Epigraphic evidence

According to its building notice, the Dome of the Rock was finished in 72 ah = 691/2 ce.Footnote ²⁵ Its arcade exhibits two mosaic inscriptions consisting of a series of Quranic segments interspersed with several instances of the basmala, various forms of the shahāda, and blessings on Muhammad and Jesus. The Umayyad portions of the copper plaques over the eastern and northern entrance to the Dome also string together a number of apparently Quranic phrases.Footnote ²⁶ In both cases, the Quranic material diverges in a number of instances from the standard rasm: for example, a phrase from Q 64:1 and two others from Q 57:2 are conflated into a statement of divine omnipotence that appears twice;Footnote ²⁷ and on one of the copper plaques, Q 7:156, a divine first-person statement appears in the third person. (Incidentally, similar observations apply to the use of Quranic material in early Islamic graffiti.Footnote ²⁸) To Robinson, all of this suggests that “Qur'anic texts must have remained at least partially fluid through the late seventh and early eighth century”.Footnote ²⁹ Shoemaker follows suit by qualifying the inscriptions of the Dome of the Rock as “perhaps the most prominent and inescapable” support for the “relative instability” of the Quran at the time of the building's construction.Footnote ³⁰ To be sure, in view of the high probability of a pre-660 date of Ṣanʿā' 1 such instability must have had clear limits, but as far as the palimpsest has so far been published it is not, for example, irreconcilable with a hypothetical claim that Q 112 may be a Marwanid addition to the Quran,Footnote ³¹ or that the statement of divine omnipotence on the Dome's arcade may only subsequently have been reworked into the opening verses of sūras 64 and 57.

Still, one must obviously ask whether other interpretations are possible. Strikingly, de Prémare is much less confident than Robinson and Shoemaker of the probative force of the epigraphic data.Footnote ³² And indeed there is much to recommend such an assessment: as Estelle Whelan has argued, divergences of the kind described above may be viewed as resulting from an adaptation of Quranic quotations to their epigraphic context, a procedure that is also observable in later inscriptions.Footnote ³³ For instance, a conversion of Q 7:156 from the first to the third person could have served to bring it into line with the preceding quotation (Q 6:12), also in the third person. Even if Whelan's explanation may not be the only tenable one, it certainly constitutes a perfectly satisfactory way of accounting for the evidence: Shoemaker's curt dismissal of her article as “special pleading” is therefore worryingly cavalier.Footnote ³⁴ In essence, then, the epigraphic data is indeterminate and compatible both with a traditional view of the Quran's codification and with the emergent canon model.

Al-Ḥajjāj and the Quran

At least two Umayyad governors of Basra and Kufa appear to have played some role in the Quran's textual history. First, there is the case of ʿUbaydallāh ibn Ziyād (killed 67/686). According to a report that Ibn Abī Dāwūd traces back to his scribe Yazīd al-Fārisī, ʿUbaydallāh added alfay ḥarfin to the codex, which could be translated either as “two thousand letters” or, more remarkably, as “two thousand words”.Footnote ³⁵ Ibn Abī Dāwūd explains that what ʿUbaydallāh ibn Ziyād did was to change the orthography of the words qālū and kānū from q-l-w and k-n-w to q-'-l-w-' and k-’-n-w-’. It is not prima facie obvious, of course, that this captures the original meaning of the tradition. De Prémare, obviously attracted to a maximalist construal of alfay ḥarfin as “two thousand words”, rejects Ibn Abī Dāwūd's interpretation on the grounds that a plene spelling of ā is already found in Muʿāwiya's inscription on a dam near al-Ṭā’if, which to him suggests that by the time of ʿUbaydallāh ibn Ziyād this spelling must also have become standard in Quranic manuscripts and therefore no longer needed promoting.Footnote ³⁶ However, this reasoning is refuted by the variation in the spelling of ā in early Quranic manuscripts.Footnote ³⁷ On the face of it, then, there is as much reason to view the report as associating ʿUbaydallāh with an increasing switch-over to plene spelling as to consider it to reflect a major overhaul of scripture.Footnote ³⁸

Let us turn to the second case: the strong interest reportedly taken by ʿAbd al-Malik's Iraqi governor al-Ḥajjāj in the Quranic text.Footnote ³⁹ Al-Ḥajjāj is said to have once convoked a group of Quran readers in order to count the text's consonants, words and verses, and to divide it into sections of equal length.Footnote ⁴⁰ He is also credited with requesting his scribes, or more specifically one Naṣr ibn ʿĀṣim (d. 89/707–8), a student of Abū al-Aswad al-Du'alī, to introduce diacritical signs into Quranic manuscripts.Footnote ⁴¹ What may be a secondary synthesis of such reports is given by the exegete Ibn ʿAṭiyya (d. 541/1146–7), according to whom al-Ḥajjāj, on the order of ʿAbd al-Malik, instructed al-Ḥasan al-Baṣrī and Yaḥyā ibn Yaʿmar to supply the Quran with diacritical marks and vowel signs, then had the text divided up into aḥzāb, and finally initiated the composition of a book on reading variants.Footnote ⁴² In addition, Ibn Abī Dāwūd, on the authority of the Basran ʿAwf ibn Abī Jamīla (d. 147/764–5),Footnote ⁴³ transmits a list of eleven passages for which al-Ḥajjāj allegedly “changed” the ʿUthmānic rasm, mostly by adding or subtracting single letters.Footnote ⁴⁴

Non-Islamic sources are more radical and portray al-Ḥajjāj as straightforwardly rewriting the Islamic scripture. The Christian apologist ʿAbd al-Masīḥ al-Kindī (early ninth century) asserts that “there is not a single codex which al-Ḥajjāj did not gather and from which he did not omit many things and to which he did not add many others” (the omitted passages allegedly concerned the Umayyads and the ʿAbbasids). He then had six master copies sent to Egypt, Damascus, Medina, Mecca, Kufa and Basra, while the “previous codices” were effaced with boiling oil, “thus imitating what ʿUthmān had done before him”.Footnote ⁴⁵ A similar accusation appears in a purported letter by the Byzantine emperor Leo III (717–41) to the caliph ʿUmar II (717–20) which is cited by the eighth-century Armenian chronicler Łewond: “one knows, among others, of a certain Ḥajjāj, named by you as Governor of Persia, who had men gather up your ancient books, which he replaced by others composed by himself, according to his taste, and which he propagated everywhere in your nation”.Footnote ⁴⁶

Similar steps, albeit not quite as drastic, are also alluded to in Islamic texts. The historians Ibn Shabba (d. 262/875–6) and al-Samhūdī (d. 911/1506) state that al-Ḥajjāj had copies of the Quran sent to the major cities of the empire.Footnote ⁴⁷ Al-Samhūdī, basing himself on Ibn Zabāla (d. after 199/814),Footnote ⁴⁸ additionally informs us that al-Ḥajjāj was the first to distribute maṣāḥif not only to the metropolises (ummahāt al-qurā), as ʿUthmān had done before, but also to smaller towns (qurā).Footnote ⁴⁹ That al-Ḥajjāj dispatched a Quranic codex as far as Egypt is confirmed by Ibn ʿAbd al-Ḥakam (d. 257/871), Ibn Duqmāq and al-Maqrīzī.Footnote ⁵⁰ Apparently al-Ḥajjāj's strategy of dissemination also included the novelty of instituting codex-based Quran recitation in mosques.Footnote ⁵¹ Islamic sources register some repressive measures, too: for example, al-Ḥajjāj reportedly established a small task force charged with inspecting Quranic codices and destroying those which were found to “disagree with the ʿUthmānic codex” (for which their owners received compensation of sixty dirhams);Footnote ⁵² and al-Farrā' (d. 207/822–3) mentions that the codex of al-Ḥārith ibn Suwayd, which was apparently based on the recension of Ibn Masʿūd, was “buried during the days of al-Ḥajjāj”.Footnote ⁵³

Finally, de Prémare has compiled a number of utterances ascribed to ʿAbd al-Malik and al-Ḥajjāj that would prima facie seem to lend support to the supposition that the two were engaged in significant redactional activity.Footnote ⁵⁴ Among them is a report from al-Balādhurī’s Ansāb al-ashrāf in which ʿAbd al-Malik describes the important role of Ramaḍān in his life by saying that it was during this month that he “collected” (alternatively, “memorized”) the Quran (jamaʿtu l-Qur’āna).Footnote ⁵⁵ And al-Ḥajjāj, according to a tradition in Muslim's Ṣaḥīḥ, once charged his audience during a sermon: allifū l-Qur’āna ka-mā allafahu Jibrīl ...Footnote ⁵⁶ – an injunction which de Prémare takes to mean “Compose the Quran as Gabriel has composed it!” and considers to have been “addressed to the scribes entrusted with the task of ta'līf al-Qur’ān”.Footnote ⁵⁷

What, then, are we to make of all this? Casanova's and Mingana's pleas for privileging al-Kindī’s account over the Islamic sources was partly based on their conviction that the former predated these latter, an assumption which is now obsolete.Footnote ⁵⁸ Nevertheless, the fact that two Christian texts which are not obviously interdependent, as well as various Islamic reports, concurrently ascribe to al-Ḥajjāj measures of textual dissemination and suppression strongly indicates that something of the sort really was afoot.Footnote ⁵⁹ It should also be noted that the fact that al-Kindī and Łewond depict al-Ḥajjāj as having revised a scripture that was already in the public domain (rather than as having compiled it in the first place) is perfectly consistent with the emergent canon model as outlined above, according to which the activity of ʿAbd al-Malik and al-Ḥajjāj would have constituted the terminus of a process that must have begun earlier. So should the material on al-Ḥajjāj that has just been presented be seen as supporting the emergent canon model?

In working through the evidence, it would be a mistake, I think, to set too much store by the statements ascribed to ʿAbd al-Malik and al-Ḥajjāj. For if the former's claim to have undertaken the jamʿ of the Quran in Ramaḍān referred to codification instead of memorization, the tradition would presumably not document an unguarded biographical reminiscence but is likely to have originated as a proud claim on the caliph's part to deserve credit for having collected the Islamic scripture. This requires that the Quranic text completed under ʿAbd al-Malik would not have been passed off as an ʿUthmānic text from the start, but would for a certain period have been openly flaunted as a Marwanid achievement, a stage of which the jamʿ tradition would constitute the last vestige. Only subsequently would this approach have been replaced by the spread of fictitious narratives about ʿUthmān's promulgation of the Quranic rasm. Such a picture, however, invites the query why the only trace of ʿAbd al-Malik's responsibility for the codification of scripture, which would once have been part of official state propaganda, is now contained in one isolated and highly ambiguous report. In the absence of more unequivocal evidence it thus seems entirely possible that ʿAbd al-Malik's statement does simply mean “In Ramaḍān I finished learning the Quran by heart”.Footnote ⁶⁰ As for al-Ḥajjāj's command allifū l-Qur’āna ka-mā allafahu Jibrīl, the innocuousness of the reading “Order the Quran as Gabriel ordered it!” = “Recite the Quran in its canonical order!”Footnote ⁶¹ must not be mistaken for implausibility: given that the tradition explicitly describes al-Ḥajjāj as “delivering a sermon from the pulpit”, de Prémare's suggestion that we are here confronted with an instruction given to an editorial team seems out of place (an early case of crowdsourcing?), while an exhortation about how to recite scripture would clearly be more appropriate.

The remaining material on al-Ḥajjāj, in particular reports about the destruction of codices and the dissemination of others, could perhaps be read as oblique reverberations of the distressing memory that the Quranic text had once undergone a significant makeover. It bears pointing out, though, that again there is nothing to preclude a more sedate understanding. The report that al-Ḥajjāj ordered the destruction of codices which “disagreed with the ʿUthmānic codex” tallies with information about his and ʿUbaydallāh ibn Ziyād's staunch opposition to the recension of Ibn Masʿūd, which was particularly popular in Kufa.Footnote ⁶² Consequently, al-Ḥajjāj's motive for the suppression of certain Quranic manuscripts may simply have been to buttress the position of one among several other existing recensions of the Quran.Footnote ⁶³ As for the codices that al-Ḥajjāj dispatched to various cities and towns, if we take the Islamic sources at face value our best guess would seem to be that these codices constituted a re-edition of the ʿUthmānic text that utilized (some) diacriticsFootnote ⁶⁴ and perhaps marked out the Quran's subdivision into sections of equal length. Whether al-Ḥajjāj's text also contained deliberate, albeit minute, changes, as reported by Ibn Abī Dāwūd, is less certain: in some cases he may only have given preference to an already existing variant, while in other cases the supposedly original reading may in fact be secondary, as Sadeghi has argued.Footnote ⁶⁵ Like the destruction of non-ʿUthmānic codices, the underlying aim of such a re-publication of the ʿUthmānic textFootnote ⁶⁶ would have consisted of bolstering its status over and against rival recensions. This would have made political sense as an assertion of Umayyad control, in particular over unruly Kufa with its strong pro-ʿAlid faction, where Ibn Masʿūd's version remained in use. Al-Ḥajjāj's solicitude for the text of the Quran would also have cast him and the caliph as pious guardians of revelation treading in the footsteps of the first member of the Umayyad family to have become caliph. On such a sedate reading, while al-Ḥajjāj could have played a role in the official imposition of the ʿUthmānic text, he was not necessarily responsible for a significant revision of it. Finally, the testimony of al-Kindī and Łewond could be accounted for as polemical attempts to harness these events, still remembered a century later, in order to cast doubt on the integrity of the Islamic scripture.

For a second time, then, our result is inconclusive: like the epigraphic data, the material on al-Ḥajjāj is compatible with the emergent canon model and with the traditional view that the standard rasm of the Quran existed by the mid-seventh century. The latter scenario would allow us to take most of what the Islamic sources say at face value, and it is not clear why, in the absence of compelling evidence to the contrary, this should not be our default position. The former view, of course, is much more attuned to the hermeneutics of suspicion that has become such an instinctive part of modern scholarly habits of reading. Nonetheless, it seems questionable to maintain, as a matter of principle, that when confronted with more than one adequate way of explaining our evidence we ought to choose the more iconoclastic one.

Even on a minimalist reading, however, it appears that as late as 700 manuscripts diverging from what was to become the standard rasm were still sufficiently prevalent in order for measures aimed at reinforcing the position of the so-called ʿUthmānic text to make sense.Footnote ⁶⁷ The latter's ultimate displacement of all other versions of scripture thus cannot have come in the immediate wake of the actions of ʿUthmān, even if these are viewed as historical. So did the standard rasm receive a major push from al-Ḥajjāj instead of ʿUthmān? The fact that according to ʿAwf ibn Abī Jamīla, al-Ḥajjāj's recension of the Quran differed in two places (23:87.89: allāhu instead of li-llāhi) from the standard rasm would seem to indicate that the text that we have is not identical with the version endorsed by al-Ḥajjāj.Footnote ⁶⁸ One should also take note of Ibn ʿAbd al-Ḥakam's report that al-Ḥajjāj's dispatch of one of his codices to Egypt was perceived as an affront by the governor ʿAbd al-ʿAzīz ibn Marwān, who then had his own codex produced; this suggests that al-Ḥajjāj's authority in the matter was regional at most and that he was not in a position to carry out an empire-wide standardization of scripture.Footnote ⁶⁹ Hence, one should probably not underestimate the significant role that an uncoerced attainment of acceptance “from the bottom up” is likely to have played in ensuring the ultimate ascendancy of the canonical rasm, albeit in tandem with al-Ḥajjāj's official measures.

Other Christian sources

In addition to the testimony of ʿAbd al-Masīḥ al-Kindī, Mingana also surveys other Christian writings on early Islam, such as the Dialogue between Patriarch John of Antioch (631–48) and a Muslim Emir Footnote ⁷⁰ or the chronicle of John Bar Penkāyē (probably written 687–8),Footnote ⁷¹ and concludes that “the Christian historians of the whole of the seventh century had no idea that the ‘Hagarian’ conquerors had any sacred Book”.Footnote ⁷² Such an argument from silence is of course easy to impugn,Footnote ⁷³ especially since it is now decisively contradicted by the early date of Ṣanʿā’ 1. A chronicler like John Bar Penkāyē, for example, concentrates on “recording current events as they impacted on the Christian communities”;Footnote ⁷⁴ it is therefore questionable whether we may expect him to discuss the scriptural canon of the Muslims.

De Prémare, too, accords an important position to Christian sources.Footnote ⁷⁵ Following Crone and Cook,Footnote ⁷⁶ he draws attention to a Syriac text from the first half of the eighth century, the Debate between a Monk of Bêt Ḥālê with an Arab Notable,Footnote ⁷⁷ which speaks of the Quran and of Sūrat al-Baqara as two distinct texts: “I think that even in your case, Muḥammad did not teach all your laws and commandments in the Quran, but you learned some of them from the Quran; some of them are in Sūrat al-Baqarah, and in G-y-g-y, and in T-w-r-h”.Footnote ⁷⁸ The statement might be construed as implying that the Quran that was known to the text's author was not identical with our Quran and perhaps formed a literary precursor of the latter. On the other hand, in view of the fact that the lower layer of the Ṣanʿā’ palimpsest does contain sections from Q 2, the author of the Debate may simply have misconstrued the way he heard Muslims speak about certain revelations being contained “in the Quran”, and others “in Sūrat al-Baqara”. Hoyland remarks that “in the Muslim tradition too there are indications that it [Sūrat al-Baqara] had a certain distinctiveness” and draws attention to the battle cry allegedly used at Ḥunayn: yā aṣḥāba sūrati l-baqarah.Footnote ⁷⁹ To an outsider, such a slogan might well imply that Sūrat al-Baqara is an independent Muslim scripture.

De Prémare attempts to strengthen his case with the chapter about Islam in John of Damascus' (d. mid-eighth century) De haeresibus,Footnote ⁸⁰ which first refers to a Muslim “book” (biblos) and later mentions four “writings” (graphê) composed by Muḥammad.Footnote ⁸¹ Only three of these “writings” bear titles corresponding to Quranic sūras, while the fourth, the “writing of the camel of God” (hê graphê tês kamêlou theou), can plausibly be connected to the Quranic story of the “camel of God” (nāqat Allāh, see Q 7:73, 11:64, and 91:13) that was killed by the Thamūd.Footnote ⁸² De Prémare suspects that this “writing of the camel of God” may have been a proto-Quranic text of which only fragments made it into the canonical recension of the Islamic scripture.Footnote ⁸³ He also notes that in his discussion of “the writing of the women” (= Q 4, Sūrat al-Nisā'?) John of Damascus refers to Muḥammad's marriage to the wife of Zayd, which is mentioned not in Q 4 but in Q 33:37, and to the Quranic statement that “your women are a tilth for you”, which occurs in Q 2:223. Hence, according to de Prémare, John of Damascus must be talking about “a text the organisation of which is noticeably different from that of the present sūra 4”.Footnote ⁸⁴

However, the De haresibus is unlikely to have been written before the 730s, i.e. at least three decades after al-Ḥajjāj had supposedly overseen the final redaction of the Quranic standard rasm. To postulate that John of Damascus would, in the 730s, still have based his presentation of Islam on a by then outdated pre-Marwanid version of the Quran strains credulity. Did the whole enterprise pass him by? Why would he be invoking an older version of the text without polemically capitalizing on al-Ḥajjāj's measures in a way similar to al-Kindī?Footnote ⁸⁵ It seems preferable, then, to suppose that the reference to a “writing of the camel of God” simply attests an early Islamic sūra name (for either Q 7, Q 11, or Q 91) which subsequently fell out of use. The fact that John of Damascus ascribes passages from other sūras to Q 4 could be a simple mistake caused by that sūra's title (“women”) and the consequent misconception that all important Quranic statements about marriage are concentrated therein.

Considerations of historical likelihood

Chase Robinson's cautious espousal of a Marwanid date for the codification of the Quran primarily relies on general considerations of historical likelihood.Footnote ⁸⁶ According to Robinson, the imposition of a standardized text of the Quran is difficult to envision under ʿUthmān, who was “deeply unpopular” in many quarters and ruled “a polity that lacked many rudimentary instruments of coercion and made no systematic attempt to project images of its own transcendent authority – no coins, little public building or inscriptions”.Footnote ⁸⁷ By contrast, ʿAbd al-Malik's coinage reform and his construction of the Dome of the Rock bespeak both his interest in deploying a specifically Islamic idiom and his disposal of the means required to carry out such measures, all of which makes his reign a more suitable context for the official promulgation of a uniform text of scripture.

There is no gainsaying the acuteness of these remarks. One way of accommodating them would obviously be to deny that ʿUthmān ever undertook the promulgation of a standard version of scripture. Still, it is not evident that this is the only possible conclusion. As was pointed out above, it appears that al-Ḥajjāj still found it necessary to repress Ibn Masʿūd's text and to promote the ʿUthmānic one. This does indeed create a strong impression that ʿUthmān did not achieve, or did not entirely achieve, the establishment of a uniform version of the Quran, but it hardly implies that he could not have tried. Robinson could therefore well be right to insist that ʿUthmān may not have been in a position to enforce the sole bindingness of one recension of the Quran, while ʿAbd al-Malik, given his imperial self-presentation and the more centralized nature of the Marwanid state, would have had both a motive and the means to give the Quranic recension favoured by him a considerable push. However, all of this primarily concerns the aspect of imposition: it has important implications for the question of when and how the standard rasm of the Quran became the sole authoritative version of scripture, and not necessarily for the question of when and how this recension reached its final form.

Discontinuities between Quranic legislation and early Islamic law

In the last section of this part I turn to Patricia Crone's case for a “mid-Umayyad” arrival of the Quran as presented in an article from 1994. Crone begins by reviewing a number of Quranic terms and passages with respect to which Islamic exegetes are clearly relying on guesswork rather than on any genuine recollection of the text's original meaning: in other words, the exegetical tradition does not generally seem to reach back to the first addressees of the Quranic recitations. Crone then focuses on similar gaps in the legal sphere, summed up in Joseph Schacht's famous verdict that “apart from the most elementary rules, norms derived from the Koran were introduced into Muhammadan law almost invariably at a secondary stage”.Footnote ⁸⁸ To be sure, Harald Motzki has now argued that already the early Meccan scholar ʿAṭā’ ibn Abī Rabāḥ (d. 114 or 115/732–734) explicitly based some of his legal opinions on Quranic verses.Footnote ⁸⁹ Nevertheless, Schacht's observation that in a number of cases the early Islamic legal tradition departs conspicuously from comparatively unequivocal Quranic stipulations remains valid. Particularly striking examples of such legal discontinuities are the refusal to recognize written documents as legal proof (contradicting Q 2:282) and the stoning penalty for zinā (contradicting Q 24:2). Crone herself presents two additional examples: first, the expression kitāb in Q 24:33, which Islamic exegetes generally understand to refer to a manumission document, whereas the context would clearly seem to require the meaning “marriage contract”; and second, a number of early legal traditions which possibly reflect a stage in Islamic legal thinking when the Quranic pronouncements awarding the non-agnatic relatives of a deceased certain fixed shares of the estate were not yet taken into account.

Crone insists that such discontinuities, when viewed through the lens of the conventional scenario of the Quran's codification, produce an intractable quandary. For if one accepts the commonsense assumption that Muḥammad implemented, or at least made a significant effort to implement, Quranic legislation, then practices which at some point were in conformity with Quranic law (such as the acceptance of written documents as legal proof) must within a rather short period of time have come to be replaced by practices that clearly violated Quranic law (such as the rejection of written documents as legal proof), in spite of the fact that the early Muslims would presumably have known, and been concerned to follow, the Quranic rules. Similarly, the original understanding of certain Quranic passages must have been lost and replaced by ingenious speculations. Crone finds such developments baffling and instead proposes a “mid-Umayyad date for the arrival of the canonical scripture”:Footnote ⁹⁰ “if ... the Quran was codified and canonized after the conquests, it ceases to be problematic that the reception of its legislation belongs to a secondary stage”.Footnote ⁹¹

Crone's article, then, accumulates circumstantial evidence indicating a surprising absence of the Quran from early Islamic intellectual history. It must be noted that the emergent canon model, whatever its merits, does not really provide a satisfactory explanation for this: for if it were true, one might have expected the incipient norms of Islamic law to have found their way into scripture, unless one were to introduce the auxiliary hypothesis that the circles responsible for the early development of the legal tradition were distinct from the circles transmitting proto-Quranic material. Even more compellingly, the carbon dating of the Ṣanʿā’ palimpsest makes it highly likely that by 660 a broadly familiar version of the Quran had come into existence and was being transmitted at considerable expense. Adding up Crone and the palimpsest, we are thus faced with the question how the Quran could have been both absent and present during the first Islamic century.

What may be a helpful paradigm is provided by the conventional narrative of how the works of Aristotle resurfaced from near-total oblivion when they were re-edited by Andronicus of Rhodes in the second half of the first century bce.Footnote ⁹² Although Crone's article does not address the issue explicitly, it can be construed as advocating precisely such a “hidden scripture” model, according to which the Quran may well have reached closure as early as 650, but nevertheless remained absent from Islamic history until c. 700, when it was secondarily co-opted, without much revision, into an existent religious tradition.Footnote ⁹³ To be sure, the Aristotelian paradigm must be considerably toned down to fit the situation of the Quran: at least isolated codices must have circulated, both because of Ṣanʿā’ 1 and because it seems excessive to dismiss the substantial and highly specific body of information found in Islamic sources about non-ʿUthmānic recensions, or the reports about al-Ḥajjāj's destruction and burial of scriptural manuscripts. It is also probable that not all parts of the corpus would have been equally “hidden”: while early Muslims may have known some Quranic material by heart, they may not generally have had access to complete manuscripts of the text, or systematically studied them, as a result of which certain passages could have inertly sat around in the midst of Quranic codices where nobody but an occasional scribe ever ventured.

Crone herself seems to dismiss the notion that the Quran could have been both present (in the sense of being transmitted in writing and selectively used for recitation) and absent (in the sense that sections of the text were not commonly known) at the same time: how could the early Muslims “have had a scripture containing legislation without regarding it as a source of law?”, she asks.Footnote ⁹⁴ Yet even today, believers who profess allegiance to a scripture without having more than a superficial understanding of what that text actually says are not an uncommon sight. Sacred writings, even if programmatically acknowledged to be reservoirs of truth and benchmarks of virtuous conduct, are not necessarily processed as bearers of concrete linguistic information. In particular if a sacred text's primary field of use is in ritual and devotional recitation, as seems to have been the case in early Islam,Footnote ⁹⁵ its semantic function can to some extent be suspended.Footnote ⁹⁶ Admittedly, it may be doubted whether this provides a convincing explanation for ignoring the normative import of straightforward injunctions like Q 24:2 (“The woman and man guilty of fornication, flog each of them with a hundred stripes”). But as noted above, most early Muslims' acquaintance with the Quran may well have been limited to “a few favorite passages and prayers, or certain selected verses that were reiterated as proof texts in political and doctrinal disputes”,Footnote ⁹⁷ while many sections could have constituted genuine blind spots. It is perhaps not surprising, then, that the process by which the early post-prophetic Muslim community (the majority of which had not been members of the Medinan Urgemeinde) took cognizance of its scripture should have been extremely gradual, and that it should have taken the form, not of scholarly exegetes systematically working through the text, but of decontextualized Quranic segments and keywords unpredictably percolating into the collective consciousness, where they inspired, attracted or merged with a host of popular narratives.Footnote ⁹⁸

Crone is surely right to insist that a proponent of the traditional scenario must assume the Quran's status to have been very different during the lifetime of Muḥammad: given that the latter presumably promulgated the Quranic revelations in order for people to understand and follow them, legally relevant Quranic passages must to some extent have been applied, endowing the Quran with the status of a “source”, rather than just that of a “document”, to put it in terms coined by John Burton.Footnote ⁹⁹ Consequently, proponents of an early dating of the Quran find themselves committed to an evolutionary trajectory leading from a stage at which the Quran functioned as a normative source (during Muḥammad's lifetime) to a stage at which it did not – or not primarily or invariably – function as such (during the seventh century), to a stage at which it was again taken seriously as a source of behavioural norms, and subjected to systematic exegetical decoding (from the eighth century onwards). Yet such a to and fro, although messy, would not be historically incomprehensible. As a result of the rapid growth of the Islamic community and its geographical expansion over a vast area, the Quranic corpus would have undergone a far-reaching disembedding. Hence, instead of thinking of the post-conquest umma as essentially an extension of the prophetic umma, we should perhaps envisage them – in spite of certain personal continuities – as two separate communities, in much the same way as urban Hellenistic Christianity was distinct from, rather than a mere extension of, early Palestinian Christianity. Such a shift of perspective calls into question the assumption that those Quranic norms which, on the traditional model, must have been put into practice in the context of the Medinan Urgemeinde ought to have remained intact, or that the meaning of specific Quranic expressions ought to have filtered down unscathed from the prophetic to the post-prophetic community. Although the Islamic tradition is generally concerned to depict the early Muslims as meticulously passing on detailed historical and exegetical remembrances of the Prophet's companions, it seems rather more probable that during the age of the conquests the majority of converts were not sufficiently preoccupied with the interpretation of the Quran in order for the prophetic community's understanding of it to be fully preserved. As a result, later Muslims needed to rediscover and hermeneutically reinvent their scripture.

To conclude this part of the article: all the data examined so far seem compatible with the conventional dating of the Quran's codification in a suitably modified version (circulation of several rival recensions even after 650, selective and predominantly liturgical use of the Quran until the end of the seventh century). The emergent canon model certainly remains in the race, although neither the epigraphic nor the literary evidence marshalled by its supporters strictly speaking requires it, and the legal and exegetical discontinuities foregrounded by Crone can only be squared with it, by bringing in the auxiliary assumption that the circles involved in the transmission of proto-Quranic material were separate from the circles at the forefront of early Islamic legal thought and unconcerned to straighten out those bits of the text that had become unintelligible. In Part II I shall go on to discuss the weightiest arguments in support of a mid-seventh century or earlier date for the standard rasm of the Quran.