Skip to main content Accessibility help


  • Access
  • Open access


      • Send article to Kindle

        To send this article to your Kindle, first ensure is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about sending to your Kindle. Find out more about sending to your Kindle.

        Note you can select to send to either the or variations. ‘’ emails are free but can only be sent to your device when it is connected to wi-fi. ‘’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

        Find out more about the Kindle Personal Document Service.

        Hišām's ʾIbrāhām: Evidence for a Canonical Quranic Reading Based on the Rasm
        Available formats

        Send article to Dropbox

        To send this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Dropbox.

        Hišām's ʾIbrāhām: Evidence for a Canonical Quranic Reading Based on the Rasm
        Available formats

        Send article to Google Drive

        To send this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Google Drive.

        Hišām's ʾIbrāhām: Evidence for a Canonical Quranic Reading Based on the Rasm
        Available formats
Export citation
Please note a correction has been issued for this article.


In early quranic Manuscripts the name of the prophet ʾIbrāhīm occurs in two different spellings either ابرهم or ابرهيم . These two spellings are spread haphazardly throughout the Quran. Close examination of the patterns in the manuscript, however, show that the distribution of this spelling is not random or up to the whims of the scribe. The location where one spelling or the other occurs is highly correlated across the early manuscripts. Moreover, the location of one spelling or the other is highly correlated to where the quranic reader Hišām reads the name as ʾIbrāhām or ʾIbrāhīm. This paper argues that this is not because these manuscripts have been written in the reading of Hišām, but rather that Hišām based his reading on the rasm of the quranic text.



I wish to thank Sean Anthony, Hythem Sidky, Tommaso Tesei, Maarten Kossmann, Benjamin Suchard, Fokelien Kootstra for commenting on earlier versions of this paper.


The prophet ʾIbrāhīm, equivalent to the biblical Abraham, presents a conundrum in terms of quranic orthography. The name occurs in two different spellings, first as ابرهم and second as ابرهيم . In the ubiquitous Cairo edition of the Quran the distribution of these two names is striking: all attestations of the name in Sūrat al-Baqarah are spelled ابرهم , whereas all other attestations are spelled ابرهيم .

Despite this difference in spelling, the most common reading tradition today, that of Ḥafṣ (d. 180/796) from ʿĀṣim (d. 127/745)—the one used in the Cairo edition—as well as most other reading traditions read this word as ʾIbrāhīm in all its instances. This is the only case of an ī spelled defectively in the quranic orthography that cannot be explained some other way.Footnote 2

Many people rightly recognise ابرهم to be the spelling one would expect for the name ʾAbrāhām, as it was borrowed from Hebrew or Aramaic, whereas ابرهيم represents a more nativised form based on biblical names with similar patterns like ʾIsmāʿīl and ʾIsrāʾīl.Footnote 3 If ابرهم is indeed an archaic spelling, reflecting its Aramaic spelling, the distribution that we find in the Cairo edition is rather attractive. The fact that al-Baqarah—the largest, and thematically somewhat isolated, sūrah—contains a different spelling could easily be interpreted as an indication that the sūrah was perhaps written by a different scribe, who used a more archaic spelling of this name, and may even have pronounced the name differently. The explanation however does not agree with the manuscript evidence that we have. In early quranic manuscripts the spelling ابرهم is more widespread, and is found in sūrahs often even mixed with other spellings. For example, in the Codex Parisino-Petropolitanus (henceforth CPP) we find Q. 19:46 spelled ابرهيم , but Q. 19:58 spelled ابرهم .Footnote 4

Besides these issues, there is another confounding factor that complicates the issue of these two spellings. Throughout the qirāʾāt literature, there is mention of the name ʾIbrāhīm being read as ʾIbrāhām in multiple places. The famous canonisation of the seven readings, kitāb al-sabʿ fī al-qirāʾāt by ibn Muǧāhid (d. 324/935), says that ibn ʿĀmir (d. 118/736) would read ʾIbrāhām rather than ʾIbrāhīm in all of Sūrat al-Baqarah, tracing this knowledge back to al-ʾAḫfaš al-Dimašqī, who in his turn traces it back to ibn Ḏakwān (d. 242/856), one of the two canonical transmitters of ibn ʿĀmir's tradition. Ibn Muǧāhid makes no mention of the way ibn ʿĀmir's other canonical transmitter, Hišām (d. 245/859), treats the name ʾIbrāhīm.Footnote 5 The reading ʾIbrāhām exactly matches the distribution of the spelling as we find it in the Cairo edition of the Quran, which is unlikely to be a chance correspondence.

Different from ibn Muǧāhid, al-Dānī (d. 444/1053) does comment on Hišām in his al-Taysīr fī al-qīrāʾāt al-sabʿ.Footnote 6 He agrees with ibn Muǧāhid that ibn Ḏakwān would exclusively read ʾIbrāhām in Sūrat al-Baqarah, but also allows the reading as ʾIbrāhīm. However, the places where Hišām reads ʾIbrāhām is significantly more idiosyncratic. An overview of all attestations of the name, and the pronunciation associated with it by al-Dānī are in the table below.

Table 1. The readings of ʾIbrāhīm by Hišām from ibn ʿĀmir.

Ibn al-Ǧazarī’s (d. 822/1419) Našr al-Qirāʾāt al-ʿAšr describes Hišām in the same way as al-Dānī, but for ibn Ḏakwān he transmits a variety of accounts that have different approaches.Footnote 7 Several accounts say that ibn Ḏakwān read ʾIbrāhām in the same places as Hišām. Others say he always read ʾIbrāhīm. Another account says that it was only read ʾIbrāhām in Sūrat al-Baqarah, and some would say both ʾIbrāhīm and ʾIbrāhām would be possible in Sūrat al-Baqarah. Finally, one transmitter adds all cases of ʾIbrāhīm in Q. 3 and Q. 87 to forms that are pronounced as ʾIbrāhām, but ibn al-Ǧazarī says that this is a mistake.

What is interesting is that, unlike ibn Muǧāhid and al-Dānī, ibn al-Ǧazarī does not only comment on the different variations on the reading traditions, he also observes that these reading traditions seem to match spellings in quranic codices. He says that in the 33 places where Hišām reads ʾIbrāhām, the name is written as ابرهم in Syrian Codices, and that he has also seen this in Medinese Codices, and that some of these would only have such a spelling in Sūrat al-Baqarah.Footnote 8

The Medinese codices that only write ابرهم in Sūrat al-Baqarah correspond exactly in this orthographic idiosyncrasy to what we find in the Cairo edition. This then obviously causes one to wonder whether it is the case that also the Syrian distribution, as described by ibn al-Ǧazarī, shows up in old quranic codices.

In this paper, I will show that the occurrences of ابرهم in early quranic manuscripts, which are now available and easily accessible online via the Corpus Coranicum website,Footnote 9 correspond remarkably well to the places where Hišām is said to have read ʾIbrāhām. It will be shown that this cannot be because the rasm of these manuscripts was adapted to represent Hišām's reading tradition. Instead, it must be understood as a case where Hišām's reading tradition adapted the reading to the rasm.

The Manuscripts

Nearly all of the early quranic manuscripts available on the Corpus Coranicum website either match the places where ibn al-Ǧazarī reports the ابرهم spelling perfectly, or almost perfectly.Footnote 10

Kairo, al-Maktabah al-Markaziyyah li-l-Maḫṭūṭāt al-Islāmiyyah: Großer Korankodex

The Großer Korankodex corresponds closely to the distribution as described for Hišām. There are six disagreements, two of which are almost certainly later additions. The six points of disagreement are:

For Q. 2:132 we are almost certainly dealing with a later addition of the denticle for the ī, as the ductus is much thinner than other instances of the denticle, see Fig. 1. Q. 2:125, the ductus is similar to the word, but the connection of the denticle to the م is different than in other parts of the manuscript, compare Figs. 2 and 3.

Fig. 1. Q. 2:132

Fig. 2. Q. 2:127

Fig. 3. Q. 3:97

This manuscript attests 68 instances of the 69 attestations in the Quran (the only missing one being Q. 3:33). If we tabulate these results, taking Q. 2:125, and Q. 2:132, as spelling ابرهم , we get the following overview, the columns designate the spelling in the manuscript, and the rows designate the reading of the name as either ʾIbrāhām or ʾIbrāhīm in Hišām's tradition. The numbers between brackets designate the numbers if Sūrat al-Baqarah is removed from the data.

Table 2. Distribution in the Großer Korankodex.

The statistical procedure known as Fisher's exact test is able to calculate the odds of this situation occurring by chance. The resulting p value is the probability that the correspondence would be due to chance, a p value of 0.05 is equivalent to a chance of one in twenty (0.05 = 1/20) that it would be due to chance. Calculating the p value, we get a value so significant, that chances of this happening by chance are virtually zero (a chance smaller than one in 52 trillion): p = 0.000000000000019

Even if we take Q. 2:125 and Q. 2:132 as spelling ابرهيم , the p value is still virtually zero (p = 0.0000000000016).

If we remove Sūrat al-Baqarah from the calculation, as this is of course the position where we also find ابرهم in the Cairo edition, we still get a highly significant result. (p = 0.00000000009).

This extremely high level of correlation between where Hišām reads ʾIbrāhām or ʾIbrāhīm versus the spelling ابرهم and ابرهيم shows clearly that this manuscript is related to a manuscript that ibn al-Ǧazarī describes.

Tübingen, Universitätsbibliothek: Ma VI 165

Ma VI 165 has four attestations of ابرهم (Q. 19:41, 46, 58; Q. 29:31) and nine attestations of ابرهيم (Q. 21:51, 60, 62, 69; Q. 22:26, 43, 78; Q. 26:69; Q. 29:16), these match perfectly with Hišām. Interestingly, we can see in Q. 29:31 was originally spelled ابرهيم and the denticle has been purposely removed to represent ابرهم , corresponding to Hišām's ʾIbrāhām (see Fig. 4). The correspondence is highly significant correlation (p = 0.0014). Assuming that Q. 29:31 represents ابرهيم , the correlation is still significant (p = 0.014).

Fig. 4. ابرهيم with the mistaken ى partially removed in Q. 29:31.

Kairo, Nationalbibliothek: qāf 47

There are five instances of ابرهم in Qāf 47 (Q. 4:125 (twice); Q. 9:144 (twice); Q. 14:35) and six instances of ابرهيم (Q. 3:65, 67, 68; Q. 6:74, 75; Q. 21:62). They perfectly correspond to the reading of Hišām (p = 0.0021).

Berlin, Staatsbibliothek: Kodex Samarkand (Facsimile)

The Samarkand Codex attests quite a few instances of ابرهـ [ـيـ]ـم . Surprisingly, the last four attestations of ʾIbrāhīm in Sūrat al-Baqarah are written ابرهيم , disagreeing both with the Cairo edition and the tradition of Hišām. Despite this, the correlation is highly significant both including Sūrat al-Baqarah (p = 0.0000083) and excluding it (p = 0.001).

Table 3. Distribution in the Kodex Samarkand.

Codex Parisino-Petropolitanus

Dutton observed that there was a relationship between the reading tradition of Hišām and the places where the different spellings show up.Footnote 11 Déroche comments on the presence of the spelling ابرهم in three places, but his edition misses seven other places where ابرهم instead of ابرهيم is written.Footnote 12 Five of these agree with Hišām's reading (Figs. 5–9).

Fig. 5. Q. 6:161 اىرهم

Fig. 6. Q. 9:114 (first) اىرهم

Fig. 7. Q. 9:144 (second) اىرهم

Fig. 8. Q. 14:35 اىرهم

Fig. 9. Q. 42:13 اىرهم

There are two places where the CPP quite clearly has ابرهم without the yāʾ where we would expect it to be present if it followed Hišām's reading tradition.

The first one is Q. 43:26 (Fig. 10), the other is Q. 3:33 (Fig. 11). Q. 3:33, however, clearly is the correction of a later hand and close examination of the high resolution photos shows that the original form of this word, before correction, was spelled ابرهيم . Dutton is therefore correct in saying that Arabe 328(a) aligns perfectly with the reading tradition of Hišām, but the broader manuscript of the CPP shows at least one deviation (Q. 43:26 is part of Arabe 328(b) not examined by Dutton).Footnote 13

Fig. 10. Q. 43:26 اىرهم

Fig. 11. Q.3:33 اىرهم with erased ابرهيم below.

There are two more instances of ابرهم /ابرهيم in the St. Petersburg portion of the CPP (Marcel 18). Q. 57:26 based on the reading Hišām would be expected to be spelled ابرهم and Q. 26:69 would be expected to be spelled ابرهيم . Both of these are read by Déroche as ابرهيم . As I do not have access to photos of this manuscript, I have decided to not include these two forms in the calculation below. The resulting correlation is highly significant (p = 0.0004).

Table 4. Distribution in the CPP.

Istanbul, Topkapı Sarayı Müzesi: M1 (= Saray Medina 1a)

Saray Medina 1a agrees in quite a few attestations of ابرهم , although in some environments the now regular spelling is attested. While the number of corresponding ابرهم spellings is a minority, the relation is still highly significant (p = 0.00028).

Table 5. Distribution in Saray Medina 1a.

Berlin, Staatsbibliothek: Wetzstein II 1913 + Paris, Bibliothèque nationale de France: Arabe 6087

Wetzstein II 1913 and Arabe 6087 are correctly identified by the Corpus Coranicum project to be part of the same codex. This fairly complete early codex can also be said to agree with the reading tradition of Hišām. However, it is clear that the text has been retouched in many places. Surprisingly, in Sūrat al-Baqarah, it seems that all instances of ابرهم have received an added yāʾ. While in some cases, it is quite clear that these are later additions (e.g. Fig. 12), in other cases this is not as obvious (Fig. 13) and the writing is almost indistinguishable from places where we would expect the spelling ابرهيم (Fig. 14). If we take the most conservative approach and assume that all instances of ابرهيم are original and not retouched, the correlation is not significant if we include Sūrat al-Baqarah (p = 0.0951); however, if we reasonably exclude Sūrat al-Baqarah from the calculation, the correlation is significant (p = 0.0098).

Fig. 12. ابرهم with ى clearly added in Q. 2:258 (second)

Fig. 13. ابرهيم with possibly an added ى in Q. 2:135

Fig. 14. Example of genuine ابرهيم spelling in Q. 6:75.

Table 6. Distribution in Wetzstein II 1913.

London, British Library: Or. 2165 & Paris, Bibliothèque nationale de France: Arabe 328(e)

As Déroche correctly identifies,Footnote 14 Or. 2165Footnote 15 and Arabe 328(e) belong to a single codex. All 34 attestations of ʾIbrāhā/īm correspond perfectly with the reading tradition of Hišām. This was already observed by DuttonFootnote 16 for just Or. 2165. But also if we include Arabe 328(e) this is the case.

The line on which Q. 43:26 appears is written by a very different hand, and seems to correct some damage on the manuscript. I have left it out of the calculation. The correlation is highly significant (p = 0.0001).

Codex Amrensis 1

The recently published edition of the quranic codex that goes by the name of Codex Amrensis 1Footnote 17 has four attestations of ابرهم (Q. 9:114 (twice); Q. 14:35; Q. 42:13) and seven attestations of ابرهيم (Q. 9:70; Q. 11:69, 74, 75, 76; Q. 15:61; Q. 60:4 (second mention)). This is in perfect agreement with the reading tradition of Hišām, and as such the correlation is significant (p = 0.003).

An absence of early Qurans of the Baqarah-only type

An interesting observation that can be made from the early quranic manuscripts that are known to us is that there is a—perhaps surprising—lack of Qurans that have the distribution that corresponds to the tradition of ibn Ḏakwān with ابرهم in Sūrat al-Baqarah and ابرهيم everywhere else—the distribution found in the Cairo edition today. Qurans of this type must have certainly existed during al-Dānī’s lifetime (b. 371/981-2, d. 444/1053) as he describes the Quran as having such a rasm.Footnote 18

The only Quran that is of the Baqarah-only type that is available on the Corpus Coranicum website at the time of writing is the virtually complete Kufic Quran codex of ʿAbdarraḥmān bin Zīdān from the Gotthelf-Bergsträßer archive.Footnote 19 This is clearly a later quranic text than some of the ones we have discussed so far, being written in a clear Kufic B II script, rather than the Hijazi script styles of the CPP, Saray medina 1a, Ma VI 165,Footnote 20 Or. 2165 and Qāf 47. It seems therefore that the Baqarah-only type as found in the Cairo Edition is a development of the manuscript tradition that post-dates the distribution that correlates to that of the Hišām reading tradition. This should probably be seen as a regularisation of the rather haphazard distribution of the two spellings for ʾIbrāhīm as it is attested in these early manuscripts, retaining the defective spelling ابرهم only in Sūrat al-Baqarah, where the spelling was consistent and common in the early manuscripts.

Interpreting the data

In the previous sections, it has been shown that in a variety of old quranic manuscripts, the correlation of the spelling ابرهم to the pronunciation ʾIbrāhām in the Hišām reading tradition is highly significant. The observation by ibn al-Ǧazarī that such quranic manuscripts exist is confirmed by the evidence available to us. Surprisingly, however, there appear to be no early Quran manuscripts that display the pattern as it is found in the modern Cairo edition today. It therefore seems that such quranic manuscripts are in fact later than those which have the distribution that is similar to Hišām; so much later, in fact, that they seem to have fallen almost completely outside of the scope of the types of manuscripts that the Corpus Coranicum project aims to collect on their website.

Of course, the extremely high correlation with the reading tradition of Hišām requires an explanation. I see three possible options:

  1. 1. All these manuscripts represent the reading tradition of Hišām, and the rasm of the manuscripts was adapted to represent this reading tradition.

  2. 2. The variable pronunciation of ʾIbrāhīm and ʾIbrāhām in specific ayahs in the Quran used to be much more widespread in other reading traditions, and these manuscripts represent this.

  3. 3. Hišām based his reading of ʾIbrāhīm and ʾIbrāhām on the spelling that was present in a codex available to him. This codex must have been very close to the quranic manuscripts discussed here.

We will discuss these three possibilities separately.

Option 1: All these manuscripts represent the Hišām tradition

One reason why the early quranic manuscripts examined so far seem to agree rather strikingly with ibn ʿĀmir's (d. 118/736) reading tradition as transmitted by Hišām (d. 245/859) could simply be that these Qurans were written to represent this reading tradition. If this is the case, it would mean that the rasm of the text was purposely altered to accommodate this reading tradition. In light of the highly defective writing of early quranic manuscripts, this is unlikely. If the rasm was freely adapted to accommodate certain reading traditions, we might expect many more subtle changes to the rasm to accommodate such variants, which do not seem to occur.

To test this hypothesis, the most obvious approach is to examine variant readings of the quranic text in these manuscripts and see if they correspond to the reading tradition of ibn ʿĀmir, and if possible more specifically to that of his transmitter Hišām. This latter option is unlikely, as several of the early quranic manuscripts examined here certainly predate Hišām's lifetime (d. 245/859). Such an examination is not always easy, as reading traditions only seldom disagree in terms of the consonantal skeleton, and consonantal dots often are not used.

DuttonFootnote 21 suggests that the CPP must indeed be a quranic manuscript related to the ibn ʿĀmir reading tradition of the late Umayyad period based on thirteen consonantal variants of which six are associated exclusively with ibn ʿĀmir. DuttonFootnote 22 also observed that the spelling of ʾIbrāhīm/ʾIbrāhām in both the CPP and Or. 2165 accord with the reading tradition of Hišām from ibn ʿĀmir and he tentatively concludes that this may also be a piece of evidence that the CPP is representing this reading tradition.Footnote 23

Dutton bases the identification of these manuscripts as part of the Syrian tradition on the comments in the rasm literature such as al-Dānī’s Muqniʿ, which describes the differences in the consonantal skeleton of the Uthmanic text as they are found in different regional codices—primarily citing variants of the four provinces of Syria, Medina, Basrah and Kufah. These variants have been studied in detail by Michael Cook,Footnote 24 who shows that these make up a manuscript stemma. The overview of the regional variants in Cook's article allows us to identify the regional variants that are present in a manuscript, and with that assign it a region.

From the CPP and Or. 2165, both Syrian manuscripts which both have the haphazard distribution of the spelling of ʾIbrāhīm that correlates with the tradition of ibn ʿĀmir, one might indeed conclude—as Dutton tentatively does—that this feature reflects Syrian regionality. However, this would only be true if all other quranic manuscripts that clearly correlate the spelling of ʾIbrāhīm also contained the Syrian rasm variants. This, however, is certainly not the case. While Wetzstein II 1913, like the CPP and Or. 2165, is a codex of the Syrian tradition, the other manuscripts examined clearly belong to different traditions.

Using Michael Cook's overview of the different consonantal differences, we can establish to which codex traditions these manuscripts belong.Footnote 25 For an overview of the features I refer the reader to the Appendix. We can sum up here the likely provenance of the manuscripts discussed here. All four of Cook's manuscript types are present:Footnote 26

Syrian: CPP; Wetzstein II 1913; Or. 2165

Medinan: Saray Medina 1a; Codex Amrensis 1; Qāf 47; Ma VI 165 (or Syrian? or Basran?)

Basran: Großer Korankodex

Kufan: Samarkand Codex

We must therefore conclude that in the quranic manuscripts where the spelling ابرهيم /ابرهم correlates with the reading of Hišām, this is not because the quranic manuscripts are intended to represent Hišām's transmission of the ibn ʿĀmir reading tradition.

Option 2: The two variant readings of ابرهـ (ـيـ )ـم used to be more widespread

It is of course possible that originally all the reading traditions and not just that of Hišām pronounced the name as ʾIbrāhām where it is spelled ابرهم . In this case, despite the manuscripts examined not representing ibn ʿĀmir's tradition, they could still be intended to represent such readings. While this is not impossible, there is absolutely no positive evidence for this.

Moreover, in this scenario, we would still be unable to explain the source of disagreement between the tradition of Hišām and of ibn Ḏakwān, who both read ʾIbrāhām, but disagree on the positions where this is done. This is different from option 3 where, as we will see, the source of disagreement between Hišām and ibn Ḏakwān is easily explained.

Option 3: Hišām based his readings on the spelling of the codex

As the first option is not possible, and the second option is unlikely, the third option remains: a reversal of the scenario. Hišām based his reading on the spelling in his codex.

It seems likely that in the establishment and canonisation of the tradition of Hišām from ibn ʿĀmir, it was decided upon that every time the quranic manuscript had the spelling ابرهم it would be read ʾIbrāhām, whereas when it had the spelling ابرهيم it would be read ʾIbrāhīm. This tradition was then later decoupled from the spelling as present in the rasm of the manuscript it was based on, and came to take on a random-looking distribution when overlaid over the Cairo Edition rasm.

From this principle we can not only understand the Hišām tradition, but also the tradition of the other transmitter of ibn ʿĀmir – ibn Ḏakwān. We know that there are (at least) two manuscript traditions when it comes to the distribution of the ابرهيم and ابرهم spellings. The first tradition can be called the Baqarah-only tradition. Belonging to this manuscript tradition, we most notably have the Cairo Edition of the Quran. The other tradition can be called the Hishamoid tradition, which has ابرهم not only in Sūrat al-Baqarah, but also in a variety of other places throughout the Quran.

If we assume that the original principle of the tradition of ibn ʿĀmir was simply to read ʾIbrāhīm wherever ابرهيم was written, and ʾIbrāhām wherever ابرهم was written, we can understand the differences between the two transmitters of ibn ʿĀmir. The tradition of Hišām based itself on a quranic manuscript similar to, for example, the CPP and the Großer Korankodex whereas the tradition of ibn Ḏakwān based itself on a quranic manuscript similar to the Cairo Edition or the ʿAbdarraḥmān b. Zīdān Kufic Quran.

This does not mean that we can therefore date the innovation of the Baqarah-only type quranic manuscript to the period of ibn Ḏakwān's lifetime (d. 242/856). If ʾIbrāhām was read whenever ابرهم was written was indeed the rule, it is possible that the reading only started agreeing with the Baqarah-only type in later transmitters of his tradition. The disagreement amongst the transmitters on where ibn Ḏakwān read ʾIbrāhām is a strong indication that a cause for disagreement only developed after his lifetime.Footnote 27

The relation between the rasm and the reading traditions

I have argued in the previous section that the variation of the ʾIbrāhīm and ʾIbrāhām pronunciations in the ibn ʿĀmir reading tradition should be considered a reading based on the rasm of codices which had these different spellings. This finding has important implications on how we understand the relation of the rasm to the reading traditions.

The important studies by DuttonFootnote 28 form a fundamental basis for looking at early quranic manuscripts in light of their reading traditions. In these two articles, Dutton clearly shows that both the CPP and Or. 2165 display clear peculiarities of their consonantal skeleton that are also present in the descriptions of the reading tradition of ibn ʿĀmir.

Dutton suggests that because of this, the CPP and Or. 2165 were written according to the reading of ibn ʿĀmir.Footnote 29 In other words, he believes that the manuscripts of this type could only have existed in a period that the reading tradition of ibn ʿĀmir had taken shape. This is an important fact because DuttonFootnote 30 dates the manuscript on the assumption that it cannot have existed before the time that ibn ʿĀmir's reading tradition had taken its final shape “[i]f […] Ibn ʿĀmir's reading was not really fixed until the time of those later considered as the main rāwīs from him, i.e. Hishām (d. 245/859) and Ibn Dhakwān (d. 242/856), then this could simply be a late (i.e. Abbasid) example of a parchment, vertical-format, Ḥijāzī manuscript[…]”.Footnote 31

This, however, presupposes that the rasm of this codex was changed from the Uthmanic archetype in order to accommodate the reading tradition. And while this is possible, this is not the only solution. The other possibility is that when the copies of the Uthmanic codex were sent out to the provincial capitals, these variations were present in these initial copies. Due to a predominant ideology that an authoritative reading has to agree with the Uthmanic rasm,Footnote 32 the local readers would have adjusted their reading tradition to the local codices available to them.Footnote 33 As a result, the reading tradition of ibn ʿĀmir, as an authoritative Damascene reader, would naturally have a reading that agreed with the deviations as present in the Syrian codex and its copies. Had this not been the case, he would not have been eligible to be part of ibn Muǧāhid's canonisation.Footnote 34

The fact that such a scenario is likely, is already suggested by Cook's seminal workFootnote 35 on the regional codices, in which he convincingly shows that the rasm variations described by al-Dānī in his Muqniʿ fī Rasm Maṣāḥif al-ʾAmṣār Footnote 36 can be formulated and analysed through the methodologies of stemmatics, yielding convincing stemmata of the regional codices. Such a result would not at all be expected if the rasm variations found in the regional codices have their origin in oral traditions that predate these codices.

While these facts are highly suggestive in the direction of a written exemplar informing the reading traditions (and in that sense are, indeed, actual reading traditions), so far no direct evidence had been presented that proves that the reading traditions were directly influenced by the rasm that said readers were exposed to. This paper shows that the transmitters of ibn ʿĀmir based their reading of the name ʾIbrāhīm on the rasm as it was present in the codex available to them. If the rasm had هم it would be read as ʾIbrāhām, whereas if the rasm had هىم , it would be read as the commonly accepted form ʾIbrāhīm. The fact that this orthographic alternation is not a regional variant, but an orthographic idiosyncrasy attested in all regional codices proves that it cannot be taken as an example where the rasm has been changed to accommodate the reading tradition. Instead the reading tradition must have been changed to accommodate the rasm.


The data presented in this paper shows that the reading traditions of the Quran cannot be seen as the product of an exclusively oral transmission of the of the quranic text that developed independently from the standardised Uthmanic rasm. The canonisation of the reading traditions along the criterion that a tradition has to agree with the Uthmanic rasm at the time of ibn Muǧāhid did not function as a ‘filter’, in the sense that only reading traditions that by chance happened to agree with the rasm were qualified to be considered for canonisation. Instead, the presence of the official rasm warped and changed the reading traditions over time, so that they would have come closer and closer to the text their readers were using. The reading traditions were informed by the rasm and evolved from there. And the rasm of the regional codices was certainly not subordinate to the reading traditions as Dutton suggests.Footnote 37 In this case, the rasm was not changed to reflect the reading traditions. The reading traditions were changed to accommodate the rasm.

We should not conclude from this that the reading traditions are completely devoid of an oral component that predates the Uthmanic canonical text. There are many cases where the canonical readers converge upon a reading, while the bare orthography of the Uthmanic rasm is highly ambiguous, and a potentially more obvious reading is left in favour of an idiosyncratic, but precise, word. This is not what we would expect if there had not been an oral tradition associated with the text.

One such an example is found in Q. 36:49 ىحصموں ‘they are disputing’ which by all but one of the seven readers is read as an anomalous assimilated stem VIII form.Footnote 38 The different readings of this word as presented by ibn MuǧāhidFootnote 39 are as follows:

  • yaḫaṣṣimūna  Ibn KathīrFootnote 40

  • yaḫaṣṣimūna  ʾAbū ʿAmr

  • yaḫiṣṣimūna   ʿĀṣim, al-kisāʾī, ibn ʿĀmir

  • yaḫṣṣimūna   Nāfiʿ

  • yaḫṣimūna   Ḥamzah

While there are four different forms of this verb in the seven traditions, all but the one transmitted by Ḥamzah, have an anomalous verb form that does not otherwise occur in the Quran, and does not conform to the regular word formation of Classical Arabic.Footnote 41 Had the readers based themselves completely on the rasm without any accompanying oral tradition, it seems unlikely that we would encounter such broad agreement on the use of this unusual verb form. From the rasm, the reading yaxṣimūna ‘they are disputing’ is clearly more straightforward, but nevertheless is only found with one of the seven canonical readers.Footnote 42

Moreover, if it is indeed true that the Uthmanic standard text did not have any consonantal dots at all,Footnote 43 it seems highly doubtful that a sensible reading—and something that approaches a consensus on most of the words in the Quran among the different readers—could be arrived at based purely on the consonantal skeleton, without knowledge of what the text is supposed to say.

In the light of this, it seems that we must consider the reading traditions neither the result of a purely oral tradition unrestrained by the written text that belongs to it, nor a purely written text completely devoid of a priori knowledge what the holy text is supposed to have said. We may therefore consider the reading traditions or rather semi-oral traditions, not altogether different from the traditional Tiberian recitation of the Hebrew Bible, which likewise is certainly dependent on the consonantal skeleton, but nevertheless preserves features in the oral tradition that would not be recoverable if the only thing that survived of the text was the consonantal text.Footnote 44

It is hoped that this investigation encourages further research into the variant readings and their relationship to not only the orthography of the Cairo Edition, but also of the earlier quranic manuscripts, whose internal variation and interrelation have yet to be examined in much more detail.

Appendix: Identification of manuscript traditions

The table below is based on the differences in the rasm of the regional codices as described by CookFootnote 45 which in turn summarizes al-Dānī.Footnote 46

The first table gives an overview of the different rasm features. As can be seen, the features marked by S, are features unique to the Syrian codices; those marked with M are shared by the Syrian and Medinese codex; The one marked by B is unique to the Basran codex, and those marked with K are unique to the Kufan codex.

Table 7. Overview of the regional codex variants.

Not all of these features are equally good as diagnostics for identifying a Quran to belong to one regional codex tradition rather than the other. Especially the features S8, K2, K3 and K4, which rely on the plene spelling of qāla are likely not features original to regional codices as this word is, generally spelled defectively in early quranic manuscripts.

The table below gives a comparative overview of the different quranic manuscripts examined. The columns are examinations of the different quranic manuscripts examined. CPP = Codex Parisino-Petropolitanus; Or. 2165 = Or. 2165 and Arabe 328(e); W = Wetzstein II 1913; S = Samarkand Codex; K = Großer Korankodex; SM1a = Saray Medina 1a; M = Ma VI 165; Q47 = Qāf 47 and CA1 = Codex Amrensis 1. The bottom row gives my identification of each manuscript.

Table 8. Identification of the codices examined by their rasm variants.

As can be seen, occasionally these manuscripts display variants which are technically mutually exclusive, e.g. appearance of the non Syrian variant of (S10) and (S14) and (M2) in Wetzstein II 1913. Nevertheless this manuscript so overwhelmingly has the other variants that point towards the Syrian tradition, that the overall identification is unambiguous.

The CPP and Or. 2165 are completely Syrian, which in light of the fact that not a single other identification lack mutually exclusive identifications, is rather surprising.

The Samarkand Codex has one Syrian variant in (S6) but otherwise points to a Basran or Kufan variant. It has one feature that is exclusive to the Kufan codex, allowing us to tentatively identify it as a Kufan codex.

The Großer Korankodex has (K3) and (K4) in their defective spelling, which is not a strong identification in favour of the Kufan variant. (K5) and (K6) unambiguously point to the non-Kufan form. Save for (M2), (M4) and (M12) the features point to a Basran identification, and it seems fairly safe therefore to identify this manuscript as Basran.

For Saray Medina 1a the majority of the Syrian features point to a non-Syrian Codex, whereas the Medinan features point to a non-Kufan/non-Basran identification. As a result its identification as Medinan seems clear.

Ma VI 165 is too fragmentary to make a strong identification possible. It is a Codex that has the Basran variant for (B1) but the Syrian/Medinan variant for (M6) and (M7). This manuscript may therefore be identified as either Basran, Medinan or Syrian. It is difficult to decide which of the features carries more weight. Cook expresses unease with identifying (B1) as a genuine Basran feature, rather than a late feature. “As we lack evidence of early basran manuscripts that retain this section of the manuscript, it remains to be seen whether Cook's unease with this variant is justified.” This does not exclude the possibility that it is not a feature part of the Basran uthmanic codex, however.

The data of Qāf 47 is fragmentary. The data that we have points to an identification as Medinan.

The data of Codex Amrensis 1 is fragmentary, but what is there is consistent with an identification of the codex as Medinan.

2 All other cases are part of the sequence which, like , is spelled defectively word-internally, an orthographic practice ultimately derived from Aramaic orthography (Diem, ‘Schreibung der Vokale’, §37–41). There is one more case in الفهم ʾīlāfi-him (Q. 106:2), for a discussion of this form see Van Putten, ‘Hamzah’, p. 110.

3 Diem ‘Schreibung der Vokale’, §40; Nöldeke et al. History, p. 401 fn. 81; Jeffery Foreign Vocabulary, p. 44ff.; Puin has a different opinion. To him ابرهيم represents a form of ʾIbrāhām with a shift of ā to ē in the final syllable, spelling /ʾibrāhēm/ (Puin, ‘ortho-epic writing’, p. 167). This “imālah” explanation is unsatisfactory. There is no obvious conditioning factor that would have shifted this final ā vowel to ē, and words with ā in similar environments never undergo such a shift. The name همن hāmān or the noun سموت samāwāt ‘heavens’, for example, show no such shift. Invoking an unmotivated “imālah” therefore introduces more problems than it solves.

4 Déroche, La transmission écrite.

5 Ibn Muǧāhid, Kitāb al-Sabʿah, p. 169.

6 al-Dānī, Taysīr, p. 76f.

7 Ibn al-Ǧazarī, Našr, vol 4., pp. 2184–2187.

8 Surprisingly, al-Dānī makes no mention of the codices as described by ibn al-Ǧazarī in his Muqniʿ fī Rasm Maṣāḥif al-ʾAmṣār (al-Dānī, Muqniʿ, p. 96).

10 Those that do not match ibn al-Ǧazarī’s appear in two types. Those that have ابرهيم in all places, i.e. the upper text of the Sanaa Palimpsest (DAM 01-27.1), and the clearly later Rampur Raza Library: No. 1, Korankodex. And those that have ابرهم in Q. 2 and ابرهيم elsewhere, these will be discussed later in the article.

11 Dutton, ‘Oldest Qur'an Manuscript’, p. 45.

12 Déroche, La transmission écrite, p. 60.

13 Dutton, ‘Oldest Qur'an Manuscript’, p. 45.

14 Déroche, Qurʾans of the Umayyads, p. 39, fn. 5.

15 This manuscript has been accessed through the website of the British Library, rather than through the Corpus Coranicum website which does not have images of this manuscript. URL:

16 Dutton, ‘Oldest Qur'an Manuscript’, p. 45.

17 Cellard, Codex Amrensis 1.

18 al-Dānī, Muqniʿ, p. 96.

19 Minutoli 296 technically agrees with this pattern as well. All attestations in Sūrat al-Baqarah are spelled ابرهم , and the single attestation outside of it, Q. 14:35 spells it ابرهيم . This manuscript is too fragmentary to decide whether this distribution is meaningful (p = 0.0624).

20 This document is perhaps better identified as Kufic B Ia, as Déroche does (Déroche, Catalogue, p. 67). This style is quite close to the Hijazi styles (op. cit.,p. 37).

21 Dutton, ‘An Early Muṣḥaf’.

22 Dutton, ‘Oldest Qur'an Manuscript’, p. 45.

23 However, see Rabb ‘Non-Canonical Readings’. Rabb identifies this manuscript as belonging to a non-canonical Ḥimsi reading.

24 Cook, ‘The Stemma of the Regional Codices’.

25 Cook, ‘The Stemma of the Regional Codices’.

26 An examination of the features attributed to the Meccan codices indeed confirm that Cook is right to feel uneasy in accepting the Meccan codex as belonging to one of the original copies of the Uthmanic Archetype.

27 Ibn al-Ǧazarī, Našr, vol. 4, p. 2185f.

28 Dutton, ‘An Early Muṣḥaf’; Dutton, ‘Oldest Qur'an Manuscript’.

29 Dutton, ‘An Early Muṣḥaf’, p. 74; Dutton, ‘Oldest Qur'an Manuscript’, p. 48.

30 Dutton, ‘An Early Muṣḥaf’, p. 82f.

31 Dutton 2001: 74.

32 The ideology that a reading must agree with the rasm certainly existed by ibn Muǧāhid's time (d. 324/936) (Nasser, Transmission, p. 52). Nasser argues in his book that tawātur was not yet as important to ibn Muǧāhid. But ibn Muǧāhid's contemporary, Muḥamad Ḥabaš says that the rasm/ʿarabiyyah/tawātur criterion was present as early as the 3rd/9th century (Nasser, Transmission, p. 48).

33 For a similar interaction between oral tradition and agreement with the rasm see Sadeghi, ‘Criteria’, pp. 24–26.

34 It is clear that to ibn Muǧāhid agreement with the rasm was an important criterion for choosing a canonical reading. Nasser (Nasser, Transmission, pp. 56–59) convincingly argues that the absence of a majority reading in Kufa that agreed with the rasm was the reason for ibn Muǧāhid to choose three relatively minor Kufan traditions rather than the dominant non-Uthmanic reading of ibn Masʿūd.

35 Cook, ‘The Stemma of the Regional Codices’.

36 Al-Dānī, Muqniʿ.

37 Dutton, ‘An Early Muṣḥaf’; Dutton, ‘Oldest Qur'an Manuscript’.

38 For the identification of this verb as an assimilated stem VIII form, see for example al-Farrāʾ Maʿānī al-Qurʾān, vol. 2, p. 379 who moreover relates that ʾUbayy b, Kaʿb read this form as yaḫtaṣimūna.

39 Ibn Muǧāhid, Kitāb al-Sabʿah, p. 541. For a similar case, compare Q. 10:35 يهدى read variously as yahddī, yahaddī, yahiddī and yahăddī, ibn Muǧāhid, Kitāb al-Sabʿah, p. 326.

40 And Nāfiʿ according to Warš.

41 It is forms like these that place excruciating focus on the deep circularity of, e.g. al-Ṭabarī’s requirement of the reading traditions to be in accordance of the grammar of the ʿArabiyyah (Nasser, Transmission, p. 46). This form would not have been likely to have been accepted as grammatical in the ʿArabiyyah had it not been for the fact that it occurs in the Quran.

42 It seems to me that these forms should all be derived from single quranic Arabic form close to that of the tradition of Nāfiʿ, e.g. /yaxṣṣimūn/. This cluster is disallowed in the Classical Arabic phonotactics, and forms like yaxaṣṣimūna and yaxiṣṣimūna can be analyzed as different solutions to alleviate this problematic cluster. The form yaxṣṣimūna appears to be the result of the syncope of *a with subsequent assimilation of the *tṣ cluster to ṣṣ, i.e. *yaxtaṣimūna > *yaxtṣimūna > yaxṣṣimūna. The motivation for this unexpected syncope, however, is unclear.

43 This traditional view, which very often taken as fact by modern scholars of the Quran, should be doubted in light of manuscript evidence. While it is true that Kufic texts generally lack the dots to distinguish consonants in the rasm, this certainly is not true for Qurans in the earlier Hijazi script. Many, if not all, early quranic manuscripts written in this script have some amount of consonantal dotting. Combining this with the knowledge that the consonantal dotting certainly existed during Uthman's reign – as we find it in an inscription from 24/646 (Ghabban & Hoyland, ‘The inscription of Zuhayr’) and a papyrus from 22/644 (Grohmann, ‘Aperçu’) – it seems premature to accept the traditional narrative that the Uthmanic codex had no consonantal dotting at all.

44 As Khan puts it: “[…] [T]he reading was a separate layer of tradition that was closely related to, but nevertheless independent from the tradition of the consonantal text.” (Khan, A Short Introduction, 46f.). This should not be understood to mean that the Tiberian reading tradition was not in some places subordinate to the Masoretic consonantal text, as Khan shows that there are several readings clearly influenced by what the consonantal skeleton allows (op. cit., p. 49).

45 Cook ‘The Stemma of the Regional Codices’.

46 Al-Dānī, Muqniʿ.

47 The yāʾ has been removed here, but the scratch marks confirm that originally it contained the Syrian variant.

48 The text has the Syrian variant, but it was clearly added by a later hand. It is difficult to judge what the form of the original text was.

49 The هو was originally present, first removed to comply with SM and then added again by a much later hand.

50 The original text contains هو but this has been removed.

51 An ʾalif of the Basran text type has been added with red ink.

52 As defective قل can easily stand for qāla, the fact that (K3) and (K4) agree with the Kufan spelling is not significant.


Arabic sources

al-Dānī, , Kitāb al-Taysīr fī al-Qirāʾāt al-Sabʿ, ed. Pretzl, O.. (Beirut: Dār al-Kitāb al-ʿArabī, 1984).
al-Dānī, , Muqniʿ fī Rasm Maṣāḥif al-ʾAmṣār, ed. Qamḥāwī, M.A.. (Cairo: Maktabat al-Kulliyāt al-Azhariyyah, n.d.).
al-Farrāʾ, , Maʿānī al-Qurʾān, ed. Šams al-Dīn, I.. (3 vols. Beirut: Dār al-Kutub al-ʿIlmiyyah, 2016).
Ibn Muǧāhid, , Kitāb al-Sabʿah fī al-Qirāʾāt, ed. Ḍayf, Š.. (Cairo: Dār al-Maʿārif, n.d.).
Ibn al-Ǧazarī, , Našr l-Qirāʾāt al-ʿAšr, ed. ʾRušdī Suwayd, Ayman. (5 vols. Beirut & Istanbul: Dār al-Ġawṯānī. 2018).

Other sources

Cellard, E., Codex Amrensis 1, avec le concours de Sabrina Cimiotti (Leiden, 2017).
Cook, M., ‘The Stemma of the Regional Codices of the Koran’, Graeco-Arabica, 9/10 (2004), pp. 89104
Déroche, F., Les Manuscrits du Coran: Aux origines de la calligraphie coranique (Paris, 1983).
Déroche, F., La transmission écrite du Coran dans les débuts de l'islam: Le codex Parisino-petropolitanus (Leiden & Boston2009).
Déroche, F., Qurʾans of the Umayyads: A first Overview (Leiden & Boston, 2014).
Diem, W., ‘Untersuchungen zur frühen Geschichte der arabischen Orthographie. I. Die Schreibung der Vokale’, Orientalia N.S., 48 (1979), pp. 207257.
Dutton, Y., ‘An Early Muṣḥaf According to the Reading of Ibn ʿĀmir’, Journal of Qurʾanic Studies, 3:1 (2001), pp. 7189.
Dutton, Y.Some Notes on the British Library's ‘Oldest Qur'an Manuscript’ (Or. 2165)’, Journal of Qurʾanic Studies, 6:1 (2004), pp. 4371.
Ghabban, A. & R. Hoyland ‘The inscription of Zuhayr, the oldest Islamic inscription (24AH/AD 644-645), the reist of the Arabic script and the nature of the early Islamic state’, Arabian Archaeology and Epigraphy, 19 (2008), pp. 209236.
Grohmann, A.Aperçu De Papyrologie Arabe’, Études De Papyrologie, 1 (1932), pp. 4043.
Jeffery, A., The Foreign Vocabulary of the Qurʾān (Leiden & Boston, 2007).
Khan, G., A Short Introduction to the Tiberian Masoretic Bible and its Reading Tradition. Second edition (Piscataway, NJ: Georgias Press, 2013).
Nasser, S. H., The Transmission of the Variant Readings of the Qurʾān (Leiden & Boston, 2013).
Nöldeke, T., Schwally, F., Bergsträßer, G. & Pretzl, O., The History of the Qurʾān, edited and translated by Behn, W. H.. (Leiden & Boston, 2013).
Puin, G.-R., ‘Vowel letters and ortho-epic writing in the Qurʾān, in New Perspective on the Qurʾān. The Qurʾān in its historical context 2, ed. Reynolds, G. S. (London & New York, 2011), pp. 147190.
van Putten, M.The Hamzah in the Quranic Consonantal Text’, Orientalia, 87:1 (2018), pp. 93120.
Rabb, I. A., ‘Non-Canonical Readings of the Qurʾan: Recognition and Authenticity (The Ḥimṣī Reading)’, Journal of Qurʾanic Studies, 8:2 (2006), pp. 84127.
Sadeghi, B., ‘Criteria for Emending the Text of the Qurʾān’, in Law and Tradition in Classical Islamic Thought, eds. Cook, M., Haider, N, Rabb, I., & Sayeed, A. (New York, 2013).