cent of Muslims reciting according to Ḥafṣ has no identifiable source. Its most widespread version, which apportions the riwâyât by percentage, totals 109 per cent. It appears to derive from a claim about countries rather than persons. The only firmly dated observation remains that of Otto Pretzl who, in 1938, expressed astonishment at finding “almost only the reading of Ḥafṣ from ʿÂṣim still known and used” — astonishment, precisely, because the thing was recent⁴⁵. Ḥafṣ had not established itself across the Arab and Muslim world before the Ottomans made it the official reading of the Empire.

The dominance of Ḥafṣ is thus a modern fact. And the specific contribution of 1924 is not “Ḥafṣ rather than Warsh” — Ḥafṣ was already the Egyptian norm — but this Ḥafṣ rather than those: the edition resolved an indeterminacy internal to the transmission, four ṭuruq being claimed for it. Nor have the other riwâyât disappeared: the King Fahd Complex at Medina today prints Warsh for the Maghreb, al-Dûrî for Sudan and Africa, Qâlûn for Libya. Plurality is not dead; it has become regional and residual, identified as an exception where Ḥafṣ has become the default text.

Conclusion

The hadith is soundly transmitted; its object is not. This dissociation is the central datum of the problem, and the thirty-five opinions catalogued by Ibn Ḥibbân are its symptom, not its richness.

The three major answers are not compatible with one another. The dialectal thesis founders on the fact that ʿUmar and Hishâm were both Qurashîs. The thesis of registers of meaning confuses the rule with its object, since those registers are precisely what the restrictive clause forbids altering. There remains the thesis of synonymic licence, the best supported by the facts of transmission — but it carries a cost its medieval proponents saw: what the recension suppressed is then not six revelations but a freedom.

Modern philology does not deny plurality: it redistributes its causes. One share falls to the ambiguity of the rasm, one to the early freedom of the reciters, one to copying accidents. But recent research corrects this framework in both directions: van Putten establishes a single written archetype from the first century, and Sidky shows, through analysis of the 292 pointing disagreements among the ten readers, that an inherited oral tradition underlies the qirâʾât. They are not the decipherings of perplexed scribes.

The coincidence between the seven aḥruf and the seven qirâʾât is an accident of institutional history. To dispel it is no critical audacity: it is to repeat what al-Mahdawî, Makkî, Abû Shâma, Ibn Taymiyya and Ibn al-Jazarî were already saying. The last of these even diagnosed the mechanism — the canon confused with its teaching manuals. And the canonical system owes more to its late codifiers than is commonly supposed: the two-transmitter scheme is Andalusian and two and a half centuries later; the Ten are of the fifteenth century, the Fourteen of the eighteenth; and the doctrine of tawâtur, today held to be self-evident, was explicitly rejected by the very man who fixed the canon of the Ten.

What remains, finally, is a preserved memory. The tradition filed away in a specialised literature — qirâʾât, maṣâḥif, nâsikh wa-mansûkh — what it had ceased to practise, instead of effacing it. It transmitted the reports that embarrassed it: Abû l-Dardâʾ refusing the received reading, ʿUmar reciting fa-mḍû until his death, Saʿîd b. Jubayr alternating the two recensions during Ramadan, Ibn al-Jazarî admitting he had been wrong. That is a form of documentary honesty which must be granted it, and of which this article will have been but one use.

Notes and references

Al-Bukhârî, al-Jâmiʿ al-ṣaḥîḥ, kitâb Faḍâʾil al-Qurʾân, no. 4992; Muslim, al-Ṣaḥîḥ, kitâb Ṣalât al-musâfirîn, no. 818. ↩
Al-Ṭaḥâwî, Sharḥ mushkil al-âthâr, nos. 3111-3112. ↩
Al-Ṭabarî, Jâmiʿ al-bayân ʿan taʾwîl âyi l-Qurʾân, muqaddima. ↩
On the versions with three and with ten aḥruf, see al-Albânî, Ḍaʿîf al-Jâmiʿ al-ṣaghîr, nos. 1336 and 1339, who holds both to be weak; al-Suyûṭî nonetheless retained the first as sound in al-Jâmiʿ al-ṣaghîr. ↩
Al-Ṭaḥâwî, Sharḥ mushkil al-âthâr, no. 3101 (from Abû Hurayra); al-Ṭabarî, Jâmiʿ al-bayân, muqaddima (from Abû Ṭalḥa). ↩
Al-Suyûṭî, al-Itqân fî ʿulûm al-Qurʾân, nawʿ 16, section “al-aqwâl fî l-murâd bi-l-aḥruf al-sabʿa.” ↩
Ibn ʿAbd al-Barr, al-Tamhîd. ↩
Ibn al-Jazarî, al-Nashr fî l-qirâʾât al-ʿashr, muqaddima, section on the nature of the divergence among the aḥruf. ↩
Al-Suyûṭî, al-Itqân, nawʿ 16, ninth opinion; al-Qurṭubî, al-Jâmiʿ li-aḥkâm al-Qurʾân, muqaddima, chapter on the seven aḥruf. ↩
Ibn Jinnî, al-Muḥtasab fî tabyîn wujûh shawâdhdh al-qirâʾât, i, on Qur’an 9:57. ↩
Ibn Jinnî, al-Muḥtasab, ii, on Qur’an 73:6. ↩
Ibn Ḥajar al-ʿAsqalânî, Fatḥ al-Bârî, commentary on Sûrat al-Jumuʿa. ↩
Ibn Ḥibbân, Ṣaḥîḥ (Ibn Balbân’s recension), no. 75, with Shuʿayb al-Arnaʾûṭ’s conditional judgement; cf. his Takhrîj Sharḥ al-sunna, where he holds it weak. Al-Suyûṭî, al-Jâmiʿ al-ṣaghîr, no. 2712 (ṣaḥîḥ); al-Albânî, Ḍaʿîf al-Jâmiʿ, no. 1338, and Silsilat al-aḥâdîth al-ḍaʿîfa, no. 2989 (ḍaʿîf); Ibn Ḥazm, al-Iḥkâm fî uṣûl al-aḥkâm. ↩
Al-Ghazâlî, Iḥyâʾ ʿulûm al-dîn, kitâb Âdâb tilâwat al-Qurʾân, fourth bâb. ↩
Al-Bukhârî, no. 4944; Muslim, no. 824. ↩
Muslim, no. 629; Abû Dâwûd; al-Ṭaḥâwî, Sharḥ maʿânî l-âthâr; Ibn Ḥajar, al-Kâfî al-shâf. ↩
George C. Miles, “Early Islamic Inscriptions near Ṭāʾif in the Ḥijāz,” Journal of Near Eastern Studies 7 (1948), pp. 236-242. ↩
ʿAlî b. Ibrâhîm Ghabbân and Robert Hoyland, “The inscription of Zuhayr, the oldest Islamic inscription (24 AH/AD 644-645), the rise of the Arabic script and the nature of the early Islamic state,” Arabian Archaeology and Epigraphy 19 (2008), pp. 209-237. ↩
François Déroche, Le Coran, Paris, PUF, “Que sais-je?”, 2017, ch. IV. ↩
Adam Bursi, “Connecting the Dots: Diacritics, Scribal Culture, and the Qurʾān in the First/Seventh Century,” Journal of the International Qurʾanic Studies Association 3 (2018), pp. 111-157. ↩
François Déroche, Le Coran, une histoire plurielle. Essai sur la formation du texte coranique, Paris, Seuil, 2019. ↩
Ignaz Goldziher, Die Richtungen der islamischen Koranauslegung, Leiden, Brill, 1920. ↩
Marijn van Putten, “‘The Grace of God’ as evidence for a written Uthmanic archetype: the importance of shared orthographic idiosyncrasies,” Bulletin of the School of Oriental and African Studies 82/2 (2019), pp. 271-288. ↩
Hythem Sidky, “Consonantal Dotting and the Oral Quran,” Journal of the American Oriental Society 143/4 (2023), pp. 785-814. ↩
François Déroche, La transmission écrite du Coran dans les débuts de l’islam: le codex Parisino-petropolitanus, Leiden, Brill, 2009. ↩
Morteza Karimi-Nia, “A New Document in the Early History of the Qurʾān: Codex Mashhad, an ʿUthmānic Text of the Qurʾān in Ibn Masʿūd’s Arrangement of Sūras,” Journal of Islamic Manuscripts 10/3 (2019). ↩
Behnam Sadeghi and Mohsen Goudarzi, “Ṣanʿāʾ 1 and the Origins of the Qurʾān,” Der Islam 87 (2012), pp. 1-129; cf. Behnam Sadeghi and Uwe Bergmann, “The Codex of a Companion of the Prophet and the Qurʾān of the Prophet,” Arabica 57/4 (2010), pp. 343-436. ↩
Asma Hilali, The Sanaa Palimpsest: The Transmission of the Qurʾan in the First Centuries AH, Oxford, Oxford University Press, 2017; for the codicological response, Éléonore Cellard, “Les manuscrits coraniques anciens,” in Le Coran des historiens, ed. M. A. Amir-Moezzi and G. Dye, Paris, Cerf, 2019, vol. 1. ↩
Omar Hamdan, “The Second Maṣāḥif Project: A Step Towards the Canonization of the Qurʾanic Text,” in A. Neuwirth, N. Sinai and M. Marx (eds.), The Qurʾān in Context, Leiden, Brill, 2010, pp. 795-835. ↩
Nicolai Sinai, “When did the consonantal skeleton of the Quran reach closure? Part I,” BSOAS 77/2 (2014), pp. 273-292; Alain George, “Coloured Dots and the Question of Regional Origins in Early Qurʾans (Part I),” Journal of Qurʾanic Studies 17/1 (2015), pp. 1-44. ↩
Al-Ṭabarî, Jâmiʿ al-bayân, muqaddima; cf. Ibn Kathîr, Faḍâʾil al-Qurʾân. ↩
Shady Hekmat Nasser, The Second Canonization of the Qurʾān (324/936): Ibn Mujāhid and the Founding of the Seven Readings, Leiden, Brill, 2020. ↩
These four judgements are gathered by Ibn al-Jazarî, al-Nashr, muqaddima, in the chapter where he refutes the identification. ↩
Ibn al-Jazarî, al-Nashr, muqaddima. The Shâṭibiyya is the didactic poem of al-Shâṭibî (d. 590/1194); the Taysîr is the manual of al-Dânî (d. 444/1053). ↩
Christopher Melchert, “Ibn Mujāhid and the Establishment of Seven Qurʾanic Readings,” Studia Islamica 91 (2000), pp. 5-22. ↩
Gabriel Said Reynolds, “Introduction: Quranic Studies and its Controversies,” in The Qurʾān in Its Historical Context, London and New York, Routledge, 2008, pp. 1-25, n. 7. ↩
Ibn al-Jazarî, al-Nashr fî l-qirâʾât al-ʿashr; al-Dimyâṭî, Itḥâf fuḍalâʾ al-bashar bi-l-qirâʾât al-arbaʿa ʿashar. ↩
Shady Hekmat Nasser, The Second Canonization of the Qurʾān, on the formation of the two-transmitter scheme in the Andalusian tradition. ↩
Muslim, no. 1050. ↩
Ibn al-Jazarî, al-Nashr, muqaddima, section “arkân al-qirâʾa al-ṣaḥîḥa.” The quotations that follow are from the same passage. ↩
Shady Hekmat Nasser, The Transmission of the Variant Readings of the Qurʾān: The Problem of Tawātur and the Emergence of Shawādhdh, Leiden, Brill, 2013. ↩
Al-Kulaynî, al-Kâfî, ii, kitâb Faḍl al-Qurʾân, bâb al-nawâdir; gradings in al-Majlisî, Mirʾât al-ʿuqûl. ↩
Al-Ṭûsî, al-Tibyân fî tafsîr al-Qurʾân, muqaddima; al-Ṭabrisî, Majmaʿ al-bayân, muqaddima; al-Khûʾî, al-Bayân fî tafsîr al-Qurʾân. ↩
Gabriel Said Reynolds, “Introduction: Quranic Studies and its Controversies,” op. cit., pp. 1-4; on the edition itself, Gotthelf Bergsträsser, “Koranlesung in Kairo,” Der Islam 20/1 (1932), pp. 1-42. ↩
Otto Pretzl, “Aufgaben und Ziele der Koranforschung,” Actes du XX^e Congrès international des orientalistes, Louvain, 1940, pp. 328-329. ↩

The Seven Aḥruf of the Qur’an (al-aḥruf al-sabʿa): Anatomy of an Exegetical Puzzle