By Avnar Sidiche
This article is focused on arguments that are commonly made in popular discourse about Quran preservation, often a result of ill informed apologetic mythologies. Firstly, the following may be a useful categorisation of Quran variants:
- Variants reportedly read by Muhammad’s companions (a few thousand). These often would not go on to fit ʿUthmān’s rasm standard, though those that did so sometimes feature in canonical readings (particularly Ibn Masʿūd’s in the Kufan readings). See the Resources page regarding these types of variants, particularly the works of Jeffery and Noldeke.l
- Other non-canonical readings. Numerous readings with different combinations of variants (many in non-compliance with ʿUthmān’s standard) were assembled over the next two centuries. These include vast numbers of variants that are not in the canonical readings but are attributed to early reciters. See in particular the works of Nasser and the non-canonical section on the Resources page. As discussed in another article on this site, it took three centuries before Ibn Mujāhid successfully promoted his selection of seven readings, and five more for Ibn al-Jazarī to elevate another three to canonical status.
- A few dozen variants between the original regional copies of the Uthmanic mushaf (codex). These were generally incorporated even into the canonical readings of those regions and are widely considered by academics to have been scribal errors. See Hythem Sidky’s paper On the Regionality of Qurʾānic Codices for the leading academic findings on this subject.
- Variants between the ten canonical readings and their transmissions (around 1400 words). These comply almost entirely with the Uthmanic rasm which had lacked markings to distinguish homographic consonants, short vowels and other diacritics. See the Resources page for Nasser’s work and some examples in the articles on this webite.
Now let’s dispell some of the very common popular mythologies or arguments.
“The differences are just dialects”
While some scholars believed that the seven ahruf referred to Arab dialects (certainly not the most defensible position when all the hadith evidence is taken into account), no-one can maintain that the variants between the qira’at are only different dialects. The examples on this website and many other lists of variants demonstrate otherwise. Christopher Melchert in his paper on The Relation of the Ten Readings to One Another (see Resources page) found in a sample that only 26% of canonical variants were conceivably of a dialect nature (ignoring large numbers of minor extremely common pronunciation issues).
“The variants are complementary even where they differ in meaning”
Naturally, no variant would find currency unless it made some kind of sense in context. Moreover, they had to fit the Uthmanic rasm constraint. Nevertheless, the dialogue variants article on this website provides sufficient evidence against this common claim. In addition, a large dose of imagination is required to reconcile some other more well known variants.
“The variants are a miracle of eloquence, providing significant additional meaning”
This highly dubious claim is untenable in light of the kinds of examples on the Superfluous variants article on this website. While technically, even a minor grammatical change by definition carries meaning, many variants render one of the pair redundant or sometimes there is a bizarre overkill of variants for a word.
“The Uthmanic rasm was deliberately ambiguous to accommodate as many of the ahruf as possible“
The rasm was typical even of early Arabic poetry manuscripts (Adam Bursi’s paper, “Connecting the dots”) in that it had sparse use of dotting to distinguish homographic consonants, lack of vocalisation diacritics for short vowels etc., and inconsistent use of alifs. Comprehensive dotting and the introduction of other diacritics (for a variety of readings, almost all non-canonical) only start to appear in Quran manuscripts of the 2nd and 3rd centuries. There was nothing deliberate about the ambiguity, and rather than being a strength, it led to a proliferation of variants.
“The rasm text is what was preserved”
This is certainly not the orthodox view, which is that even the vocalisation and consonantal dotting variants in the canonical transmissions are fully authentic. A large number of vocalised readings can and were imposed on the rasm, as Dr van Putten remarks (see the compliation of expert comments article on this website). In early manuscripts, non-canonical and unknown readings greatly outnumber canonical ones, though they mostly involve different permutations of the same variants, while variants attributed in the literature to various early reciters are much more varied (see below). A range of nuanced and sometimes substantial meanings was possible within the rasm.
As for the rasm itself, academics agree that the Uthmanic rasm has been very well preserved from its original compilation (aside from a small number of scribal errors in the copies made for each region and some spelling conventions), though the situation prior to the Uthmanic standardisation is much less clear. For this, we need only look at the variants in the companion readings, or in the Ṣanʽā’ 1 palimpsest which includes minor differences as well as transposed, omitted, substituted and additional words and phrases, and has a surah order similar to that reported of Ubayy b. Kaʿb (See the Sadeghi papers on the Resources page). Sadeghi in his 2010 paper based on a small portion of the folios (pages) now available describes the differences in Ṣanʽā’ 1 as first tier, meaning that they could have occured during dictiation errors such as misremembered synonyms and assimilation of other Quranic words and phrases, while those in companion readings such as Ibn Masʿūd’s also included higher tier differences. Nicolai Sinai in his 2020 review article mentioned below hopes that similar investigation will be done for the other folios.
In his 2012 paper transcribing many more folios, Sadeghi states that the Ṣanʽā’ 1 lower text “constitutes direct documentary evidence for the reality of the non-‘Uthmānic text types that are usually referred to as ‘Companion codices.'” His Appendix 1 lists around 70 cases of variants similar to those reported of various companions. Apologists here would argue that these are merely evidence of scribal errors or other ahruf (Ibn al-Jazari reported the majority opinion was that the Uthmanic rasm accommodated only some of the ahruf). The problem is the very large number and range of differences in Ṣanʽā’ 1 and companion readings (which both contain apparent errors, as Sadeghi notes), precluding full confidence in Uthman’s rasm standard as well as the qira’at variants which are meant to trace back to the companions.
“The Sanaa 1 palimpsest was a student exercise book”
This was Asma Hilali’s thesis based on a third of the available folios, but as of 2021 this ill-received claim is now considered completely debunked, firstly in a review article by Nicolai Sinai, “Beyond the Cairo Edition: On the Study of Early Qurʾānic Codices”, which Hythem Sidky (in his paper on the regional codices) cites for “criticism of her reading of the lower text and overall thesis”, and then more significantly by Éléonore Cellard’s paper, “The Ṣanʿāʾ Palimpsest: Materializing the Codices”, which Sidky cites “for a codicological reconstruction of portions of the undertext demonstrating the document’s status as a codex and the product of professional scribes”. Moreover, it was pointed out that animal skins for parchment were far too expensive to be used for practise books!
“The Birmingham manuscript shows that the Quran never changed”
This fragment (Mingana 1572a) was found to be actually part of a larger manuscript fragment (BnF Arabe 328c) and together does have various types of variant including minor meaning differences within it such as subject of verbs, as Alba Fedeli detailed in her PhD thesis. In any case, it is demonstrably not pre-Uthmanic despite its early radiocardon dating due to spelling idiosyncrasies shared with other manuscripts of the Uthmanic text type. See the Wikipedia article for further details on these points. It is not unlike all such manuscripts in having small differences. Nevertheless, we know that the Uthmanic rasm was well preserved as mentioned above.
“Companion readings are other ahruf / tafsir (exegesis)”
These were attempted Islamic scholarly explanations. However, it is striking just how many there are (thousands compiled by Jeffery – see Resources page). This shows why ʿUthmān’s standardisation was necessary to limit the range. Some may be exegetical glosses, though this is not a viable explanation in most cases, such as when words were omitted or for grammatical changes etc. and no-one can certainly say which are ahruf and which are exegetical (if either). In cases of replacement or additional words where the alternative wording is clearer, the Uthmanic wording seems to lose a useful purpose. Some such cases are discussed on the external blog linked on the resources page, which also covers instances where Ibn Abbas maintained that the Uthmanic reading was a scribal error.
Unless they all had perfect memories and each memorised every surah, it is likely that companions were the originators of many variants, canonical and non-canonical. Those with recorded precedents often are attributed only to a single companion like ʿAlī or Ibn Masʿūd. Even taking the story of Zayd b. Thābit’s collection process of written materials under Abū Bakr at face value, his rasm collection would not have included the consonantal dottings, short vowels, and other diacritics which make up the variants within the canonical readings.
“The companions unanimously accepted the Uthmanic standard“
There are various reports of companions such as ʿĀʾisha, ʿAlī, and Ibn ʿAbbās (as already mentioned) commenting on scribal errors in the Uthmanic mushaf. More famously, Kufa resisted the standardised rasm for over a century, prefering the reading of Ibn Masʿūd who strongly defended his reading (see Ramon Harvey’s paper on the Resources page).
“Everything canonical is authentically traced back to the Prophet”
If that was a well proven claim, it would be the end of the story. Unfortunately, there are numerous difficulties with this claim:
- Scribal errors in the original Uthmanic mushafs
Academics such as Marijn van Putten and Hythem Sidky (see Sidky’s paper on the Resources page, published in 2021, and this archived twitter thread) have given the following main reasons why the 40 or so differences between the rasm given to Medina, Syria, Basra and Kufa are scribal errors:
- The four codexes were clearly intended to be identical, which was also their stated purpose.
- The differences are very few and minor compared to the variants reported of companions.
- The manuscript variants across the regions reported by Islamic scholars like al-Dani were identified by Michael Cook as forming a stemma (copying tree). This was confirmed by Sidky using an improved list of the well attested differences, which he further confirmed by means of a phylogenetic analysis of manuscripts for the attested differences, earlier manuscripts than the Islamic scholars had access to. There are 4 possible stemmatic arrangements (each with the same connections between codices) based on the pattern of shared and isolated variants. It’s harder to work out which of those occured, though that doesn’t affect at all the fact that there is a clear stemmatic relationship.
- They are typical of the kinds of scribal errors in later copying that no-one denies.
- Sidky also showed that awareness and knowledge of the differences was documented organically and inconsistently by Muslim scholars comparing available manuscripts (i.e. this information was not supplied by ʿUthmān’s committee).
- Small number of single chains between the companions and each canonical reader
Scholars disagreed on whether the differences or just the agreement between the readings were mutawātir. As Nasser has noted (The Transmission of the Variant Readings of the Qurʾān pp. 110-116 – see Resources page) “all the Eponymous Readings were transmitted via single strands of transmissions (āḥād) between the Prophet and the seven Readers, which rendered the tawātur of these Readings questionable and problematic.” He observes that qira’at manuals were often silent on the isnad (chain of transmission) between the eponymous reader and the Prophet, documenting instead the formal isnads from the manual author to the eponymous reader. Like Ibn Mujahid, often they separately included various biographical accounts connecting the reading back to the Prophet, while later manuals developed more sophisticated isnads. Nasser concludes that “the dominant and strongest opinion among the Muslim scholars holds to the non-tawātur of the canonical Readings”.
- Limited attestation in the early manuscript record
Canonical readers are supposed to have picked from authentic variants passed down to them. However, as per the Compilation of expert comments article on this website, Dr van Putten has stated that some of these readings are not found in vocalised manuscripts of the first few centuries, though they were well documented by the likes of Ibn Mujahid. In the case of Hafs he found that not only was this reading unknown in such manuscripts, but those variants unique to Hafs among the canonical readers usually at best only appear in early vocalised manuscripts as part of secondary non-canonical readings added later. In general though it must be noted that variants in early vocalised manuscripts for the most part involve different permutations of those found also in the canonical readings. It is in the qira’at literature that we find a much wider range of variants documented from early reciters, both within and outside the Uthmanic rasm (more on that below).
- Divine status of all canonical variants was a late consensus – before then there was criticism and confusion
This is detailed particularly in Shady Nasser‘s first book. As also seen on the compliation of expert comments article on this website, major scholars like al-Ṭabarī, grammarians such as al-Farrā, and even Ibn Mujāhid himself said some variants now considered canonical were wrong. It’s quite obvious that, faced with a large array of transmissions, the only option for scholars was to deem those considered canonical to be fully authentic and try to justify variants previously considered problematic. See also the article on the final stages of canonization on this website.
Further, Chapter 4 of Nasser’s second book, The Second Canonization of the Qurʾān (324/936), published in 2020, shows that written notes played a significant role in transmission of the readings in the 2nd century. Despite their best efforts, some canonical readers and their transmitters were said to have doubts about their (often unique) readings. Abu ‘Amr, al Kisa’i, Nafi, and the transmitters of ‘Asim (Hafs and Shu’ba) are all reported “retracting a reading and adopting a new one” in some cases. Shu’ba “became skeptical” of his teacher ‘Asim’s reading of a certain word and adopted another, and said he “did not memorize” how certain words were read. In one instance Ibn Dhakwan, the transmitter of Ibn Amir’s reading, found one reading for a word in his book/notebook, and recalled something different in his memory. When the detailed recitation of a word was unknown, “the Qurrāʾ resorted to qiyās (analogy)”, as too did Ibn Mujahid when documenting the readings as he often faced conflicting or missing information.
- Dialect evident in the Quranic Consonantal text
Dr van Putten has a book due out in 2021 called Quranic Arabic: From its Hijazi origins to its classical reading traditions in which he will expand on the findings in his papers (see Resources page) that the Hijazi dialect evident in the rasm lacks various features found in the canonical readings (which contain a mixture of dialects).
- The sheer number of variants
It is commonly reckoned that c.1400 words in the Uthmanic Quran have canonical variants, about 2% of the total c.77,000 words. These are all within the Uthmanic rasm. Yet if there is any truth in reported companion readings, those indicate that there was no rasm constraint at the time of the original recitation, which would imply yet thousands more were uttered outside the standard rasm if the Uthmanic ones are all authentic.
Further, if we suppose the canonical variants are entirely authentic, then there’s no reason to assume they represent the only authentic variants that happen to fit the rasm. The Mu’jam al qira’at compilation mentioned on the Resources page has a dozen large volumes listing variants attributed to early reciters, canonical and non-canonical, within and outside the Uthmanic rasm.
So considering these two factors, the implicit numbers are scarcely credible. Moreover, such variety would be a recipe not for perfect preservation, but for confusion, false recollections and interpolations creeping into circulation at an early stage. We can see this phenomenum also in the many thousands of variants in other non-canonical readings, which cannot all be authentic.
- The nature of the variants
As discussed in other articles on this website, some of the dialogue variants are not credible as being deliberately uttered by the same person. In terms of tracing variants back even beyond Muhammad as divine revelation, both these and the superfluous variants are highly problematic. Other more well known canonical variants led to disagreements on rules surrounding fasting and wudu (and non-canonical variants reported by Ibn Masud on such matters as amputation were only used by some jurists – see Ramon Harvey’s paper on the Resources page). Another issue is that, as can be seen in Nasser’s list (Resources page), most of the few hundred canonical consonantal dotting variants involve verbal person or gender prefixes (y-, t-, n-), the correct reading of which would naturally be the most ambiguous among the rasm letters.
In summary, the probability is essentially zero that the canonical variants are entirely authentic.