Pronominal variation in Arabic among the grammarians, Qurʾānic reading traditions and manuscripts

ABSTRACT

In the early Islamic period, the literary form of Arabic had an astounding amount of variation in its pronominal morphology. This paper examines the pronominal morphology of especially the third person masculine singular and plural pronouns, as they are described by the early Arabic grammarians, how they appear in the the descriptions of the canonical and non-canonical reading traditions and finally the pronominal systems as they appear in actual early Qurʾānic manuscripts. It is shown that the early grammarians were much more permissive of morphological variation than one would expect from strict norms that appear later. It is also shown that the pronominal systems of the reading traditions go well beyond the kinds of variation described by these grammarians. The systems of these readers show clear clusters (a Kufan, Basran, Hijazi and a ‘Classical’ cluster), even though in the details reading traditions from a single region still often differ from one another. Finally, it is shown that while some of the pronominal systems known in the reading tradition literature also show up in the Qurʾānic manuscripts, there is a significant number of manuscripts that have pronominal systems unlike any of the described systems in the literary sources.

KEYWORDS:

• Qurʾān
• Qirāʾāt
• Classical arabic
• Qurʾānic manuscripts
• Arabic grammarians
• prononominal morphology

Introduction

Classical Arabic as it is presented in its normative form has a fairly uniform system for third-person suffix pronouns. Both -hu and -hum harmonise to -hi and -him if i, ī or ay precedes, and the singular -hu/hi is lengthened after short vowels to -hū/-hī (Wright 1996 [1859-62]: 100–1; Fischer 2002 [1972]: 142-3; Thackston 1994: 39-40). This system is likewise similar to what we encounter in the most widely circulated Qurʾānic print edition, the Medina muṣḥaf that reflects the reading of Ḥafṣ ʿan (on the authority of) ʿĀṣim.

However, even within Ḥafṣ’ reading we know that this is not the entire story. Ḥafṣ on one occasion lengthens the pronoun even after a long vowel, Q. 25:69 fī-hī (therein), and does not apply vowel harmony to -hu in two other occasions Q. 18:63 ʾansā-nī-huʾillā ([and none] made me forget except [Satan]) and Q. 48:10 ʿalay-hu ḷḷāh ([what] he has covenanted with God). Further examination of another popular reading, namely that of Warš ʿan Nāfiʿ, reveals yet more variation, where such forms like ʿalay-himū and humū appear. It is clear then that the simple and uniform picture of normative Classical Arabic does not properly cover the linguistic situation of the Qurʾān in the early Islamic period. The Qurʾānic reading traditions present a much more diverse situation than the strict scholastic grammatical descriptions of the 19^th- and 20^th-century orientalists have often let on; this is likewise the case for the pronominal system present in early Qurʾānic manuscripts, as already noted by Nöldeke (1860: 328).¹ Likewise, the early Arabic grammarians report many more variants of the pronominal morphology than these modern descriptions.

As it stands there is therefore a clear lacuna in terms of the description of the pronominal system of early Classical Arabic as it is reflected in the Qurʾān and described by the Arabic grammarians. The current paper aims to develop a clear and comprehensive account of the behaviour of suffixed (third-person) pronouns as they are available to us in three different sources. First, we examine how the behaviour of these pronominal systems is described by the Arab grammarians. Second, we conduct a detailed study into the many Qurʾānic reading traditions as they are recorded in the qirāʾāt (Reading traditions) literature. Finally, we turn to Qurʾānic manuscripts and, with a large comprehensive corpus of early vocalised manuscripts, we examine the pronominal systems as they are reflected in these manuscripts.

Grammarians

While Classical Arabic is often thought of as a fairly uniform register, especially when it comes to the morphophonology, we often find that the grammarians themselves were rather lenient in allowing linguistic variation. Moreover, close examination will show that the reports and judgement of acceptability differ both by region and through time. The following section will present a fairly comprehensive treatment of the pronominal suffix system of -hū and -hum among the prominent grammarians of the first millennium CE whose works have survived, tracking the variation they report and how their reports differ from one another. We will also discuss a few later grammarians whose comments are of some interest to the topic at hand.

Sībawayhi (d. ca. 180/796)

Sībawayhi devotes two consecutive chapters to the shape of the third-person singular and plural pronouns, the first being ‘the chapter of the persistence of the yāʾ and wāw to the hāʾ that functions as a sign of the pronoun, and their removal from it’ (bāb ṯabāt al-yāʾ wa-l-wāw fī l-hāʾ allatī hiya ʿalāmat al-ʾiḍmār wa-ḥaḏfi-himā, Sībawayhi Kitāb IV, 189–95) and ‘the chapter of what is given a kasra concerning the hāʾ which is the sign of the pronoun’ (bāb mā tuksaru fī-hi l-hāʾ allatī hiya ʿalāmat al-ʾiḍmār, Sībawayhi Kitāb IV, 195–8).

He takes the lengthened forms of the third person singular masculine, i.e. -hū and -hī as the default form, and describes in which contexts these are shortened. Sībawayhi tells us this happens after long vowels and diphthongs, thus one says ʿalay-hi yā fatā (upon it, O boy), raʾaytu ʾabā-hu qablu (I saw his father before), ʾabū-hu kamā tarā (his father is as you see, Sībawayhi Kitāb IV, 195–8).

It is also possible to retain the long vowels -hū and -hī regardless of context, although he comments that the short pronunciation is the better of the two readings (of the Qurʾān) (ʾaḥsan al-qirāʾatayn)² citing Q. 17:106 nazzalnā-hu tanzīlan (we have revealed it as revelation), Q. 7:176 ʾin taḥmil ʿalay-hi yalhaṯ (if you chase him, he will pant), Q. 12:20 wa-šaraw-hu bi-ṯamanin baḫsin (they sold him for a low price) and Q. 69:30 ḫuḏū-hu fa-ġullū-hu (seize him, and shackle him). But, also the long pronunciation (i.e. nazzalnā-hū, ʿalay-hī, šaraw-hū, ḫuḏū-hū) is Arabic (wa-l-ʾitmām ʿarabī, Sībawayhi Kitāb IV, 189).

He explicitly tells us that if the preceding letter does not carry a vowel, the vowel is normally lengthened (e.g. min-hū, ʾaṣābat-hū etc). This is different from the form of Classical Arabic that later become normative,³ where after any closed syllable the suffix is shortened. Sībawayhi admits this possibility too, he tells us that ‘the speech of some of them’ (qawl baʿḍi-him) is min-hu yā fatā (from them, O boy), ʾaṣābat-hu ǧāʾiḥatun (a calamity struck him), but pronouncing it fully is better (al-ʾitmām ʾaǧwad, Sībawayhi Kitāb IV, 190).

In the same chapters Sībawayhi also discusses the long form of the plural pronouns, in this case not just the third-person suffix -hum/him, but all pronominal elements that end in the mīm, i.e. also the independent third person plural hum, second person plural suffix -kum, the verbal second-person plural -tum and the independent second-person plural ʾantum. With all of these the speakers have a choice: ‘If you wish you can remove [the final -ū or -ī], or you can maintain it’. Thus, Sībawayhi gives the examples ʿalay-kumū mālun (wealth is upon you), ʾantumū ḏāhibūna (you are going), and laday-himī mālun (they have wealth). But one may just as well say ʿalay-kum mālun, ʾantum ḏāhibūna and laday-him mālun. Interesting to note here is the harmonised long form -himī, which tends to not be considered part of normative Classical Arabic, where preference is given to the form -himū. However, Sībawayhi is quite explicit that only ‘some of them say ʿalay-himū’ (Sībawayhi Kitāb IV, 191).

Even when people normally drop the long vowel, before a two-consonant cluster the u that belongs to the long vowel -ū reappears, thus kun-tumu l-yawma (today, you were), faʿal-tumu l-ḫayra (you have done good), ʿalay-himu l-mālu (upon them is wealth). Those that say ʿalay-himi in this context have as their basic form ʿalay-himī (which may presumably be shortened to ʿalay-him) (Sībawayhi Kitāb IV, 191).

The second of the two chapters specifically addresses not the length of the suffix, but the vowel of the suffix. If a kasra or yāʾ precedes the hāʾ (of either -hū/ī or -humū/himī), then the hāʾ receives a i vowel, whether it is short or long, thus marartu bi-hī qablu (I passed by him before), laday-hi mālun (he has wealth), marartu bi-dāri-hī qablu (I passed by his house before). However, the people of the Hijaz say marartu bi-hū qablu, and laday-hū mālun and they would recite Q. 28:81 fa-ḫasafnā bi-hū wa-bi-dāri-hū l-ʾarḍa (and we caused him and his home to sink into the ground). In other words, the dialect of the Hijaz is explicitly said to have not undergone vowel harmony. Unlike other variations in reading the Qurʾān discussed earlier by Sībawayhi, this variation is not explicitly called better or worse than the general system he describes (Sībawayhi Kitāb IV, 195).

He continues, saying that whoever says wa-bi-dāri-hū l-ʾarḍa [i.e. the Hijazis] also says ʿalay-humū mālun and bi-humū ḏālika – thus connecting the long unharmonised form with the unharmonised -hū of the singular (Sībawayhi Kitāb IV, 196).⁴

While the unharmonised Hijazi forms receive no special qualification in relation to the harmonised forms, this is different from certain harmonised forms that fall outside of Sībawayhi’s normative acceptability. Thus, he informs us that people from among the Rabīʿa say min-him ‘from them’. Sībawayhi does not mince words: this is degenerate language (wa-hāḏihī luġa radīʾa). Among the Bakr b. Wāʾil tribe, some speakers go even further and harmonise the vowel of -kum saying ʾaḥlāmi-kim (your dreams) and bi-kim (with you). With this Sībawayhi is even less pleased and says ‘it is utterly degenerate’ (hiya radīʾa ǧiddan, Sībawayhi Kitāb IV, 196). Schematically we can represent the different options Sībawayhi describes as proper Arabic. A small arrow pointing downwards is placed besides any form that Sībawayhi values negatively compared to the other forms.

Table

	Default system	“Some of them”	Hijaz
v-hv	-hū, -hī		-hū
v-hv	-hu, -hi	-hū, -hī ↓	-hu
C-hv	-hū	-hu↓	-hū
-hvm	-hum(ū), -him(ī)	-hum(ū), -him(ū)	-humū
-hvmv CC	-humu, -himi	-humu, -himu	-humu

Farrāʾ (d. 207/822)

Farrāʾ in his Luġāt al-Qurʾān reports on the various ways the third person singular masculine, dual and plural pronouns are treated in terms of vowel harmony. He states, ‘Regarding ʿalay-hvm: there are two linguistic practices. As for Qurayš and the people of Hijaz, and those eloquent [people] that surround them [who are from] Yemen, they say: ʿalay-hum, with an u after the h (rafʿ al-hāʾ), and (likewise) ʿalay-humā, and ʿalay-hunna, and Q. 15:6 yā-ʾayyuhā llaḏī nuzzila ʾalay-hu ḏ-ḏikru (O you to whom the reminder has been revealed), and Q. 2:2 lā rayba fī-hu hudan li-l-muttaqīna (there is no doubt in it, a guidance to the God-fearing), and Q. 26:210 wa-mā tanazzalat bi-hu š-šayāṭīnu (the devils did not reveal it), and nazaltu bi-hu (I lodged with him), so they [always] have an u after the h (rafʿ al-hāʾ)’ (Farrāʾ Luġāt 10).

He continues, ‘And the people of Naǧd from ʾAsad and Qays and Tamīm, they apply an i vowel to it (yaksirūnahā), so they say: ʿalay-hi, and ʿalay-himā, and ʿalay-him. As for Kināna and some of Banū Saʿd b. Bakr – and they are the foster family of the prophet – they also apply an i vowel to it, but when it is followed by a [definite article] they apply the u to the h and m (of the third person plural masculine), like Q. 6:111 ʾilay-humu l-malāʾikata ([we bestowed] the angels upon them), and Q. 28:63 ʿalay-humu l-qawlu ([those] upon whom the word [has come into effect]). This is what Kisāʾī chooses⁵ and it is, in our eyes, the most eloquent of linguistic practices because the prophet said “I am the most eloquent of you [all], I grew upon among my maternal uncles”’ (Farrāʾ Luġāt 10–1).

Finally, he concludes ‘Some of Banū ʾAsad apply an i to the h in ʿalay-him and an u to the m when [it is followed by] the definite article al-, so they say: “ʿalay-himu l-malāʾikata”. All of the above is correct and good’. What Farrāʾ does not mention explicitly is the alternative, ʿalay-himi l-malāʾikata, although it is implied by singling out some of the Banū ʾAsad (Farrāʾ Luġāt 11).

When it comes to the vowel of the plural pronouns such as ʾantumū, -kumū and -humū, Farrāʾ tells us that ‘the Arabs are united on [both options:] not placing a vowel after the m, or placing an u after it. In their speech it is: min-hum or min-humū; ʿalay-kum or ʿalay-kumu; kun-tum or kun-tumu.⁶ We do not know it as being exclusive to anyone (to have) one of the two linguistic practices. All of them say it in both manners’ (Farrāʾ Luġāt 33).

Finally, Farrāʾ also mentions the forms with extreme vowel harmony such as min-him, which he attributes to Kalb, ʿUḏra, Banū l-Qayn, Banū Taġlab, and Namr which is a linguistic practice that he considers to be rejected (wa-hiya luġa marfūḍa), and he adds that Namr even says as-salāmuʿalay-kim (Farrāʾ Luġāt 23).

While in Luġāt al-Qurʾān Farrāʾ does not discuss the conditioning of the shortening of the singular pronoun to -hu/hi explicitly, he does discuss this at some length in his Maʿānī l-Qurʾān. Here he says that ‘they choose to remove the wāw from the hāʾ’ thus saying daʿ-hu yaḏhabu (let him go), min-hu, ʿan-hu, and they hardly ever say [wa-lā yakādūna yaqūlūna] min-hū, ʿan-hū. Here Farrāʾ therefore clearly disagrees with Sībawayhi, and advocates the form that becomes standard in Classical Arabic today (Farrāʾ Maʿānī I, 223–4).

ʾAḫfas al-ʾAwsaṭ (d. 215/830)

In his Maʿānī l-Qurʾān, ʾAḫfaš al-ʾAwsaṭ (henceforth ʾAḫfaš) begins by discussing suffix vowel length. For singular pronouns following a yāʾ, according to him, the Arabs shortened the yāʾ and the wāw that comes after the hāʾ, which is consistent with Sībawayhi. The same applies to ʾalif-s and wāw-s before the pronoun like Q. 7:107 ʿaṣā-hu (his stick) and Q. 10:73 fa-kaḏḏabū-hu (so they disbelieved him). He does recognise that ‘some Arabs maintain [the length], because that is from the base (al-aṣl), so they say: fa-kaḏḏabū-hū, or Q. 7:64 fa-ʾanǧaynā-hū (so we saved him), or ʿaṣā-hū, or Q. 2:2 fī-hū (in it), and this is the reading of the people of Medina’ (ʾAḫfaš Maʿānī 26–7).⁷ Note also that here he suggests that the people of Medina read fī-hū without vowel harmony.

For singular suffixes following consonants, however, ʾAḫfaš states that ‘some people have said: Q. 51:50 ʾinnī lakum min-hu naḏīrun mubīnun (I am for you, from him, a clear warner), dropping the wāw [of min-hū]. They likened the consonant to the yāʾ, wāw, and ʾalif. But this is not good in Arabic. The best is min-hū naḏīrun with the addition of a wāw, even if it is not written’.⁸ Here, he is in clear agreement with Sībawayhi and more opinionated at that, against Farrāʾ, his Kufan contemporary (ʾAḫfaš Maʿānī 27).

Vowel harmony after ī, ay or i is referred to as the practice of Banū Tamīm. ‘As for the people of Hijaz, they do ḍamm after i and after yāʾ as well. [God] says [Q. 2:92]: ṯumma ttaḫaḏtumu l-ʿiǧla min baʿdi-hī wa-ʾantum ẓālimūna (then you took the calf after it). The people of Hijaz [read]: min baʿdi-hū, maintaining the wāw in all locations’ (ʾAḫfaš Maʿānī 28).

ʾAḫfaš disapproves of shortened suffixes after short vowels as well: ‘Of the Arabs are those who drop the wāw or yāʾ in these instances as well, but this is rare and obscene. They say: marartu bi-hi qablu (I passed by him before). They do the same with ḍamm, saying: raʾaytu-hu qablu (I saw him before), not following with a wāw. We have also heard some of this from eloquent Arabs’. He ends his discussion of the singular pronoun by mentioning that some readers assimilate across the short vowel Q. 2:2 fī-h=hudan ([there is no doubt] in it, a guidance [for the mindful]),⁹ and other Arabs do not vowel the pronominal hāʾ at all, which is claimed to be attributed to ʾAzd al-Sarāt¹⁰ (ʾA ḫfaš Maʿānī 28).

With respect to the realisation of the plural pronoun, the speakers are split into two groups: those who do not add any vowel to the -hvm suffix in context and those who do. For the latter group, ʾAḫfaš mentions some who add a wāw to form -hvmū unconditionally, both in Qurʾānic reading and in regular speech.¹¹ If it is preceded by an i, ī or ay, such as in ʾalay-him, bi-him, or baʿdi-him, then ‘some Arabs say ʿalay-himī, with the addition of a yāʾ, and the kasr of the mīm and hāʾ’. Alternatively, he mentions those who say ʿalay-himū, or ʿalay-humī. Forms such as ʿalay-kimī and bi-kimī, he refers to as obnoxious (qabīḥ), practically unheard of, and attributed to some of the Bakr b. Wāʾil; he cites ʾAḫfaš al-ʾAkbar (d. 177/793) as having heard this in some of their poetry (ʾAḫfaš Maʿānī 29).

Regarding suffixes followed by consonant clusters, ʾAḫfaš prescribes shortening the long vowel, both for plural and singular, yielding marartu bi-hi l-yawma (I passed by him today) and raʾaytu-hu l-yawma (I saw him today). He concludes by dismissing the shortening of the vowels of plural pronouns like -humū, stating that ‘it has been claimed that some Arabs vowel the mīm but do not follow it with a yāʾ or a wāw in poetry. But this is practically unheard of’ (ʾAḫfaš Maʿānī 30).

Mubarrid (d. 285/899)

For the singular, Mubarrid in his Muqtaḍab starts by stating that the base (ʾaṣl) of the pronoun is always -hū and in pause -h, ‘so if you pronounce it in context, you say: ʾaʿṭaytu-hū yā raǧula (I gave it to him, O man!), ǧāʾa-nī ġulāmu-hū fa-ʿlam (his servant came to me, so know it!), raʾaytu ġulāma-hū yā fatā (I saw his servant, O boy!), marartu bi-ġulāmi-hū (I passed by his servant), marartu bi-hū (I passed by him), Q. 28:81 fa-ḫasafnā bi-hū wa-bi-dāri-hū l-ʾarḍa (and we caused him and his home to sink into the ground), ʿalay-hū mālun (he has wealth), hāḏihī ʿaṣā-hū yā fatā (this is his stick, O boy), and hāḏā ʾaḫū-hū fa-ʿlam (this is his brother, so know it!, Mubarrid Muqtaḍab I, 399)’.

Different from the other grammarians, Mubarrid specifically starts his description with the ‘base’ forms, which Sībawayhi and others only describe as the Hijazi forms. He continues, however, saying that ‘if there is a yāʾ or a kasra before the hāʾ, then it is best (al-ʾaḥsan) that you replace the ḍamma with a kasra [...] and the wāw with the yāʾ. But even if you bring the base form, like we started with, that is [also] good Arabic (fa-ʿarabī ǧayyid, Mubarrid Muqtaḍab I, 399)’.¹² Here then, he gives explicit normative preference to harmonised forms, something that is lacking in the descriptions of the grammarians before him.

Moreover, for the shortening he says ‘know that when there is an unvoweled yāʾ or unvoweled wāw or ʾalif before the masculine hāʾ, there is one who chooses to remove the wāw and the yāʾ after it […] and thus his speech is Q. 7:107 fa-ʾalqā mūsā ʿaṣā-hu (so Moses cast down his stick), Q. 24:54 ʿalay-hi mā ḥummila (and upon him is that with which he has been charged), [and they say:] fī-hi baṣāʾiru (in it are keen insights), raʾaytu qafā-hu yā fatā (I saw the back of his neck, O boy!) If you pronounce it fully, that is excellent Arabic (fa-ʿarabī ḥasan), for it is the base (ʾaṣl) and it is an option’ (Mubarrid Muqtaḍab I, 401).

He continues to comment on the preferences of his Basran predecessors and explicitly disagrees with them, siding with Farrāʾ: ‘in the case that before the hāʾ there is an unvoweled letter that is not among these letters (i.e. ʾalif, yāʾ or wāw) then Sībawayhi and Ḫalīl preferred full pronunciation. But to me removal is the best (al-ḥaḏf ʿindī ʾaḥsan), thus His speech is: Q. 3:7 min-hu ʾāyātun muḥkamātun (among it are clear revelations), [and they say:] min ladun-hu yā fatā (from his part, O boy!, Mubarrid Muqtaḍab I, 401)’.

In the sections where Mubarrid discusses the morphology of the plural, he first introduces the second person plural forms, which he says can be either -kumū, -tumū or shortened to -kum, -tum. For the third person masculine plural, one switches the kāf with the hāʾ and ‘you say: raʾaytu-humū yā fatā (I saw them, O boy!), marartu bi-humū fa-ʿlam (I passed by them, so know it!), and it is allowed to remove (the wāw), and this is good and what most people do […]. However, it is preferable with the hāʾ to apply a kasra (to it) when before it there is a kasra or a yāʾ, so you say: marartu bi-himī (I passed by them), nazaltu ʿalay-himī (I visited them). One who removes (the wāw or yāʾ) says: marartu bi-him, nazaltu ʿalay-him’. One may also decide not to apply the vowel harmony on the vowel that follows the mīm, and ‘so one says marartu bi-himū, but following [the vocalisation of the preceding syllable] is better (ʾaḥsan), and thus he says: marartu bi-himī, nazaltu ʿalay-himī’ (Mubarrid Muqtaḍab I, 403–5).

Similar to Sībawayhi, Mubarrid finishes up mentioning that Bakr b. Wāʾil would apply vowel harmony even with -kum. But this is a mistake to him as well.

Mubarrid’s description is significantly more normative than the previous grammarians, and he gives specific instructions on how to treat the pronominal system. His most normative system is essentially identical to what comes to be considered normative Classical Arabic, with the exception that he prefers the harmonised long form -himī over -himū.

Zaǧǧāǧ (d. 311/923)

Zaǧǧāǧ (Maʿānī I, 50–3) discusses the pronouns starting with the plural pronoun, stating: ‘and you have, regarding ʿalay-hvm, [two options]: adding an u to the h or adding an i. Thus, you say: Q. 1:7 allaḏīna ʾanʿamta ʿalay-him ([the path] of those you have blessed) or ʿalay-hum. And the majority of the readers adhere to these two forms (luġatayn)’. He also discusses the optional ending, ‘it is also permitted [to read] ʿalay-h[i/u]mū with a waw. The base (ʾaṣl) for this is the hāʾ in your speech: ḍarabtu-hū yā fatā (I hit him, O boy) or marartu bi-hū yā fatā (I passed by him, O boy!) that is articulated with a wāw in context. However, when you pause, you say: ḍarabtu-h or marartu bi-h’ (Zaǧǧāǧ Maʿānī I, 50).

In his description, Zaǧǧāǧ takes the unharmonised form of the singular pronoun to be the default while allowing the harmonised suffix as well: ‘when you say marartu bi-hū yā fatā, if you so choose you can also say marartu bi-hī. You change the wāw into a yāʾ due to the i that precedes it’. Citing Qurʾānic verses, he states ‘it was also read Q. 28:81 fa-ḫasafnā bi-hī wa-bi-dāri-hī l-ʾarḍa or bi-hū wa-bi-dāri-hū l-ʾarḍa (and we caused him and his home to sink into the ground) – which is from the reading of the people of Hijaz’. He follows up with a summary of the possible combinations: ‘Therefore, you have four options: if you wish, you add an i to the hāʾ or if you wish you maintain [the original] ī. It is the same in ḍamm: if you wish, you add an u to the hāʾ, or if you wish you maintain [the original] ū. So, you can say: ʾalay-hi, ʿalay-hī, ʾalay-hu,ʿalay-hū’ (Zaǧǧāǧ Maʿānī I, 50–1).

Although Zaǧǧāǧ contextualises the unharmonised pronominal suffix as being the base (ʾaṣl), that status does not translate into preference in Qurʾānic readings. In fact, while previous grammarians are descriptive in their treatment of the pronominal suffix forms, with Farrāʾ explicitly approving all forms he lists as being ‘correct and good’, Zaǧǧāǧ expresses explicit preference for particular forms, and distinguishes between possible forms and permitted forms in Qurʾānic reading on the basis of pre-existing tradition, a condition which was not previously stipulated. He says, ‘The [proper] reading is i without a full ī in ʿalay-hi; it is the best of these four options. It is also not allowed (lā yanbaġī) to read with what is possible (bimā yaǧūz) except if there exists an authentic transmission using it or it is read by many readers’ (Zaǧǧāǧ Maʿānī I, 51)¹³

He then explicitly ranks the singular pronominal suffixes, preferring shortening in closed syllables and harmony over lengthening and disharmony, ‘The most superior of languages is what is in the Qurʾān, which is His speech: Q. 3:75 ʿalay-hi qāʾiman (standing over him).¹⁴ Then what is next in superiority is ʿalay-hu mālun, with ḍamm. Then it is followed by ʿalay-hī mālun, and finally ʿalay-hū mālun – and it is the worst (ʾardaʾ) of the four’ (Zaǧǧāǧ Maʿānī I, 51). Zaǧǧāǧ’s expression of preference for the shortened harmonised form is reflected in the canonical reading traditions which mostly follow this norm as we will see below, and is likely to indicate a more developed notion of proper Arabic present in the late 3^rd/early 4^th century compared to just a century prior. We also find, once again, explicit disapproval of the lengthening of all singular pronominal suffixes after long vowels and diphthongs, which Sībawayhi expressed distaste for as well.

Moving on to the plural pronominal suffix, he says: ‘As for their speech: ʿalay-hum, the base form of the hāʾ as we have described it, that it is followed by an u, with a loss of the [full] ū. However, the hāʾ can also be followed by an i due to the yāʾ which precedes it. [To summarise:] what is before the pronominal mīm is followed by an u, which is present due to the mīm, but it is switched to an i due to the yāʾ’ (Zaǧǧāǧ Maʿānī I, 51–2). Zaǧǧāǧ continues by noting the popularity of ʿalay-him and ʿalay-hum in the Qurʾān. He then chastises Sībawayhi for quoting ‘aberrant poetry’ containing ʾaḥlāmi-kim (your dreams), which, in his estimation, is in fact ʾaḥlāmi-kum.

Zaǧǧāǧ concludes: ‘As for ʿalay-hvmū, the original form of the plural is that it has an ū […]. As for those who read Q. 1:7 ʿalay-humū wa-lā ḍ-ḍāllīna ([not] those [who have earned your anger] nor those who have gone astray) – it is rare. And it is not appropriate (lā yanbaġī) to read with anything but what is popular, despite there being people who have read like this, for it is far rarer than dropping [the wāw] in the speech of the Arabs’ (Zaǧǧāǧ Maʿānī I, 52–3).

It would appear then, that in late third century Baghdad, where Zaǧǧāǧ is writing, the contextual -ū had all but disappeared. Once again, this is in stark contrast to what Sībawayhi and Farrāʾ mention, both of whom testify to the popularity of both forms.

Naḥḥās (d. 338/949)

Naḥḥās was an Egyptian grammarian who travelled to Baghdad to study under Zaǧǧāǧ. In his ʾIʿrāb al-Qurʾān, he does not add significantly new information to the discussion of pronouns, but he offers a survey of regional and individual readers associated with the different options (Naḥḥās ʾIʿrāb 14–5, 17). It is significant to note that Naḥḥās is the first grammarian in our discussion who is potentially writing after Ibn Muǧāhid authored his work on the seven readers, and he was also based in Baghdad.

He indicates that there are five options regarding the plural pronominal suffix, and cites Ibn ʾAbī ʾIsḥāq (d. 117/735), an early Basran grammarian and the teacher of the canonical reader ʾAbū ʿAmr (d. 154/770), as reading ʿalay-humū; Naḥḥās refers to this as being the base form (ʾaṣl) of the plural. He then reports that Ḥasan al-Baṣrī (d. 110/728) read ʿalay-himī for the reason of vowel harmony (lit. “because he hated to combine a yāʾ and a ḍamma”). The people of Medina are said to have read ʿalay-him, which is ‘the language of the people of Naǧd’. Ḥamza and the people of Kufa reportedly read ʿalay-hum, and, finally, ʾAʿraǧ (d. 130/748) read ʿalay-himū (Naḥḥās ʾIʿrāb 14–5).

Naḥḥās reports two additional aberrant forms (luġatān šāḏḏatān), which are ʿalay-humu and ʿalay-himi, with shortened final vowels. It appears that this was also the position of Mubarrid (d. 286/898), who authored an earlier but now lost work in the ʾIʿrāb al-Qurʾān genre, whom Naḥḥās cites as disapproving of these final two forms. The attribution of the ʿalay-him form to Najd is consistent with what Farrāʾ relays, and the readings ascribed to Ḥasan and Ḥamza are consistent with what we find in other sources, including the latter’s canonical reading (see below). However, Naḥḥās does not seem to recognise the prevalence of the ʿalay-himū form, mentioning it as ‘the fifth (and final) form’ (al-luġa l-ḫāmisa) and attributed only to a single reader (al-ʾAʿraǧ). This is odd considering works from within the reading tradition transmit this from numerous Hijazi readers including Qālūn ʿan Nāfiʿ, Ibn Kaṯīr, and ʾAbū Ǧaʿfar, as we will see in the next section on the reading traditions (Naḥḥās ʾIʿrāb 15).

Regarding the singular pronominal suffix, Naḥḥās also lists five options, which he ranks in the same order as Zaǧǧāǧ while adding in mentions of readers, ‘Regarding the hāʾ there are five options: the most superior is fī-hi, followed by fī-hu […] and it is the reading of Zuhrī and Sallām ʾAbū l-Munḏir. Next is fī-hī … which is the reading of Ibn Kaṯīr. Also possible are fī-hū and fī-h when assimilated’ (Naḥḥās ʾIʿrāb 17).¹⁵ Here, Naḥḥās expresses awareness of a number of readers, including, Ibn Kaṯīr, which is surprising considering his silence regarding the plural form ʿalay-himū, also read by Ibn Kaṯīr.

Fārisī (d. 377/987)

Fārisī’s Ḥuǧǧa is primarily a work in defence of the readings canonised by his teacher Ibn Muǧāhid. The data of interest here are reports that he relays on the authority of ʾAbū Ḥātim al-Siǧistānī (d. 255/869), whose works are largely lost. Fārisī transmits through ʾAbū Ḥātim that the reading of ʾAʿraǧ (d. 130/748) was ʿalay-himū, which is consistent with what Naḥḥās reports. ʾAbū Ḥātim also said that he had ‘never heard anyone recite with kasr of the mīm without following it with a yāʾ in context, nor anyone recite with ḍamm of the mīm except that they followed it with a wāw in context’ (Fārisī Ḥuǧǧa I, 151). In other words, ʾAbū Ḥātim is indicating that those who vowel the plural suffix all lengthen it. The final relevant report from ʾAbū Ḥātim is that, during his time, the kasr of the hāʾ along with the ḍamm of the mīm before a consonant cluster, as in Q. 2:61 ʿalay-himu ḏ-ḏillatu (humiliation [struck] them), is ‘a widespread (fāšiya) language in the two holy cities’ (Fārisī Ḥuǧǧa I, 149).

A very interesting and revealing comment from Fārisī regarding the unconditional ḍamm of the hāʿ is that ‘it is the language of Qurayš, and it is the old reading’ (Fārisī Ḥuǧǧa I, 149). Here, Fārisī is recognising that unharmonised forms represent the oldest readings which apparently fell out of fashion for one reason or another over time.

Later grammarians

Ibn al-ʾAnbārī (d. 577/1181) only has a brief discussion of plural pronouns, mentioning that the base of the plural is ʿalay-humū, and that while harmonising to ʿalay-him is well and good, some unnamed grammarians expressed disapproval of the harmonised forms (Ibn al-ʿAnbārī Bayān I, 39–40). He discusses the singular pronoun in greater length, indicating that two popular ways of reading fīh (in context) are fī-hi and fī-hī. Though both are approved, he expresses preference for the short -i over -ī, ‘the reading […] fī-hi is better (ʾawǧah) than […] the reading fī-hī’ (Ibn al-ʾAnbārī Bayān I, 44–5).

ʿUkbarī (d. 616/1219) enumerates ten options for the plural pronominal suffix and indicates that all ten have been used in Qurʾānic recitation. These correspond to all possible combinations of i and u vowels following the hāʾ with i, ī, u, ū, and no vowel following the mīm.¹⁶ He does not discuss any readers, except for the passing reference to Ibn Kaṯīr who is said to have read with a long -ū. No preference is given to any particular form either, although ʿalay-himī is noted as being ‘acceptable despite being weak (ǧāʾiz ʿalā ḍaʿf)’. This is a striking deviation from the norms as presented by Sībawayhi and later Mubarrid who clearly viewed the ʿalay-himī form as the default and even preferable form to ʿalay-himū. For the singular suffix, ʿUkbarī lists all four possibilities: -hu, -hū, -hi, and -hī, once again without expressing any preference for a particular form (ʿUkbarī Tibyān I, 11–4).¹⁷

Hamaḏānī (d. 643/1245) reproduces an amalgamated discussion from ʿUkbarī and Zaǧǧāǧ in his work, without any new additions (Hamaḏānī Farīd I, 91–4).

Raḍī l-ʾAstarabāḏī (d. 686/1287) in his famous commentary on Ibn al-Ḥāǧib’s Kāfiya first mentions the Hijazi unconditional ḍamm of the singular pronominal suffix, giving laday-hū and bi-hū as examples, while others harmonise with kasr. Words such as min-hu, which are not preceded by a vowel are typically articulated as such by almost everyone, are reported on the authority of ʾAbū ʿAlī l-Fārisī to have been pronounced as min-hī by ‘some people from Bakr b. Wāʾil’ (ʾAstarabāḏī Šarḥ II, 133). As for the vowel length, ʾAstarabāḏī tells us to look to what precedes the hāʾ: if it is a short vowel, then the vowel following the hāʾ is lengthened as in bi-hī or bi-hū and ḍaraba-hū. An exception to this are reportedly the tribes of Banū ʿAqīl and Kilāb, both of whom approve of the shortened forms, and even the complete dropping of the vowel. He does note that others are not so approving, and only allow it under poetic licence.

When the hāʾ follows a vowelless consonant or weak consonant, such as ʿalay-hv or min-hv, then, he says, ‘the choice is to shorten the vowel’. Here he sides with Mubarrid and Farrāʾ against Sībawayhi. He also notes that Ibn Kaṯīr read with unconditional vowel lengthening.

ʾAstarabāḏī enumerates the different vowel combinations while indicating their popularity, ‘The addition of i to the hāʾ without lengthening is most popular after a yāʾ and less so after an i vowel, because the former would be like a consonant cluster. The addition of ī after the hāʾ, such as bi-hī or ʿalay-hī, is more popular after an i vowel and less so after a yāʾ for the reasons we stated’ (ʾAstarabāḏī Šarḥ II, 134–5). The -hu and -hū forms are mentioned for reference only, with no indication of them being used.

A new scenario is also described by ʾAstarabāḏī for weak consonants that are dropped in jussive such as yarḍa-h and nuṣli-h, and pausal forms like ʾalqi-h and aġzi-h, where there are a number of possibilities. Vowel lengthening is allowed out of consideration for the preceding vowel in speech, and shortening is also allowed out of consideration for the consonant that was dropped. The final allowance is to drop the vowel entirely, treating the contextual form as pausal. He makes explicit mention that all three of these options have been used to read the Qurʾān (ʾAstarabāḏī Šarḥ II, 135).¹⁸

Discussing the hāʾ in dual and both masculine and feminine plural suffixes, ʾAstarabāḏī repeats the aberrant forms like min-himā and min-him on the authority of Fārisī. He states that those who have unharmonised singular pronouns like bi-hū and ʿalay-hū, namely the Hijazis, also have unharmonised duals and plurals like ġulāmay-humā, ġulāmay-hum, and ġulāmay-hunna. Then he cites Ḥamza’s reading as doing ḍamm of the three plurals: ʾilay-hum,ʿalay-hum, and laday-hum, stating that ‘it was said that the reason behind this is that the yāʾ in those words replace an ʾalif, so the yāʾ was treated as if it was the original [letter, i.e. ʾalif]’. However, ʾAstarabāḏī is not convinced by this and responds, ‘Using this logic, it would have been necessary to read the singular, dual, and feminine plurals as ʿalay-hu, ʿalay-humā, and ʿalay-hunna –but it was not read so’. He tempers his frustration by reasoning that ‘perhaps this [mismatch] is due to following a precedent (wa-laʿalla ḏālika li-ttibāʿ al-ʾaṯar, ʾAstarabāḏī Šarḥ II, 135–6)’.

Moving on to the mīm in plural pronouns of the -him form, he affirms the contextual nature of final vowel, and opines that if it is followed by a consonant cluster, then ‘the kasr of the mīm is more fitting [ʾaqyas], as in Q. 28:23 min dūni-himi mraʾatayni (besides them, two women) and in Q. 2:61 ʿalay-himi ḏ-ḏillatu (upon them is humiliation), in accordance to the reading of ʾAbū ʿAmr. The remaining readers employ ḍamm in consideration of the base form’. As for vowel lengthening, he reports that it is less popular than dropping the vowel in context, and uses Ibn Kaṯīr’s reading of ʿalay-himū as an example. He tacitly disapproves of this, indicating that ‘the lengthening with a -ī (to yield -himī) is more fitting [ʾaqyas] to be followed’. However, he quotes Fārisī as disapproving of lengthening with an -ī (ʾAstarabāḏī Šarḥ II, 136–7).

Samīn al-Ḥalabī (d. 756/1355) reiterates that the unharmonised hāʾ in the plural is a Hijazi feature and that vowelless mīm-s are most popular along with -i before consonant clusters (for those who harmonise -him).¹⁹ He once again mentions the ten options, though, unlike ʿUkbarī, Ḥalabī states that only some of them were used in recitation (Ḥalabī Durr I, 70–1). This enumeration is attributed by Ḥalabī to ʾAbū Bakr b. al-ʾAnbārī (d. 328/940),²⁰ who is perhaps a source for ʿUkbarī as well.

For singular forms, Ḥalabī again cites the disharmony of the Hijazis, and quotes Ḥamza as reading Q. 20:10 li-ʾahli-hu mkuṯū ([he said] to his family: ‘stay’), and Ḥafṣ’ reading Q. 48:10 ʿāhada ʿalayhu ḷḷāha ([what] he has covenanted with God) and Q. 18:63 wa-māʾansānī-huʾillā (and none made me forget except [Satan]), all in ‘the language of the Hijaz’ (Ḥalabī Durr I, 88). He also comments on vowel length, stating that ‘the most popular [option], if it is not followed by a consonant cluster, and is preceded by a consonant, such as fī-hv²¹ and min-hv, is to shorten (al-iḫtilās), though lengthening is permitted, and it was read so by Ibn Kaṯīr. If what is before it is vowelled, then it is lengthened, although it may be shortened or dropped entirely (tusakkan, Ḥalabī Durr I, 88)’. In terms of the shortening after closed syllables, ʾAstarabāḏī and Ḥalabī both follow Farrāʾs and Mubarrid’s preference over Sībawayhi’s, while the lengthened form is acknowledged and considered good. It is not until Suyūṭī (d. 849/1445) that we get an explicit opinion in terms of correctness (faṣāḥa), he tells us that ‘the most correct (al-ʾafṣaḥ) is the shortening [of the vowel of -hū] after a sākin’ (Suyūṭī Hamʿ II, 38).

Summary

The grammarians discussed here all present a fairly consistent system. They all discuss both the option of vowel harmony in the pronouns, as well as the option without. The unharmonised forms are universally attributed to the Hijazi dialect. Especially the earlier grammarians do not seem to clearly prefer the forms with vowel harmony over those without.

From Sībawayhi onwards, we find that when a preference is expressed explicitly, all grammarians are in agreement that having the long forms of the singular pronoun -hū/-hī after long vowels and diphthongs is less good than shortening it. However, when it comes to preceding heavy syllables due to them ending on a consonant, there is clear disagreement between the grammarians. Sībawayhi is explicit in considering min-hū better than min-hu, and his student ʾAḫfaš follows him in this, going against what becomes the later classical norm. Farrāʾ, however, already advocated for the shortened form min-hu, an opinion shared by Mubarrid who explicitly says that forms like min-hu are better.

For the plural pronouns it is frequently affirmed that using either the short forms or the long forms is purely up to the speakers. Some grammarians explicitly discuss the existence of the harmonised form -himī, and to Sībawayhi that even seems to be the default option if one lengthens and harmonises. While some grammarians describe the -himū form as well, it seems clear that the -himī was considered the default for some time after the earliest grammarians, particularly among the Basrans.

The most striking thing that we see in this discussion is a clear shift in ideology of what is ‘correct’ classical Arabic. Where Sībawayhi, Farrāʾ and ʾAḫfaš still list most forms as being equally acceptable (with the exception of the noted disagreement on whether one says min-hu or min-hū) this changes with Mubarrid who establishes a much more normative paradigm. Mubarrid explicitly calls harmonised singular -hi/ī the best practice (al-ʾaḥsan), harmonising the plural -him better (ʾaḥsan), using short forms after heavy syllable excellent Arabic (ʿarabī ḥasan) and the use of short pronouns good and what most people do. His normative system is identical to the now commonly accepted classical norms. This clearly shows that by the second half of the 3^rd/9^th century the norms of Classical Arabic had begun to crystallise. This is in clear contrast with the grammarians before him who are much less normative and more accepting of morphological variation in the classical language. The grammarians we examined that come after him are actually somewhat less explicitly prescriptive, but seem to take the norms prescribed by Mubarrid as an implicit norm as well.

Reading traditions

The emergence of Qurʾānic readers (qurrāʾ) as a distinct class of people must have coincided with the explosion of vocalised Qurʾānic manuscripts from the second century onwards. Of the countless readers who emerged across the various metropoles of the Muslim world, seven were canonised by the influential 4^th/10^th century scholar ʾAbū Bakr b. Muǧāhid (d. 324/936). Ibn Muǧāhid’s efforts can be seen as an attempt to curb the multiplicative growth in reading traditions witnessed during his lifetime (Melchert 2000: 18). In his book, Kitāb al-Sabʿa, Ibn Muǧāhid documents the readings of the Meccan Ibn Kaṯīr (d. 120/738), the Medinan Nāfiʿ (d. 169/785), the Syrian Ibn ʿĀmir (d. 118/736), the Basran ʾAbū ʿAmr (d. 154/770), and three Kufans, Ḥamza (d. 156/773), Kisāʾī (d. 189/804), and ʿĀṣim (d. 127/745). It is not entirely clear why Ibn Muǧāhid chose those particular readers – he does not inform us in his book – but some have suggested it may have been informed by the popularity of those readers in their respective regions, at least among the qurrāʾ (Nasser 2013a: 54–61; Shah 2004).²²

Although Ibn Muǧāhid’s selection of only seven readers relegated the remainder to the status of šāḏḏ (non-canonical),²³ many continued to be documented and transmitted in written and oral form, as evidenced by the works cited in this paper. The form of transmission and documentation varied, however. Some readings remained alive as complete systems; their ʾisnād-s (chains of transmitters) and specific features documented with the same degree of thoroughness as the canonical readings. Others only survived in fragmentary form, usually without ʾisnād-s and limited to individual words. They are commonly found as part of lists given in works dedicated to šāḏḏ readings or cited in grammatical works.²⁴

Over time, it became clear that three additional readings were still widely recognised for their transmission and reliability, despite their non-canonical status. The first was the reading of the Medinan ʾAbū Ǧaʿfar (d. 130/747) who was also a teacher of the canonical-reader Nāfiʾ. The second was that of the Basran reader Yaʿqūb al-Ḥaḍramī (d. 205/820). The third was the reading of Ḫalaf (d. 229/844), himself a transmitter of Ḥamza’s reading, who had synthesised his own reading (iḫtiyār). All three of those were eventually incorporated into the canon by Ibn al-Ǧazarī (d. 833/1429) in the 9^th/15^th century.

Individual readers rarely have a single uniform reading attributed to them. Different transmissions going back to a single reader often disagree with one another to various degrees. These differences can be due to variation on a reader’s part, intended differences (iḫtiyār) introduced by a transmitter, or errors in transmission. Increasing disagreements among transmitters, primarily due to the first two categories, were consolidated through a process of transmitter canonisation. A convention emerged wherein two canonical transmitters were chosen for each reader (Nasser 2013b). The resulting ten canonical readers along with their respective twenty transmitters are summarised in the table below.

Table

District	Reader	Transmitters
Medina	Nāfiʿ (d. 169/785)	Warš (d. 197/812)	Qālūn (d. 220/835)
Mecca	Ibn Kaṯīr (d. 120/738)	Bazzī (d. 250/864)	Qunbul (d. 291/904)
Damascus	Ibn ʿĀmir (d. 118/736)	Hišām (d. 245/859)	Ibn Ḏakwān (d.242/856)
Basra	ʾAbū ʿĀmr (d. 154/770)	Dūrī (d. 246/860)	Sūsī (d. 261/874)
Kufa	ʿĀṣim (d. 127/745)	Ḥafṣ (d. 180/796)	Šuʿba (d. 193/809)
Kufa	Ḥamza (d. 156/773)	Ḫalaf (d. 229/844)	Ḫallād (d. 220/835)
Kufa	Kisāʾī (d. 189/804)	Dūrī (d. 246/860)	Layṯ (d. 240/854)
Medina	ʾAbū Ǧaʿfar (d. 130/747)	Ibn Wardān (d. 160/776)	Ibn Ǧammāz (d. 170/786)
Basra	Yaʿqūb (d. 205/820)	Ruways (d. 238/852)	Rawḥ (d. 234/849)
Kufa	Ḫalaf (d. 229/844)	ʾIsḥāq (d. 286/899)	ʾIdrīs (d. 292/905)

Below we examine the pronominal suffix systems of the canonical readers named above in addition to non-canonical readers. We rely on the works of Ibn Muǧāhid, Ibn Ḫālawayhi (d. 380/981), Ibn Mihrān (d. 381/992), Ibn Ǧinnī (d. 392/1002), Ḫuzāʾī (d. 407/1017), Dānī (d. 444/1053), Huḏalī (d. 465/1073), ʾAbū Maʿšar al-Ṭabarī (d. 478/1086), Rūḏbārī (d. 489/1096), Nawzawāzī (d. ca. 6^th/12^th c.), ʾAhwāzī (d. 446/1055), and Kirmānī (d. 6^th/12^th c.). When it comes to the canonical readers there is significant agreement among these sources, but with the lesser-known readers we occasionally encounter disagreements. Whenever disagreements occur these will be discussed individually.

When discussing the canonical readings, and whenever there is no further reference, we cite from Ibn Muǧāhid (Kitāb al-Sabʿa 108–11, 130–2). For the non-canonical readers, and transmission of canonical readers not treated by Ibn Muǧāhid we give the citation in the appropriate location.

Ibn Muǧāhid’s Kitāb al-Sabʿa fī l-Qirāʾāt is today widely regarded as the first to canonise the seven readings and its structure and organisation served as a model for many descriptions of the reading traditions that followed. It is moreover the earliest extant work on the reading traditions, although it seems to have been by no means the first work of this type (Nasser 2013: 37).

Ibn Ḫālawayhi, a student of Ibn Muǧāhid, authored the Muḫtaṣar fī šawāḏḏ al-Qurʾān, which is an abridgement of non-canonical readings to go along with his work al-Badīʿ, where he describes Yaʿqūb al-Ḥaḍramī’s reading in addition to the canonical seven. The Muḫtaṣar is the first extant dedicated collection of non-canonical readings after his teacher Ibn Muǧāhid’s canonisation. Shortly after him, we have the works of Ibn Mihrān which are perhaps the earliest surviving books to completely document the additional readings of Yaʿqūb al-Ḥaḍramī, ʾAbū Ǧaʿfar, and Ḫalaf in great detail. He additionally relays the reading of ʾAbū Ḥātim al-Siǧistānī (d. 250/864), a renowned Basran scholar of Qurʾānic orthography and readings.

Ibn Ǧinnī studied under ʾAbū ʿAlī l-Fārisī, whose Ḥuǧǧa work on the seven canonical readings we have already discussed. As Fārisī had been unable to author a similar work on šāḏḏ readings, Ibn Ǧinnī did so instead. He produced al-Muḥtasab fī tabyīn wuǧūh šawāḏḏ al-qirāʾāt wa-l-ʾīḍāḥʿanhā, known simply as al-Muḥtasab. In it he discusses the grammatical and linguistic aspects of various non-canonical readings, much in the same way the Ḥuǧǧa literature does for canonical readings.

Ḫuzāʿī was a controversial but influential figure who was ostracised for allegedly fabricating a reading containing blasphemous variants and attributing it to ʾAbū Ḥanīfa.²⁵ Nonetheless, this did not stop later authorities, including Ibn al-Ǧazarī from relying on his works. He authored al-Muntahā in which he documents fifteen readings. While most of the readings are of no particular interest, since we have already documented them, one in particular stands out. Ḫuzāʿī transmits the complete reading of Sallām ʾAbū l-Munḏir (d. 171/788).

Dānī is among the most famous writers on the readings, mostly due to his extremely lucid simplified learner’s guide to the seven readings al-Taysīr fī l-qirāʾāt al-sabʿ. This work was eventually recomposed into verse form by Šāṭibī (d. 590/1194). This work, known as al-Šāṭibiyya, is still one of the main didactic poems used for the study of the seven readings today. For our purposes, though, al-Taysīr does not innovate at all upon the account of Ibn Muǧāhid. However, his expert manual on the seven Qirāʾāt – Ǧāmiʿ al-bayān fī l-qirāʾāt al-sabʿ al-mašhūra – is of interest. Unlike al-Taysīr, this work gives descriptions of the readings of many non-canonical transmitters of the seven, some of which show deviations from the standard two-rāwī canon.

ʾAbū ʿAlī l-ʾAhwāzī was another controversial scholar. He had numerous teachers – over ninety-five by some accounts – and as many students. He also amassed a suspiciously large number of extremely short isnād-s going back to earlier readers. A number of the individuals in those isnād-s had never been heard of (Ḏahabī Ṭabaqāt 612 #491). Still, he is an extremely important source for later authorities. He authored many works, many of which have not survived. However, some of his smaller books, such as Mufradat al-Ḥasan al-Baṣrī have been edited and we rely upon it here.

Within a decade of one another, both ʾAbū l-Qāsim al-Huḏalī and ʾAbū Maʿšar al-Ṭabarī authored the most expansive collections of reading traditions known. Huḏalī wrote al-Kāmil, where he documents fifty different readings; we only report relevant details. It is worth noting that Huḏalī’s work, which we rely upon a lot, is known for having many suspect transmissions and errors (Ḏahabī Ṭabaqāt 16 #529). ʾAbū Maʿšar wrote Sawq al-ʿArūs, which stands as the largest collection of transmitted reading routes (ṭuruq) to have survived (and perhaps ever written). In it he collects many single strand transmissions not appearing elsewhere.

ʾAbū Bakr al-Rūḏbāri’s book, Ǧāmiʿ al-Qirāʾāt, was recognised by Ibn al-Ǧazarī as a work ‘the likes of which have never been written’ (Ibn al-Ǧazarī Ġāyat II, 82). In it he documents the ten readings, collecting many unique single strand transmissions, in addition to the readings of Zuhrī, Ibn Muḥayṣin, ʾAʿmaš and numerous others. What sets it apart is Rūḏbārī’s predominant reliance on oral transmission rather than previous written works; al-Ǧāmiʿ was recently edited and published for the first time in 2017.

Not as much is known about ʾAbū ʿAbd Allāh al-Kirmānī, who authored works on Qurʾānic orthography and reading traditions. His book on non-canonical readings, Šawāḏḏ al-Qirāʾāt, as noted in the introduction, was compiled from earlier sources. Asides from those mentioned above, Kirmānī relied upon the works of ʾAbū l-Faḍl al-Rāzī (d. 454/1062), ʾAhwāzī, Zaǧǧāǧ, and other unknown individuals whose works are lost. The same is true for Nawzawāzī, about whom virtually nothing is known.

For the following overview it is useful to define several terms formally, so that they can be described succinctly. When these terms are used below, they should be understood as defined here, unless explicitly provided with further qualification for exception.

Singular vowel harmony: -hu/ū becomes -hi/ī after i, ī, ay.

Singular length disharmony: -hū/ī occurs after light syllables, -hu/i occurs after heavy syllables.²⁶

Long plural pronouns: -humū, ʾantumū etc. rather than -hum,ʾantum (which will be called short plural pronouns). Harmonised long plural pronouns can be either -himū or -himī and this will be mentioned explicitly.

Plural pronouns before consonant clusters can be harmonised either as -himu or -himi or unharmonised -humu. In non-harmonising environments the pronouns are always -humu.

ʾAbū Baḥriyya l-Ḥimṣī (d. ca. 80/700)

The reading of ʾAbū Baḥriyya l-Ḥimṣī was documented by Ḫuzāʿī in much detail. However, given that he lived exceptionally early and his reading is poorly attested,²⁷ we include his reading here with some scepticism. ʾAbū Baḥriyya is reported to have had singular vowel harmony and length disharmony, short harmonised plural pronouns, and an unharmonised -humu before consonant clusters, thus following a system identical to those of the canonical readers Kisāʾī and Ḫalaf (Ḫuzāʿī Muntahā 550).

Ḥasan al-Baṣrī (d. 110/728)

The reading of Ḥasan al-Baṣrī²⁸ is recorded by ʾAhwāzī in a dedicated Mufrada. Ḥasan had singular vowel harmony and length disharmony. He used long plural pronouns where the harmonised form was -himī rather than himū, e.g. ʿalay-himī, min-humū, and bi-himī (ʾAhwāzī Mufradat 71). Plural pronouns before consonant clusters harmonised as -himi (i.e. bi-himi l-ʾasbāb, humu l-ġālibūn). Nawzawāzī (Muġnī 368) attributes the form ʿalay-humī to Ḥasan.

Ibn ʿĀmir (d. 118/736)

The Damascene reader Ibn ʿĀmir²⁹ has singular vowel harmony and length disharmony, short harmonising plural pronouns and the plural pronoun harmonised before consonant clusters as -himu.

There is a non-canonical transmission path of Ibn Ḏakwān (d. 242/856) through Ibn ʿAbd al-Razzāq that lacks vowel harmony of -humu before consonant clusters (Ḫuzāʿī Muntahā 550).

Ibn Kaṯīr (d. 120/738)

The Meccan Ibn Kaṯīr³⁰ has singular vowel harmony, but no length disharmony, always using the long pronominal forms, thus both bi-hī and fī-hī. He always employs long plural pronouns, which are harmonised as -himū. Before consonant clusters it is -himu.

ʾAbū Maʿšar al-Ṭabarī (Sawq al-ʿArūs II, 469ff.) and Nawzawāzī (Muġnī 368) transmit that ʾAbū ʿAmr (d. 154/770) ← Ibn Kaṯīr and Ḫuzaymī (d. 4^th/10^th c.) ← Ibn Fulayḥ (d. ca. 270/883) ← Ibn Kaṯīr³¹ lengthen the pronouns and would not harmonise the plural pronoun at all, i.e. ʿalay-humū, yuzakkī-humū, samʿi-humū. Before consonant clusters they also transmit an invariably unharmonised -humu.

Surprisingly, Ibn Ḫālawayhi reports ʿalay-humū for Ibn Kaṯīr, even though he traces his ʾisnād through the transmission of Ibn Muǧāhid ← Qunbul (d. 291/904) ← […] ← Ibn Kaṯīr. This is the case in his Kitāb al-Badīʿ³² but the same is reported in his ʾIʿrāb al-Qirāʾāt as well (Ibn Ḫālawayhi ʾIʿrāb 51) which makes it unlikely that this is a mistake of the copyist of the Badīʿ. Ibn Muǧāhid, however, is unambiguous in his wording that Ibn Kaṯīr read ʿalayhimū.

Huḏalī (Kāmil 466) reports length disharmony through Ibn ʿUyayna (d.198/814) ← Ibn Kaṯīr.

Ibn Muḥayṣin (d. 123/741)

Ibn Muḥayṣin,³³ another Meccan reader, is reported to have had the same system as Ibn Kaṯīr, lacking singular length disharmony, and using long plural pronouns. An exception to this is found from one transmitter, Ibn [ʾAbī]³⁴ Yazīd (d. ca. 2^nd/8^th c.), who had vowel length disharmony. Also, he transmits short harmonised plural pronouns (Huḏalī Kāmil 466–7, 469).

Muḥammad b. Šihāb al-Zuhrī (d. 124/741–2)

Al-Rūḏbārī documents the reading of the Medinan-turned-Damascene Ibn Šihāb al-Zuhrī, a figure of course primarily known as one of the central early ḥadīṯ tradents. He lacks singular vowel harmony, and follows length disharmony, i.e. ʿalay-hu, fī-hu, bi-hū (al-Rūḏbārī Ǧāmiʿ II, 318–9). He however exceptionally reads with vowel harmony for the words Q. 19:95 ʾātī-hi ([each of them] will come to him), and Q. 70:11 bi-banī-hi (by his children) and (Q. 70:12) wa-ʾaḫī-hi (and his brother), Q. 70 [sic!]³⁵ ʾummi-hī (his mother), Q. 70:12 ṣāḥibati-hī (his wife) and Q. 70:13 wa-faṣīlati-hī (and his nearest kin)

For the plural Zuhrī has the short forms, which normally harmonised but lack harmony in three specific words ʿalay-hum, ʾilay-hum and laday-hum alone – a conditioning he shared with the canonical reader Ḥamza, for which see below. Before consonant clusters his pronouns are always unharmonised -humu.

ʿĀṣim (d. 127/745)

The treatment of the pronominal system by the Kufan ʿĀṣim³⁶ is identical to that of Ibn ʿĀmir. He has singular harmony and length disharmony, and short harmonised plural pronouns. Before consonant clusters the harmonised plural pronoun is -himu.

While this is the basic system, Ḥafṣ (d. 180/796) ← ʿĀṣim deviates from this system in several fixed locations. He reads -hu unharmonised in two locations: Q. 48:10 ʿalay-hu ḷḷāh ([what] he has covenanted with God) and Q. 18:63 ʾansā-nī-hu ([and none] made me forget [except Satan]). And he reads Q. 25:69 fī-hī (therein) without the expected length disharmony. Šuʿba (d. 193/809) ← ʿAṣim does not follow him in this.

Dānī (Ǧāmiʿ 164) brings a transmission from al-ʾAʿšā (d. ca. 200/815) ← Šuʿba who is said to have read -himi before consonant clusters with the words ʿalay-himi and ʾilay-himi whenever they occur and in Q. 5:63 wa-ʾakli-himi s-suḥta (and their consuming of that which is unlawful) and nothing else.

ʿĀṣim al-Ǧaḥdarī (d. 128/745)

Huḏalī (Kāmil 467–8) reports that the Basran ʿĀṣim al-Ǧaḥdarī,³⁷ had singular vowel harmony and length disharmony. For the plural, however, he lacked vowel harmony after -ī and -ay, thus reading fī-hum and ʿalay-hum. Also, the dual -humā and feminine plural -hunna are affected by this.

ʾAbū Ǧaʿfar (d. 130/747)

The Medinan ʾAbū Ǧaʿfar³⁸ has singular vowel harmony and length disharmony, but the plural has long plural pronouns and harmonises this as -himū. Before consonant clusters the harmonised form is -himu.

Ḫuzāʾī (Muntahā 547) reports that ʿUmarī (d. ca. 270/883) ← Qālūn ← Qālūn’s teachers ← ʾAbū Ǧaʿfar read the long form of the pronoun only before a following hamza, a conditioning we will encounter again among the transmissions of Warš ← Nāfiʿ and for two non-canonical transmissions of Kisāʾī.

ʾAʿmaš (d. 148/765)

One of the oldest of the Kufan reciters and a teacher of Ḥamza, ʾAʿmaš³⁹ has singular vowel harmony and length disharmony and harmonising short plural pronouns. He however lacks vowel harmony with the words ʿalay-hum and ʿalay-humā (Huḏalī Kāmil 467–8). Like most other Kufans, before consonant clusters the plural pronoun is always an unharmonised -humu.

ʾAbū ʿAmr (d. 154/770)

The Basran ʾAbū ʿAmr⁴⁰ has singular vowel harmony and length disharmony and harmonising short plural pronouns. Before consonant clusters the harmonised plural pronoun is -himi.

Dānī (Ǧāmiʿ 163) reports that Ibn Ǧubayr (d. 258/872) ← Yazīdī (d. 202/818) ← ʾAbū ʿAmr read penultimate words of a verse with lengthened forms of the plural pronouns, e.g. Q. 2:91ʾin kuntumū muʾminīn# (if you are believers); Q. 2:22 wa-ʾantumū taʿmalūn# (while you know). It is not mentioned explicitly how the harmonised pronoun would be treated in this case.

ʾAbū Maʿšar al-Ṭabarī (Sawq II, 479–80) reports for Ḫāriǧa (d. 168/784) and Hārūn (d. 170/786) ← ʾAbū ʿAmr long plural pronouns with full harmony. This gives humū, laʿalla-humū, ʿantumū, and la-humū, but also ʿalay-himī, fī-himī, qulūbi-himī, and so on. This system is identical to that of Ḥasan al-Baṣrī.

Ḥamza (d. 156/773)

Ḥamza⁴¹ has singular vowel harmony and length disharmony and harmonising short plural pronouns. Before consonant clusters the plural pronoun is the unharmonised -humu. He lacks vowel harmony with three specific words: ʿalay-hum, ʾilay-hum and laday-hum whenever they occur.

Huḏalī (Kāmil 467) transmits from ʾAbū l-Zaʿrāʾ (d. 283/896) ← Ḥamza that he exceptionally read Q. 16:106 fa-ʿalayhim ġaḍabun (so wrath is upon them).

Ḥamza breaks with the vowel harmony rule of -hu in two places. He reads Q. 20:10, Q. 28:29 fa-qāla li-ʾahli-hu mkuṯū (he said to his family: ‘stay’).

ʾAbū Maʿšar al-Ṭabarī (Sawq II, 469–72) reports a single strand transmission from ʾAhwāzī ← Kisāʾī ← Ḥamza with a different pronominal system. He transmits that Ḥamza always reads the plural pronoun unharmonised (e.g. ʿalay-hum, fī-hum, and samʿi-hum), and that the singular pronouns was unharmonised after ī and ay (i.e. bi-hī, but ʿalay-hu and fī-hu).

Nāfiʿ (d. 169/785)

The transmitters of the Medinan reader Nāfiʿ⁴² agree on the treatment of the singular pronouns as having vowel harmony and length disharmony. They all agree that before consonant clusters the harmonised pronoun is -himu.

When it comes to the length of the plural pronoun, there is disagreement. Qālūn ← Nāfiʿ has the option to use the short forms -hum/-him, or the long forms -humū/-himū.⁴³ Ibn Muǧāhid reports the same option for the non-canonical transmitters ʾIsmāʿīl b. Ǧaʿfar (d. 180/796), Ibn Ǧammāz, and Musayyabī (d. 236/850). Dānī (Ǧāmiʿ 157) reports it for the same readers but does not mention Ibn Ǧammāz.

Warš (d. 197/812) ← Nāfiʿ uses both the long and short plural pronouns with a specific phonetic conditioning. By default, the plural pronoun is -hum/-him, but when this pronoun precedes a word that starts with hamza, he uses the long forms -humū/-himū.

Dānī (Ǧāmiʿ 160) reports one single strand transmission from Ḫalaf b. ʾIbrāhīm (d. 402/1011) ← […] ← Yūnus (d. 264/877) ← Warš that he left as an option whether he harmonised long form as -himū or -himī.

There are reports of several fixed locations that have exceptions to the harmonisation rule of the singular. Ibn Saʿdān (d. 231/845) ← Musayyabī ← Nāfiʿ follows Ḥamza in reading Q. 20:10 (and Q. 28:29) fa-qāla li-ʾahli-hu mkuṯū (he said to his family: ‘stay’, Ibn Muǧāhid Sabʿa 417). Ibn al-Musayyabī ← Musayyabī reads Q. 6:46 yaʾtī-kum(ū) bi-hu nẓur (to bring them [back]. Look!). Later this variant is also attributed to one of the transmission paths of Warš, that of ʾAṣbahānī (d. 296/908) (Ibn al-Ǧazarī Našr II, 994–5). Additionally, Ibn Saʿdān ← Musayyabī, and Ibn Baḥr (d. unknown) ← Musayyabī add Q. 28:81 wa-bi-dāri-hu l-ʾarḍa ([and we caused] the earth [to swallow him up] and his home) (Huḏalī Kāmil 466).

There are also some exceptions reported on the vowel length disharmony of the singular pronoun. Musayyabī ← Nāfiʿ read Q. 20:32 wa-ʾašrik-hū fī ʾamrī (and let him share my task). Both Ibn Saʿdān ← Musayyabī and Kisāʾī⁴⁴ ← ʾIsmāʿīl b. Ǧaʿfar ← Nāfiʿ always read ʿalay-hī with a lengthened vowel.

Dānī (Ǧāmiʿ 158–9) reports a more complex conditioning for the use of the long plural pronoun for Ḥulwānī ← Qālūn. He lengthens it if 1. the next word starts with a hamza, e.g. Q. 2:78 wa-min-humū ʾummiyyūn (and among them are gentiles); 2. the next word starts with a m, e.g. Q. 2:201 min-humū man yaqūlu (but among them is he who says); 3. the next word is the end of the verse according to the Medinan verse count, e.g. Q. 2:8 mā humū bi-muʾminīn# (but they are not believers).

Sallām ʾAbū l-Munḏir (d. 171/788)

Ḫuzāʿī transmits the complete reading of Sallām ʾAbū l-Munḏir,⁴⁵ who already receives brief mention by Naḥḥās. Importantly, Sallām was a Basran-turned-Kufan reader who was a student of the Kufan eponymous reader ʿĀṣim and the Basrans Ǧaḥdarī and ʾAbū ʿAmr (Ḫuzāʿī Muntahā I, 417); he was also the teacher of the canonical reader Yaʿqūb.

Sallām had singular length disharmony, but did not apply vowel harmony after ī or ay, so fī-hu, ʾilay-hu, ʿalay-hu but bi-hī (Huḏalī Muntahā III, 573).

This distribution is parallel to the one reported for the plural and dual, where vowel harmony is likewise blocked on the short pronouns if ī or ay precedes, i.e. ʿalay-hum, fī-hum, ʿalay-humā, fī-humā (Ḫuzāʿī Muntahā II, 545). Before a consonant cluster the harmonised pronoun is -himi (Ḫuzāʿī Muntahā II, 550).

ʾAbū Maʿšar al-Ṭabarī (Sawq II, 479–80) adds that Sallām, like Rawḥ ← Yaʿqūb does not harmonise after the short -i of apocopates and imperative final weak verbs.

Kisāʾī (d. 189/804)

The Kufan Kisāʾī⁴⁶ has singular vowel harmony and length disharmony, and always reads unharmonised -humu before consonant clusters. For his canonical transmitters, the plural pronoun is always short and harmonising.

ʾAbū Maʿšar al-Ṭabarī (Sawq II, 469–472) brings several chains of transmission through Rāzī ← Kisāʾī that make an exception to the general rule of vowel harmony with the words ʿalay-hum, ʾilay-hum and laday-hum, a conditioning that is typical of his teacher Ḥamza.

The non-canonical transmitters Nuṣayr and Qutayba are fairly broadly transmitted in the works that we have examined. The earliest transmission is by Ibn Mihrān (Ġāya 141–2; Mabsūṭ 89), and a much more detailed description is found in Dānī’s Ǧāmiʿ al-Bayān. While they follow all the rules typical of the canonical transmissions of Kisāʾī they both occasionally make use of the long plural pronoun -humū when it is in a non-harmonising environment. This however only occurs in highly specific environments which are discussed in detail in the sections below.

Nuṣayr (d. ca. 240/855)

The accounts of Dānī and Ibn Mihrān differ somewhat and we will discuss both of them here. Ibn Mihrān (Mabsūṭ 89) tells us that Nuṣayr ⁴⁷ would use long pronouns when the pronoun occurs:

1. next to the final word of an ʾāya, e.g. Q. 27:3 humū yūqinūn# (and they are certain).

2. before hamza, e.g. Q. 2:140 ʾa-ʾantumū ʾaʿlam (are you more knowing?).

3. Before m, e.g. Q. 2:92 ǧāʾa-kumū mūsē (and Moses brought you).

But this lengthening is blocked if:

1. The pronoun stands in a harmonising context, e.g. Q. 6:1 bi-rabbi-him yaʿdilūn# (they equate with their lord).

2. If the word is long,⁴⁸ e.g. Q. 2:23 šuhadāʾa-kum min dūni llāh (your witnesses other than God), or if the word contains a geminate (ṯaqulat), e.g. Q. 7:176 laʿalla-hum yatafakkarūn# (so perhaps they will reflect).

Dānī (Ǧāmiʿ 162–3) follows the same three main conditions but the context in which the harmony is blocked is slightly different. Like with Ibn Mihrān, vowel harmony of the pronoun blocks the lengthening, but gemination in a word does not. The only condition involved with blocking the lengthening besides that is that the word ought to be five letters or less according to the spelling of the muṣḥaf. In practice, this means that e.g. Q. 2:21 laʿalla-kumū tattaqūna (so that perhaps you may become righteous) is lengthened, while in Ibn Mihrān’s description it would not be.

Ibn Muǧāhid (Sabʿa 98) also transmits Nuṣayr ʿan Kisāʾī but does not report anything unusual about the plural pronouns for him (Ibn Muǧāhid Sabʿa 108–9).

Qutayba (d. ca. 200)

Ibn Mihrān (Mabsūṭ 89) and Dānī (Ǧāmiʿ 160–1) transmit the same conditions for Qutayba.⁴⁹ When the pronoun is not harmonised it is lengthened whenever:

1. It is directly followed by the last word of the verse. wa- nor any other short word like min or lā may intervene, e.g. Q. 2:3 mimmā razaqnā-humū yunfiqūn# (and they spend of what We have provided for them), but Q. 26:94 fa-kubkibū fī-hā hum wa-l-ġāwūn# (they will be hurled into it, together with the deviators).

2. If it is directly followed by a word that begins with a hamza, e.g. Q. 2:6ʾa-ʾanḏarta-humū ʾam lam (whether you have warned them or have not [warned them]).

Yaʿqūb (d. 205/820)

The earliest detailed description of his pronominal system is provided by Ibn Mihrān (Ġāya 140–3).⁵⁰ Yaʿqūb⁵¹ has singular vowel harmony and length disharmony. Its plural pronouns are short, and the harmonised form before consonant clusters is -himi. Like his teacher Sallām, Yaʿqūb lacks vowel harmony of the plural pronoun and the dual after ī and ay, thus he reads fī-hum and ʿalay-hum.

There is disagreement between the two canonical transmitters of Yaʿqūb, Rawḥ and Ruways, on how to treat the plural pronoun after an apocopate/imperative form of a final weak verb that ends in i. Rawḥ treats those forms as if they have a shortened ī, and therefore treats -hum as if it is preceded by ī, e.g. Q. 7:169 yaʾti-hum ([and if an offer like it] comes to them) whereas Ruways treats this as any normal i and therefore applies vowel harmony, i.e. yaʾti-him.

There are two cases where Yaʿqūb harmonises against his harmonisation rule, namely in the phrase Q. 7:17 (and Q. 41:14) min bayni ʾaydī-him wa-min ḫalfi-him ([I will come to them] from before them and from behind them).⁵² Ruways makes an exception to his rule concerning apocopates in Q. 8:16 wa-man yawalli-him (and whoever turns his back on them).

Huḏalī (Kāmil 466) reports that all readers he knows of lengthen the singular pronoun -hū/-hī after a single short vowel, with the exception of Naḥḥās ← Ruways ← Yaʿqūb, who is said to shorten it in the phrase bi-yadi-hi whenever it occurs.

It seems that Yaʿqūb chose to stick to Sallām’s plural forms but opted for the more usual harmonised form in the singular. That would put some degree of classicisation at the level of this eponymous reader. Alternatively, it could be that the transmitters of Yaʿqūb’s reading classicised his singular form.

ʾAbū ʿUbayd al-Qāsim b. Sallām (d. 224/838)

Ḫuzāʾī (Muntahā 550) described ʾAbū ʿUbayd’s⁵³ treatment of pronouns as being identical to ʾAbū ʿAmr’s. Thus, the singular pronoun has vowel harmony and length disharmony, the plural pronoun is harmonised and short and before consonant clusters the harmonised pronoun is -himi.

Ḫalaf (d. 229/844)

Ḫalaf ⁵⁴ is a transmitter of Ḥamza, but also one of the ten canonical readers in his own right. In terms of his pronominal system, his personal reading is identical to the canonical transmissions of Kisāʾī: the singular pronoun has vowel harmony and length disharmony, the plural pronoun is short and harmonised, and before consonant clusters the plural pronoun is always an unharmonised -humu (Ibn al-Ǧazarī Našr II, 868–76, 971–3).

ʾAbū Ḥātim al-Siǧistānī (d. 255/869)

ʾAbū Ḥātim⁵⁵ was a famous Basran philologist, who had constructed his own reading. Ibn Mihrān (Ġāya 140–3) gives the earliest description of his reading. The singular pronoun is harmonised and has length disharmony. Before consonant clusters the harmonised plural is -himi (Ḫuzāʿī Muntahā 550). The plural pronoun is short and Ḫuzāʿī (Muntahā 547) reports that unlike Yaʿqūb and Sallām, ʾAbū Ḥātim only lacks vowel harmony after -ay and not after -ī, i.e. ʿalay-hum but fī-him.

Yaḥyā b. Ṥubayḥ

Huḏalī (Kāmil 468) dedicates a section to the reading of Yaḥyā b. Ṣubayḥ. He documents that after ay, the singular pronoun is shortened and harmonised in contexts such as ʿalay-hi, ʾilay-hi, and laday-hi.⁵⁶ If the pronoun is preceded by an ī vowel, such as nuṣlī-hu and nūḥī-hu, it is read as an unharmonised -hu, with the exception of the words fī-hi, ʾaḫī-hi, ʾabī-hi, and banī-hi, which are harmonised, and another location specific exception is reported for Q. 18:95 makkan-nī fī-hu rabb-ī ([that] in which my Lord has established me [is better]) which is read unharmonised.

Apocopate final weak verbs do not trigger vowel harmony, not in the plural (as with Ruways ← Yaʿqūb) nor in the singular, thus he reads: nuṣli-hu, wa-qi-hum, and yuḫzi-hum, with a single exception of Q. 20:75 yaʾti-hī muʾminan ([but whoever] comes to him as a believer) which he harmonises and lengthens in context.

Summary

The table on the following page gives an overview of the pronominal systems of the readers we have discussed above. It does not capture all the details of the readings, specifically lacking isolated deviations from the general system of several of the readers. The readers are organised by the city in which they were active. The labels of the different columns refer to the following:

sg. length: Lists whether the singular pronoun is short -hu/-hi or long -hū/-hī after heavy syllables.

i/ī/ay + sg.: these denote the harmonisation of the singular pronoun in the three harmonising environments of Classical Arabic.

i/ī/ay + pl.: these denote the shape of the plural pronoun in the three harmonising environments of Classical Arabic.

ʿalay-him/ʾilay-him/laday-him: Lists the shape of the plural pronoun with these specific lexical items.

pl. + CC: lists the shape of the pronoun before a two consonant cluster in harmonising environments.

pl. + hamza: shape of the plural pronoun (hvm or hvmū) when the next word starts with a hamza

Table

Reader	City	sg. length	i + sg.	ay + sg.	ī + sg.	i + pl.	ʿalay-him	ʾilay-him	laday-him	ay + pl.	ī + pl.	pl. + CC	pl. + hamza
ʾAʿmaš	Kufa	short	-hi	-hi	-hi	-him	-hum	-him	-him	-him	-him	-humu	short
Kisāʾī	Kufa	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-humu	short
Ḫalaf	Kufa	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-humu	short
Ḥamza	Kufa	short	-hi	-hi	-hi	-him	-hum	-hum	-hum	-him	-him	-humu	short
Kisāʾī < Ḥamza	Kufa	short	-hi	-hu	-hu	-hum	-hum	-hum	-hum	-hum	-hum	-humu	short
ʿĀṣim	Kufa	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-himu	short
ʾAbū ʿUbayd	Kufa	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-himi	short
ʾAbū ʿAmr	Basra	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-himi	short
ʾAbū Ḥātim	Basra	short	-hi	-hi	-hi	-him	-hum	-hum	-hum	-hum	-him	-himi	short
Yaʿqūb	Basra	short	-hi	-hi	-hi	-him	-hum	-hum	-hum	-hum	-hum	-himi	short
ʿĀṣim al-Ǧaḥdarī	Basra	short	-hi	-hi	-hi	-him	-hum	-hum	-hum	-hum	-hum	-himi	short
Sallām	Basra	short	-hi	-hu	-hu	-him	-hum	-hum	-hum	-hum	-hum	-himi	short
Ḥasan	Basra	short	-hi	-hi	-hi	-himī	-himī	-himī	-himī	-himī	-himī	-himi	long
ʾAbū Baḥriyya	Syria	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-humu	short
Ibn ʿĀmir	Syria	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-himu	short
Ibn Kaṯīr	Mecca	long	-hi	-hi	-hi	-himū	-himū	-himū	-himū	-himū	-himū	-himu	long
Ibn Muḥayṣin	Mecca	long	-hi	-hi	-hi	-himū	-himū	-himū	-himū	-himū	-himū	-himu	long
Warš	Medina	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-himu	long
Qālūn (1)	Medina	short	-hi	-hi	-hi	-him	-him	-him	-him	-him	-him	-himu	short
Qālūn (2)	Medina	short	-hi	-hi	-hi	-himū	-himū	-himū	-himū	-himū	-himū	-himu	long
Ḥulwānī < Qālūn	Medina	short	-hi	-hi	-hi	-himū	-him	-him	-him	-him	-him	-himu	long
ʾAbū Ǧaʿfar	Medina	short	-hi	-hi	-hi	-himū	-himū	-himū	-himū	-himū	-himū	-himu	long
Zuhrī	Medina	short	-hu	-hu	-hu	-him	-hum	-hum	-hum	-him	-him	-humu	short
Yaḥyā b. Ṣubayḥ	Unknown	short	-hi	-hi	-hu	-him	-him	-him	-him	-him	-hum	-humu	short

Incomplete non-canonical descriptions

The canonical and non-canonical readings elaborated above have descriptions which are quite complete. Qirāʾāt works also document the remnants of additional non-canonical readers, though not in full detail. We do not attempt to verify the veracity of the transmissions, nor exhaustively collect all transmissions of a particular reader, but we give an overview of the reports here.

Huḏalī (Kāmil 466–7) reports the pronominal system of the famous early exegete Muǧāhid b. Ǧabr (d. 104/722),⁵⁷ who is said to have simply had singular vowel harmony and vowel length disharmony, thus reading forms like fī-hi, ʿalay-hi, min-hu, ʿan-hu. The plural pronouns were short and harmonising, and before consonant cluster the harmonised pronoun was -himu (Huḏalī Kāmil 468).

Šibl b. ʿAbbād al-Makkī (d. ca. 160/777),⁵⁸ is reported to have had the option to either read with singular vowel length disharmony or have long vowels in all environments (Huḏalī Kāmil 466–7), with vowel harmony in both cases. He likewise has the option to use the long plural forms -humū/-himū or the short ones instead (Huḏalī Kāmil 468)

Before two consonant clusters the harmonised plural pronoun is -himu. Ṭalḥa b. Muṣarrif b. ʿAmr (d. 146/764),⁵⁹ used -himi, not -himu in harmonising environments before two consonant clusters (Huḏalī Kāmil 467).

Like Sallām ʾAbū l-Munḏir, Ibn Miqsam (d. 354/965) would not apply vowel harmony to -hu if an ī or ay preceded (Huḏalī Kāmil 466). Muslim b. Ǧundab (d. 110/729) reportedly had an unharmonised third person masculine singular pronoun -hu for Q. 2:2 lā rayba fī-hu (there is no doubt in it) (Ibn Ḫālawayhi Muḫtaṣar 2). Ibn Ǧinnī (Muḥtasab I, 44) adds that the plural for him is ʿalay-humū. Ibn Ḫālawayhi (Muḫtaṣar 1) attributed ʿalay-humu (no mention of length of the vowel after the m) to Ibn ʾAbī ʾIsḥāq (d. 117/735). Ibn Ǧinnī (Muḥtasab I, 44) explicitly attributes length to his reading: ʿalay-humū.

ʿIsā b. ʿUmar al-Ṯaqafī (d. 149/766)⁶⁰ read ʿalay-humū (Ibn Ǧinnī Muḫtasab I, 44; Nawzawāzī Muġnī 367). ʿAbd Allāh b. Yazīd (d. 213/829) read ʿalay-humū (Ibn Ǧinnī Muḥtasab I, 44). ʿAmr b. Fāʾid (d. 201/816) is said to have read ʿalay-himi according to Ibn Ḫālawayhi (Muḫtaṣar 1),⁶¹ with no comment on length and ʿalay-himī according to Ibn Ǧinnī (Muḥtasab I, 44). Between Ibn Ǧinnī (Muḥtasab I, 44) and Nawzawāzī (Muġnī 367–8), several forms of the plural pronoun are attributed to ʿAbd al-Raḥmān b. Hurmuz al-ʾAʿraǧ (d. 117/736): ʿalay-humū, ʿalay-himū, ʿalay-himu and ʿalay-humu.

Analysis

The data collected on the readers with complete descriptions contain sufficient information for detailed numerical analysis. We use a statistical procedure known as principal component analysis (PCA) which allows us to visualise and interpret the data present in the summary table. PCA produces a two-dimensional map whose axes capture the maximum amount of variance in the data. For an excellent non-technical introduction of PCA, applied to Qurʾānic data, see Sadeghi (2011: 247–52); we will forgo reproducing a description of the technique here.

Figure 1 shows the PCA of the pronominal suffix treatment of readers enumerated in the summary table. Each dot represents a reader coloured by their respective region. The clustering of readers into groups is immediately apparent. We have shaded and labelled the four primary groups based on the majority of readers and corresponding pronominal features. The Hijazi group contains the expected readers, both Meccan and Medinan, with the additional presence of Ḥasan al-Baṣrī, who we will return to later. What unifies the Hijazi group is plural vowel harmony (min-hum but ʿalay-him), the use of a u before consonant clusters (ʿalay-himu ḏ-ḏillatu), and lengthening of the plural pronoun (min-humūʾilay-him-ū). The separation between Meccans and Medinans is driven by the unconditional lengthening of the singular pronominal suffix -hū/-hī for the former readers. Warš, a transmitter of the Medinan Nāfiʿ, is distinguished from the remaining Hijazis by conditioning lengthening of the plural in the presence of a subsequent hamza. The lack of a complete application of plural lengthening places him farther removed from the rest of the group and closer to the classical group.

Figure 1. Principal component analysis (PCA) of main Qurʾānic readers’ treatment of singular and plural pronominal suffixes.

The Basran cluster contains exclusively Basran readers, all of whom have unharmonised plurals after ay (i.e. ʿalay-hum, miṯlay-hum) and harmonise with -himi in front of consonant clusters (i.e. bi-himi l-ʾasbāb). Indeed, this latter feature is found among all Basran readers, including Ḥasan and the canonical ʾAbū ʿAmr who are not found in the Basran cluster. ʾAbū ʿAmr’s reading is far closer to the classical group which we discuss below. The Basran group, apart from ʾAbū Ḥātim, also lacks harmony after ī (fī-hum).

The Kufan group is not as distinct as the others and exhibits a greater degree of variability. The Kufans ʿĀṣim and Abū ʿUbayd are also excluded from it. In general, they harmonise their plurals, except before consonant clusters where they read -humu (i.e. bi-him min but bi-humu l-ʾasbāb). ʾAʿmaš additionally reads ʿalay-hum while Ḥamza adds ʾilay-hum and laday-hum, putting the latter slightly closer to the Basran group.

The fact that regional pronominal traditions do emerge from the data allows us to say something about the readers who are outside of their region. We find that Zuhrī, who is Medinan, has a pronominal system far closer to the Kufan one, Ḥasan al-Baṣrī is closer to Hijazi, and Abū Baḥriyya, a Homsi reader, has a decidedly Kufan system. What stands out is that all three readers are among the earliest, with death dates of 124/741, 110/728, and ca. 80/700 respectively. This implies that at least in the late 1^th/7^th century, regional pronominal systems had not yet crystallised into their characteristic forms. Not all early readers have anomalous pronominal systems though, as evidenced by Ibn Kaṯīr, who died in 120/738, and ʿĀṣim al-Ǧaḥdarī, who died in 128/745, who have prototypical ones for their respective regions. Of course, this is not unexpected since the emergent regional systems must have come from somewhere. Nonetheless, we can tentatively suggest that the stabilisation of pronominal systems around the mid-2^nd/8^th century may prove to be related to other aspects of the reading traditions or perhaps even with the beginning of the grammatical tradition, and both warrant further investigation.

Another corollary of strong regional traditions is the ability to use pronominal features as a regional litmus test. In our case, we would apply this to the reader Yaḥyā b. Ṣubayḥ, about whom very little is known. He is said to have been the first person to popularise Qurʾānic readings in Nishapur (Mizzī Tahḏīb XXXI, 382). His pronominal system suggests that he has a Kufan reading. However, a glance at his students and teachers in ḥadīṯ suggests a Basran connection, and Huḏalī states that he got his reading from the Meccan ʿAmr b. Dīnār (Huḏalī Kāmil I, 293). However, Ibn al-Ǧazarī explicitly rejects this and instead informs us that he actually learned from ʾIbrāhīm b. Ṭahmān who learned from the Kufan canonical reader ʿĀṣim (Ibn al-Ǧazarī Ġāya II, 374).

This validates the use of pronominal features as a means of regional identification – with one major caveat: our collection of non-canonical reader transmissions has demonstrated that pronominal features are remarkably fluid. Readers often mix features associated with different regions, and in particular lengthening of the plural pronoun. Huḏalī states, ‘there is not a single one of the readers who did not have lengthening of the mīm transmitted from them, either by many transmitters or few, except for Ḥamza and his transmitters [who only transmitted short forms]’ (Huḏalī Kāmil III, 579). Therefore, in identifying unknown readings in Qurʾānic manuscripts, where pronominal features become especially handy, one must keep this in mind.

The fourth and final group is the classical group. As the name suggests, the classical group represents the system which becomes part of normative Classical Arabic, namely short harmonised plurals (min-hum,ʿalay-him) and harmonised singulars with length disharmony (bi-hī, fī-hi). Three of the readers have completely identical systems also agreeing on using the vowel u before consonant clusters (ʿalay-himu ḏ-ḏillatu). ʾAbū ʿAmr and ʾAbū ʿUbayd (who have the same system) only deviate from the other three in front of consonant clusters, where they use -himi instead. The main distinguishing factor between the Hijazi and the classical cluster is the latter group’s use of the short plural pronouns. It is worth noting that, unlike the other three groups, the classical group contains a reader from every single region in our analysis (Kufa, Basra, Hijaz, Syria). The fact that the pronominal system on which readers from different regions converge appears to be identical to what later becomes Classical Arabic could very well be an early testimony of an emerging transregional linguistic norm that ends up becoming the strict Classical Arabic as we know it today.

As we saw above, striking regional trends appear when we examine the pronominal systems of Qurʾānic readers. However, when we compare the regional trends as found in the reading traditions to what the Arab grammarians describe for dialect geography, we notice that these regional trends do not map onto the grammarian dialect geography at all. This is best exemplified by comparing the reading tradition of the Hijaz to what is described for the Hijaz by Farrāʾ in his Luġāt al-Qurʾān. He reports that the people of the Hijaz and Qurayš in particular lack vowel harmony for both the singular and plural (bi-hū, fī-hu, ʿalay-hu and ʿalay-hum) (Farrāʾ Luġāt 10). However, a total lack of vowel harmony is not attested in any regional group of readers at all, least of all the Hijazi cluster, none of whom lack vowel harmony in any environment. Instead, the defining feature of the Hijazi cluster is the lengthened pronouns, a feature which Farrāʾ explicitly describes as an option for all dialects of Arabic (Farrāʾ Luġāt 33). Instead, we find that the use of unharmonised pronouns is typical for the Basran cluster – although not a single reader applies it unconditionally to both the singular and the plural as described for the Hijaz.⁶²

General harmony (ʿalay-hi and ʿalay-him) is ascribed to the Najdi tribes of ʾAsad, Qays, and Tamīm, with ʾAsad reportedly breaking consonant clusters with the vowel u (ʿalay-himu l-malāʾikatu).⁶³ The behaviour of ʾAsad is what we find among the crystallised Hijazi and classical groups.⁶⁴ Farrāʾ also states that Kināna and some of Banū Saʿd b. Bakr harmonise plurals except before consonant clusters (ʿalay-him but ʿalay-humu l-qawlu); this is a pattern we find among the Kufan group.

Most significantly, however, is the fact that a large number of readers have pronominal systems that are simply not described by the grammarians at all. The Basran clusters’ phonetically conditioned disharmony, where ī and ay (generally) block vowel harmony in the plural only, whereas i triggers vowel harmony is a system which not only never gets a dialectal attribution, but does not even get acknowledged at all. Likewise, the lexical exceptions as we find with the reader Ḥamza who only blocks harmony on the words ʿalay-hum, laday-hum and ʾilay-hum and not for example on miṯlay-him or ʿalay-himā is a type of conditioning that is unique to reading traditions and outside of the scope of the grammarians. Likewise, the complex conditioning of the long forms of the plural pronouns such as found in the reading of Warš ← Nāfiʿ and Qutayba and Nuṣayr ← Kisāʾī seem to be exclusive to Qurʾānic readings.

The fact that the reading traditions do not align with dialect geography, and indeed frequently show patterns that do not belong to dialects at all is difficult to understand as a failing of the Arabic grammarians who are generally thorough, show significant agreement between each other and are quite tolerant of variation in the pronominal system. Instead, it seems that the reading traditions were not the object of study for the grammarians, but rather the result of it. The readers seem to have composed their pronominal systems quite deliberately combining different morphological options from different dialects into a single reading. Such forms must be seen as artificial linguistic practice as part of a performance register of recitation rather than forming an accurate portrayal of the varieties of Arabic that the grammarians described.

The artificiality of the readers is especially transparent with the conditioned use of the long pronouns by Warš. He only uses long pronouns when the next word starts with a hamza. However, phonetically there is no obvious reason why a hamza would cause the epenthesis of a long vowel -ū between m and hamza. In fact, within Warš’ phonological system we would rather expect the hamza to be dropped in such an environment as Warš drops word-initial hamza if the previous word ends in a consonant (Ibn al-Ǧazarī Našr II, 1282–4). Even if we admit the possibility that every variety of Arabic had access to both long and short forms of the plural pronouns as the grammarians repeatedly affirm, there is no obvious path to explain why Warš would have started using the long forms exclusively in front of hamza.

From a perspective of performance, however, this behaviour is easier to understand. Warš’ reading is well-known for its extensive use of overlong vowels (madd). In Qurʾānic recitation it is customary to pronounce any long vowel overlong when it stands in front of a hamza (malāāʾikatu, šāāʾa ḷḷāhu) or long consonant (aḍ-ḍāāllīn). There is disagreement on how long this vowel should be, and whether the overlong pronunciation before hamza should be just as long word-internally as across word-boundaries. Warš treats those two environments the same and pronounces them the longest of all readers (Dānī Taysīr 30). Moreover, he is unique in also making vowels overlong when a hamza precedes them (ʾāādam) (Dānī Taysīr 31). Using long pronouns before hamza gave him yet another opportunity to apply his excessively overlong vowels. It stands to reason that Warš’ conditioned long pronouns were a conscious choice to achieve the artistic effect of introducing even more opportunities of overlong vowels in his reading.

Such evidently artificial features of the reading traditions are by no means limited to the pronominal system, Van Putten (2018: 97–101) shows, for example, that the treatment of the loss of hamza in certain reading traditions cannot be explained as the result of regular sound change and must likewise be artificial.

Manuscripts

So far, we have looked at the pronominal systems in the literary sources, both from the perspective of the Arab grammarians and the Qurʾānic reading traditions. In both of these types of sources we find a large amount of variation, which deviates markedly from what becomes the Classical norm. However, especially in researching the Qurʾānic readers, we are stuck with literary sources reporting on the readings of people that lived at least a century before even the earliest authors. Were the eponymous readers really the most popular during their lifetimes and beyond? Where do readers outside of the canon fit in?

There is a way to answer these questions. It was already noticed by Nöldeke (1860: 328) more than 150 years ago, that Qurʾānic manuscripts frequently reflect pronominal systems that fall outside of the canonical systems, these manuscripts can give us essential insights into these central questions of the linguistic history of the Qurʾān. There are hundreds of vocalised Kufic manuscripts that span a period from around the 2^nd/8^th to the 4^th/10^th century. These manuscripts can give us a great deal of information about how the literary sources relate to the reality from the time that the readers were active.

For this research we have collected information of the pronominal system of all the Qurʾānic manuscripts we could find access to. In some cases, this meant we had many pages to check, in other cases we were only able to access a single photo of a single folio. However, very often even such fragmentary material was enough to establish several important facts about the vocalisation of the pronouns due to the fact that pronouns are, of course, rather common in the corpus. We have collected data in a corpus of 402 distinct manuscripts, ranging from single folios to near-complete muṣḥaf-s. We have made the data and analysis available on Github, and these can be accessed at https://github.com/hsidky/quranic_pronouns. In general, we have accessed these manuscripts either through the Corpus Coranicum website,⁶⁵ the Gallica website for manuscripts from the Bibliothèque nationale de France,⁶⁶ the Cambridge University Library website,⁶⁷ or the Chester Beatty Library website.⁶⁸ In the case of any other manuscript, we give the full reference in the notes of our database.⁶⁹

Vocalisation

Several studies have already looked at the vocalisation systems of Qurʾānic manuscripts, most notably Dutton (1999, 2000) and George (2015a; 2015b). Muehlhaeusler (2016) presents some additional remarks to vocalisation. While Dutton in particular has studied the pronominal systems present in these manuscripts, none of the authors have looked at which systems appear in these manuscripts in a systematic way.

The majority of the manuscripts we have looked at use the Kufic vocalisation system. This makes use of red dots to mark different forms of the pronoun. A red dot above a letter marks fatḥa, a red dot on the baseline, usually after a letter, marks the ḍamma, and a red dot below a letter marks the kasra. This system has no way to mark vowel length. Therefore, we cannot distinguish -hu from -hū or -hi from -hī, -humū from -humu or -himī from -himi.⁷⁰

For the singular, this means that it is not possible to distinguish Ibn Kaṯīr’s system from that of the other canonical readings. For the plural the lack of distinction is fairly insubstantial. While forms with short vowels are reported, these are extremely marginal and not significant. We will therefore assume that vowels on the mīm denote the long vowel by default in context, and a short vowel when it stands in front of a consonant cluster.

Some of these manuscripts use a variety of different colours at the same time to denote different readings. For example, below from Paris, BnF Arabe 350a, f. 120 v, l. 5 denoted rabbi-himi (l-hudā) in red, rabbi-hum(u) in green and rabbi-himu in gold. Occasionally, secondary colours form clear and consistent readings, but most of the time they simply seem to be designed to show off the vocaliser’s knowledge of other options, and are marked rather inconsistently. In such cases, reading the different options often leaves room for ambiguity. The green vocalisation in this manuscript, for example, could technically also be read as the very rarely reported rabbi-humi.

Another complication is that many manuscripts have a very minimalist approach to writing word-internal vowels. While case vowels before the suffix are written more often than not and final vowels are almost always written, the internal vowel of hum is not commonly written.

As such, -hī and -hū can always be distinguished, and the presence of long pronouns such as ʾantumū, -kumū, -humū can easily be distinguished from ʾantum, -kum and -hum. But manuscripts do not always give enough information to allow -himū to be distinguished from -humū nor -hum from -him.

Pronominal features

In order to carry out our analysis, we defined categories for each pronominal feature, bearing in mind the limitations of the vocalisation system we noted above. For the singular pronoun, we consider it either harmonised or unharmonised. In her study of vocalisation in several C.I⁷¹ manuscripts Cellard (2015: 368) noticed that some manuscripts that lack vowel harmony after -i in the singular do harmonise bi-hī while others do not. We quickly recognised that this indeed is the case more broadly across different manuscripts and we included it as a separate feature. Four different features of the plural pronoun were also considered. The length and vowel of the mīm is either short, long, or conditioned in the same way as the reading of Warš – that is, present only before hamza. The third person plural after an ay diphthong (i.e. ʿalay-him/ʿalay-hum,ʾilay-him/ilay-hum) is either -him or -hum. The same classification is made after an ī vowel (i.e. fī-him/fī-hum). Finally, the behaviour of the vowel after the hāʾ and mīm in harmonising environments prior to ʾalif al-waṣl (i.e. ʿalay-himu l-ʿaḏāb/ʿalay-humu l-ʿaḏāb). The following table summarises the described features:Examples of what these looks like in the vocalised Qurʾānic manuscripts are shown below.

Table

Feature	Categories	Example
Singular pronoun	Harmonised	ʾamri-hī
	Unharmonised	ʾamri-hū
bi-hī	Harmonised	bi-hī
	Unharmonised	bi-hū
Plural mīm	Short	kuntum
	Long	kuntumū
	Warš	kuntum tušrikūn/kuntumū ʾiyyāh
-ayhim	Harmonised	ʿalay-him
	Unharmonised	ʿalay-hum
-īhim	Harmonised	fī-him
	Unharmonised	fī-hum
-ay-him + ʾalif al-waṣl	-himi	ʿalay-himi l-ʿaḏāb
	-himu	ʿalay-himu l-ʿaḏāb
	-humu	ʿalay-humu l-ʿaḏāb
	-humi	ʿalay-humi l-ʿaḏāb

Analysis

While the vast majority of manuscripts contain features that are classifiable into the categories enumerated above, a few contain more complex conditioning or appear to be inconsistent.⁷² Since they only amount to a very small fraction of the total number of manuscripts, we have chosen to omit them from our analysis. In addition, because plural pronouns before waṣl and conditioning of the plural hāʾ occur relatively infrequently, we have no observations for many of the fragments surveyed. Therefore, we initially consider three features: singular pronoun harmony, bi-hī, and plural mīm for which we have the most data. The remaining features we shall return to later.

The table below shows the prevalence of the different features combinations across 287 manuscripts which contain enough data to unambiguously identify the system. Out of the twelve theoretically possible combinations, only six are actually attested.

Table

Style	Singular	Harmony of bi-hī	Form of plural mīm	Number	%
Classical	Harmonised	bi-hī	Short	149	51.9%
Long style	Harmonised	bi-hī	Long	24	8.4%
Warš	Harmonised	bi-hī	Warš	62	21.6%
Undescribed	Unharmonised	bi-hī	Long	46	16.0%
Hijazi	Unharmonised	bi-hū	Long	5	1.7%
Arabe 340d	Unharmonised	bi-hī	Short	1	0.4%

What immediately catches the eye is the fact that the combination of long pronouns with unharmonised singular pronouns, excluding bi-hī, is surprisingly common. In fact, it is almost as common as manuscripts that have a pronominal system in the style of Warš and significantly more common than manuscripts that have canonical systems with long pronouns such as ʾAbū Ǧaʿfar and Ibn Kaṯīr. While the reading(s) characterised by this specific combination of features is quite common in the manuscripts we surveyed, it is completely absent in the descriptions of the reading traditions, canonical or non-canonical. This therefore puts into strong focus just how much manuscripts have still to teach us about the reading of the Qurʾān in the early Islamic period and how small a portion of the readings that existed was documented.

The ‘Classical’ pattern is the most dominant in the manuscript record. This is to be expected, because if we define this category by vowel harmony, the harmony of bi-hī and absence of a long vowel ū after the mīm of plural forms, this covers the reading of eight of the ten canonical readers (ʿĀṣim, Ibn ʿĀmir, ʾAbū ʿAmr, Ḥamza, Kisāʾī, Ḫalaf, Yaʿqūb, and some transmissions of Nāfiʿ). It is, in fact, somewhat surprising that this system only shows up in a bit more than half of the manuscripts. Considering as it makes up 80% of the canonical readers, one would probably expect that percentage to be even higher.

We may look at the manuscripts with the Classical pattern at a slightly higher resolution, examining their treatment of the harmonised pronoun before a two consonant cluster. Here there are three patterns. The normative style with -himu, the Basran style with -himi and the Kufan style with -humu. The sample of manuscripts that attest a word in such a context along with the other three options is 46 manuscripts. Of these, 23 ( = 50%) have the Basran style, 8 ( = 17%) have the Kufan style and 15 ( = 33%) have the normative style. Thus, we see that the Basran style is the most popular, followed by the normative system, then finally the Kufan style.

It is possible to further define manuscripts that unambiguously have one of the systems of the canonical readings, but by then the number of manuscripts become so small that they are not worth tabulating. We may add that there is one manuscript that seems to have the pronominal system similar, but not quite identical to that of Yaʿqūb (Cod. Arab. 38),⁷³ five manuscripts with the pronominal system of Kisāʾī or Ḫalaf (Paris, BnF, Arabe 339, Arabe 361f, Arabe 366c, Cod. Arab. 37, CUL, MS Add. 1137) and eight manuscripts with the pronominal system that corresponds to that of Ḥamza (Paris, BnF, Arabe 330b, Arabe 6140d, Arabe 325j, Arabe 364b, Wetzstein II 1918, 1928, 1921a, 1915). All of them thus make up only a small percentage of our corpus.

Correlation to script style

There is a variety of different popular styles of writing in Qurʾānic manuscripts that have been used in at different times and in different places, whose classification has been pioneered by Déroche (1983). Comparing the five vocalisation styles as they appear in the manuscripts to the Kufic styles in which these manuscripts have been written reveals several strong correlations between writing style and vocalisation style as shown in the table below. We have only included styles which had a minimum of five vocalised manuscripts with the relevant data available for it.

Table

The presence of strong correlations between script style and vocalisation style tells us something about the manner of production of these manuscripts. As vocalisation almost invariably uses a different ink applied with a different tool from the text, it is not a priori obvious that vocalisation would be contemporary with the production of the text, or would even have been applied in the same scribal context. However, the fact that we can see clear correlations between ‘schools’ of script style and ‘schools’ of pronominal morphology reflected in the vocalisation, make it highly unlikely that vocalisation happened in a context independent of the manuscript production. These separate schools were likely one single centre of production. The manuscript would be produced in a style proper to that school, and vocalised according to the norms of those schools.

B.II and the Undescribed pronoun style

There is a strong correlation between the B.II script style and the undescribed pronoun vocalisation style. Out of 30 B.II manuscripts, 20 of them exhibit this unique pronominal system. The initial basis for the script style classification used above is Déroche (1983). However, when we look at several of the diverging B.II manuscripts that do not follow the Undescribed style, it quickly becomes clear that they deviate from the standard B.II style in a variety of ways. Here it is worth also examining Milo (2010) who describes the B.II style not just from the perspective of letter shapes, but also its script grammar. We will see that several of the putative B.II manuscripts in Déroche’s classification should probably be reconsidered as either substyles of B.II, or unclassified altogether. The result of this is a further increase in the correlation between the Undescribed pronoun style and the B.II script.

Ms. Add. 1149 is classified by the Cambridge University Library website as B.II style. While it has several features that are palaeographically similar to B.II, such as a hāʾ that straddles the baseline, there are many other features both of letter shape and script grammar that make it look closer to some subtype of D. The mīm generally is a semicircle that sits on the baseline, rather than circular straddling it (Déroche 1983: 38); the ǧīm/ḥā/ḫāʾ stacks with all consonants (Milo 2010: 282–3), and the returning stroke on the final yāʾ and qāf consists of two fully horizontal strokes (Milo 2010: 273). These are all features not typical for the B.II style.

CUL, Ms. Add. 1145 is likewise identified as B.II, but is written in a highly informal script style, which seems closer to a subtype of the D style than B.II. The medial hāʾ is triangular and does not clearly straddle the baseline, the yāʾ and qāf returns fully most of the time. However, we classify this manuscript exactly, it is clear that its production is quite distinct from the highly controlled B.II styles that have the Undescribed pronominal system.

Paris, BnF, Arabe 338b is likewise a quite informal looking manuscript, with surprisingly few lines to the page for a B.II manuscript (only 10), while more typical examples in this style have 15 or 16 lines. Nevertheless, in terms of script grammar and letter shapes, this manuscript should probably be considered a form of the B.II style.

Paris, BnF, Arabe 6982, while classified by Déroche (1983: 74, no. 43) as B.II should clearly not be classified in this group. While its circular mīm and straddling hāʾ indeed are similar to B.II, the shape of the strokes is very distinct with very sharp points, unknown for B.II. More importantly there are several features in its script grammar that clearly set it apart from typical B.II manuscripts, such as the stacking ǧīm/ḥāʾ/ḫāʾ, fully returning final yāʾ and qāf (typical of the D style), and an extremely long initial ʿayn/ġayn. Medial fāʾ/qāf frequently stands far above the baseline, while in the B.II style it typically lies on the baseline, being very similar in shape to the medial mīm of the D style (Milo 2010: 271).

The remaining manuscripts in this style that do not have the Undescribed pronominal system are all quite typical examples of the B.II style.

D.III, the Warš system, and its maghrebi origins

Another striking correlation is the very high percentage of the Warš system of pronouns present in D.III, D.I/D.III (and the less formal ‘D Commune’ style) manuscripts. George (2015b: 97) notes a certain correlation between the D.III style and what is reported in the Classical sources as the Maghrebi vocalisation system. Warš ʿan Nāfiʿ (and to a lesser extent Qālūn ʿan Nāfiʿ) quickly became the dominant reading in the Maghreb and it still is today. This is because North Africans adhere to the Maliki legal school whose eponymous founder, Mālik b. ʾAnas, explicitly stated that he considered Nāfiʿ’s reading to be sunna (communal practice).⁷⁴ Thus, the correlation between the D.III style and the Warš pronominal system might be considered further confirmation that indeed D.III is (in part) a typical Maghrebi style of Kufic.

Of course, identifying only the pronominal system of the manuscript is not necessarily a foolproof way to identify the reading tradition as that of Warš. However, a cursory examination of some of the manuscripts within this group indeed identifies several other typical features that make the Warš identification fairly certain. Paris, BnF, Arabe 325 g, Arabe 341a, Arabe 375d, Arabe 334 l, and Dublin, CBL Is. 1411 (all D.III manuscripts) all show the typical Warš feature of naql, i.e. the dropping of the hamza when it immediately followed a word that ends in an unvowelled consonants, e.g. al-ʾalbāb → alalbāb (Ibn al-Ǧazarī Našr II, 1282–4). This is marked in all of these manuscripts as a line above the vowel sign that cancels the hamza, as seen for example in Paris, BnF, Arabe 325 g:

Paris, BnF, Arabe 325 g, f. 30 v: al-ʾalbāb → alalbāb and ǧunāḥun → ǧunāḥunan (Q. 2:197-8).

Marking waṣl using a red stroke is indeed associated with the Maghrebi vocalisation style (George 2015a: 15).⁷⁵ Although not explicitly discussed by George or Dutton, Dānī’s work on dotting, al-Muḥkam, also dedicates a section to the specific features of the reading of Warš. Dānī states that naql is treated in the same manner as waṣl, i.e. with a horizontal stroke (Dānī Muḥkam: 214–5). We therefore wish to carefully suggest that the production of D.III manuscripts with the vocalisation of the Warš reading tradition indeed took place in the Maghreb/al-Andalus.

We also note that the NS.III style has a quite strong association with the Warš pronominal system. The Palermo Qurʾān (dated 372/982) – clearly Maghrebi in origin in its broader sense –, likewise is transparently identifiable as the reading of Warš, and as George (2015b: 81) observes follows a vocalisation system that is clearly identifiable as Maghrebi (using a blue dot with red line above to mark naql). It seems therefore quite possible that the NS.III style also has some association with the Maghreb.

Decrease in pronominal system variety over time

As a general expectation, with the passage of time and the eventual rise of the canonical readings, one should ultimately see a convergence upon canonical pronominal systems. And this effect does appear to bear out when looking at D and later script styles. As far as the palaeographic chronology of the D script style, Déroche indicates on the basis of waqf-s (endowment notice) that the D.I style was in full use at the beginning of the 3^rd/9^th century (Déroche 1992: 36f.). However, he notes the challenge of establishing a relative chronology for the remaining D scripts due to the limited number of dated artefacts and overlapping intervals for those which are dated. Perhaps an exception to this is the D.V subgroup where he observes an ‘external influence’ on the prototypical D style (Déroche 1992: 37).

On the basis of the distribution of pronominal systems in each script style, we would like to tentatively suggest an earlier terminus post quem for the D.I and D.IV styles relative to the rest, followed by D.III, then the remaining sub types follow. Both D.I and D.IV contain a significant number of examples of the Undescribed system to the exclusion of all other styles. Examples of the Long style system are also present, shared by D.III, which we have discussed above.

There are two D.I manuscripts with the Undescribed system worthy of note. The first is the ʾĀmāǧūr Qurʾān, commissioned by the Turkic governor of Damascus ʾĀmāǧūr (d. 265/878). This is approximately a generation prior to Ibn Muǧāhid who dies 324/936. The second is one donated to the Great Umayyad Mosque of Damascus in 298/911 by one ʿAbd al-Munʿim b. ʾAḥmad (George 2015a: 20). It too is from around the lifetime of Ibn Muǧāhid, and we can see that perhaps as late as the turn of the 4^th/10^th century, this totally undocumented system was still in use. A comprehensive investigation of reading traditions that are present in manuscripts before and after Ibn Muǧāhid to determine the exact extent of his impact remains a topic for future research.

As for an explanation of the shared pronoun similarity between D.I and D.IV, we suggest that they are palaeographically very closely related and that perhaps the salient differences between them were largely motivated by layout constraints. D.I is a large hand, with typically 6–7 lines, and occasionally even fewer per page while D.IV is a much smaller hand with a layout closer to that of B.II with generally 15 or 16 lines per page; examples of the two are shown below.

A more detailed comparison of letter shapes for D.I, D.IV, and D.III is presented below as well. The D.III ʾalif has a noticeably longer and tapered return compared to D.I and D.IV. The lām of the lām-ʾalif is also vertical. Final nūn does not have an initial flare or sharp angle in D.III and is constant thickness throughout, with a horizontal bottom stroke. Finally, initial ʿayn and ǧīm for D.III have a very fine initial taper as opposed to a diagonal edge for D.I and D.IV. While there are subtle differences between D.I and D.IV as present in the ʾalif and nūn for example, the D.IV shapes look very much like compressed versions of their D.I counterparts, as there is less room for broad curves and angular shapes in the compressed 15–16 line layout. We think that the similarities in the letter shapes of D.I and D.IV as well as the pronominal systems their vocalisations reflect, suggest that they were produced in similar time spans or regions of production.⁷⁶

Table

Letter	D.I	D.IV	D.III
ʾalif
lām-ʾalif
nūn
ʿayn
ǧīm

Moving beyond the D.I, D.IV, and D.III styles, we find an overwhelming preference for the Classical system in D.V (especially D.Vb) and NS.I. With the D.III/NS.III group above we saw that the style fairly strongly correlated with the reading of Warš ʿan Nāfiʿ which – along with other palaeographical considerations – gives the impression that at least a significant part of the manuscripts in these styles were produced in the Maghreb/al-Andalus. The overwhelming preference in these later styles towards the Classical pronominal system would suggest that they were rather produced and vocalised in the east, where readers with the classical system would become dominant, especially ʾAbū ʿAmr would gain great popularity (see footnote 19).

Also on palaeographical grounds, there is reason to connect D.Vb and NS.I to the eastern realm. What comes to be known as the ‘Persian’ Kufic style (variously called Bent Kufic, tall-Kufic, semi-Kufic and even, inappropriately Qarmatian), a script of the 5^th/11^th – 6^th/12^th century in Iraq and Persia, such as found in the British Library’s Or. 6573 has very obvious links especially with the angular NS.I style.⁷⁷ The focus on long and slanted ascenders on the ṭāʾ and final kāf are clearly anticipated in the D.Vb style, which shares clear affinities with the later NS.I style.

The New Style script is also close to early non-Qurʾānic calligraphic styles used in the Mashreq, which are quite distinct from the script that comes to dominate the Maghreb. Compare, for example, the copy of Ibn Ḫālawayhi’s Kitāb al-Badīʿ held at the Chester Beatty Library (Ar. 3051), which dates to 370/980 – the same year Ibn Ḫālawayhi died in Syria. For this reason, it is believed that the manuscript was likely copied in Syria, or at least outside of the Maghreb.⁷⁸ Had this been a Qurʾānic manuscript, it would probably have been classified as NS.I. The similarities of this manuscript to, for example, QUR286 of the Khalili collection (Déroche 1992: 142–3, no. 78) are apparent and warrant further investigation.

Before concluding this discussion, it is worth mentioning that the Long style, despite being canonical, clearly fell out of favour in the later script styles. Our survey contained over 100 manuscripts in such later script styles and not a single one contained long pronouns. This nicely illustrates that, despite the seven readings all enjoying canonical status, this did not necessarily translate to equal popularity. There are undoubtedly manuscripts not included in our survey which do contain long style pronouns, but our sample seems large enough to register a marked drop in its popularity. This is also consistent with the statement by Zaǧǧāǧ quoted earlier who notes the popularity of shortened forms and the comparatively rare lengthened long pronouns.

Conclusion

In this paper we have conducted an extensive survey of the morphology of the third person singular and plural pronouns of Arabic from different sources. The medieval grammarians in general present a consistent picture of the variation in pronominal forms. Nonetheless, we do observe a few important differences. Apart from Sībawayhi and his student ʾAḫfaš, the remainder do not prefer that the pronoun be lengthened after a closed syllable (e.g. min-hu vs. min-hū). Mubarrid appears to mark the beginning of strongly endorsing what becomes the normative Classical Arabic system. He states that harmonised singular -hi/ī and plural -him is the best practice, as is the use of short forms after heavy syllables and the use of short plural pronouns. This system is identical to normative Classical Arabic.

Moving on to the reading traditions, we find them to be in stark contrast to what the grammarians describe. The readers have a vast number of highly specifically conditioned environments in which allomorphs of the pronouns appear, probably most clearly exemplified by the transmitters Qutayba and Nuṣayr from Kisāʾī but likewise present in the conditioning of harmonised forms of the pronouns in the Basran group of readers. But the grammarians never describe such systems. Their model is quite clear: either you have vowel harmony of -hū and -hum after i, ī and ay (the Najdi practice) or you do not (the Hijazi practice). Compare this to Yaʿqūb, who only applies vowel harmony to the singular -hū in all those environments, whereas for the plural -hum, he only does it after ī and ay but never after i. Such a rather unusually conditioned system is never described by the grammarians. This would perhaps not be altogether surprising for grammarians that are roughly contemporary with the early readers, but it is striking that even in the centuries that follow, later grammarians never feel the need to describe such systems, nor are such reading systems ever explained as a ‘dialect’ of one tribe or the other. All this suggests is that what the medieval scholars believed to be ʿarabī was something self-evidently different from qirāʾa.

That being said, the readers clearly do draw upon variation in morphology and theoretical concepts that they describe. Warš could not have decided to use long forms of the plural pronouns, had he not known about their existence, and he could not have decided to use them only when hamza followed if he had had no concept of what a hamza was.

Besides the disconnect between the grammarians and the reading traditions, there is yet another disconnect that deserves attention. The literary tradition transmits a large number of readings with vastly different pronominal systems, some of which retain quite complex conditioning. While some of the details between the different readers are lost in the somewhat limited tools that vocalised Qurʾānic manuscripts have to express them, it is still clear that there is a major gap between what we find in the literary tradition and what we find in vocalised manuscripts.

While three of the ten canonical readers have harmonised pronouns and long plural pronouns (the ‘Hijazi cluster’), this system is almost entirely absent in the manuscript record; this is despite surveying over 400 vocalised manuscripts. Whenever a reading has long plural pronouns, it overwhelmingly lacks vowel harmony in the pronouns (with the exception of bi-hī). Other canonical systems like that of Yaʿqūb and others from the ‘Basran cluster’ are barely found at all. In light of these absences, one comes to wonder to what extent the 7 (and later the 10) readings that came to be accepted as canonical were very representative of the popular recitation of the Qurʾān.

Another point of disconnect is the striking fact that after the Classical and Warš systems, the system that lacks vowel harmony in all words except bi-hī along with long plural pronouns is extremely popular and even more common than ʾAbū ʿAmr’s, Ḥamza’s or Kisāʾī’s systems. Despite this extraordinary popularity, this system is not known in the literary tradition at all, not among the canonical readers nor among the non-canonical ones. This lack of description cannot be seen as the result of this tradition simply having died out by the time the readings start being described. We have at least two dated manuscripts from the lifetime of Ibn Muǧāhid, namely the Qurʾān of ʾAmāǧūr dated shortly before 262/876 and the Qurʾān of ʿAbd al-Munʿim dated before 298/911, that have this pronominal system. It seems that the description of this reading was decided against despite its apparent popularity in the manuscript record.

During our analysis of the manuscripts, we found a correlation between dotting and script style. For example, we find that two thirds of the B.II manuscripts surveyed contain the undescribed pronominal system discussed above. The Warš system of pronouns is also frequently present in the D.III, D.I/D.III, and D Commune style manuscripts. The fact that there is a correspondence between script style and vocalisation indicates that the dotting is contemporary with the manuscripts.⁷⁹

Despite the fact that we carried out the most extensive survey of vocalised manuscripts to date, we still maintain that our findings are to be interpreted with caution. Many of the 400 manuscripts surveyed are of unknown provenance, while others are known to have been acquired from Syria, Egypt, or North Africa. The number of manuscripts surveyed does not matter if they are not representative of the overall population of manuscripts during those time periods. Recent information on the manuscript collections of Iran may prove this to be the case. While right now we only have very limited access to these manuscripts, a preliminary examination suggests that they have different vocalisations despite belonging to the standard script style classifications. Time will tell whether we will see a larger concentration of the pronominal systems of the readers which were nearly entirely absent from the manuscripts we surveyed. This however, does not change our major findings; namely that the grammarians do not describe the systems used by the readers, and that manuscripts contain pronominal systems not described by either the grammarians or qirāʾāt works. Thus, the mystery remains as to how and why pronominal systems were used into the 4^th/10^th century that evaded characterisation in the vast literature presented above.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Sadly, the highly insightful and detailed discussion of the reading tradition that appears in vocalised manuscripts that Nöldeke had examined, for whatever reason, did not end up in the final updated version of Geschichte des Qorâns.

² While Sībawayhi is not explicitly mentioning the reader by name, we will see below that one of the seven canonical readers, Ibn Kaṯīr, adopts this rule. It is clear that in Sībawayhi’s time, it was apparently unproblematic to call this reading inferior.

³ See Fischer (2002 [1972]:142), surprisingly Wright (1996 [1859-62]:100–1) only mentions the short versions of the pronominal forms, an option that is universally considered unacceptable by all medieval grammarians.

⁴ Rabin (1951:99) cites Sībawayhi as saying the people of the Hijaz said bi-ġulāmi-hū, bi-ġulāmi-hum. These examples are not mentioned by Sībawayhi. In fact, Sībawayhi does not mention the unharmonised short form -hum explicitly at all, only -humū.

⁵ This is indeed the reading as it is reported for Kisāʾī. See the discussion in the section below.

⁶ The edition only writes min-humū with a wāw to denote lengthening of the mīm. The remaining words are only marked with a ḍamma. Although this may convey a difference in vowel length, we do not believe this to be the case here. However, since the edition is based on a lone manuscript and there is no way to corroborate the text, we reproduce it as such.

⁷ No Medinan reader is reported to lengthen those forms, although the Meccan reciters such as Ibn Kaṯīr and Ibn Muḥayṣin do (see discussion under Reading traditions below).

⁸ Note ʾAḫfaš’ striking argumentation here that highlights the disconnect between Classical Arabic orthography and the language that it is supposed to represent.

⁹ Such an assimilation is still found in ʾAbū ʿAmr’s reading, see the discussion under Reading Traditions.

¹⁰ This should be ʾazd al-sarāt, but the edition actually has ʾasad al-sarāt.

¹¹ This remark by ʾAḫfaš concerning regular speech at least provides some evidence that he was not (exclusively) sourcing their information from Qurʾānic readings.

¹² Mubarrid does not explicitly identify these base forms as being Hijazi here, unlike most other authors. But elsewhere in the book it is identified as specific to the people of the Hijaz (Mubarrid Muqtaḍab I, 175).

¹³ Note that this, and his subsequent remark, explicitly suggests that a reading may be valid not only through authentic transmission but also through broad popularity of a reading. This is a striking deviation from what comes to be the orthodox view of the acceptability of a reading which requires authentic transmission.

¹⁴ This statement is striking, as this seems to imply that to Zaǧǧāǧ only the harmonised form ʿalay-hi is Qurʾān. But as we will see in the next section, ʿalay-hī and ʿalay-hu are both reported for readers of the Qurʾān, even among the canonical readers.

¹⁵ For fī-h Naḥḥās is undoubtedly referring to ʾAbū ʿAmr’s reading Q. 2:2 fī-h=hudan ([there is no doubt] in it, a guidance [for the mindful]), with major assimilation across a short vowel.

¹⁶ Such unlikely forms like ʿalay-humī should be interpreted as paradigmatic fabrications, which some readers or grammarians may have experimented with, rather than actual attested forms.

¹⁷ ʿUkbarī also has a work dedicated to the ʾiʿrāb of aberrant (šāḏḏ) readings where he simply reproduces the same discussion.

¹⁸ This is indeed the case. All jussive/imperative weak verbs that take the pronominal suffix -h are read without a final vowel among at least one of the ten canonical readers and many of them are read with a short vowel by some. There seems to be no patterns as to when any one reader decides to read a form vowelless, shortened form or long form (Ibn al-Ǧazarī Našr II, 974–92).

¹⁹ This almost certainly is a reference to the popularity of ʾAbū ʿAmr’s reading, which had spread throughout much of the Medieval Muslim world. Ibn al-Ǧazarī mentions, ‘the reading that people employ today in Syria, Hijaz, Yemen, and Egypt is the reading of ʾAbū ʿAmr. You would be hard pressed to find anyone teaching Qurʾān except according to his reading, especially when it comes to farš (i.e. specific variants), though they might make mistakes in the ʾuṣūl (i.e. general principles)’ (Ibn al-Ǧazarī Našr I, 265).

²⁰ Not to be confused with Abū l-Barakāt b. al-ʾAnbārī who dies nearly two centuries later.

²¹ Arabic grammatical theory conceives of the long vowel ī as /iy/, hence fī-hi is considered as having a consonant preceding the -hv suffix.

²² There is the question of whether the popularity of a reading among specialists was shared by the general public. For an initial inquiry on that subject, see Saleem (2019).

²³ It is worth mentioning here that, despite the status of these readings as šāḏḏ, they are, as a whole not non-ʿUṯmānic. While many early reports of companions of the prophet did not follow the standard consonantal skeleton of the now standard ʿUṯmānic Text, the vast majority of the non-canonical readings – especially those of which we have complete transmissions – are by-and-large in agreement with the ʿUṯmānic text. ʾAʿmaš and Ḥasan al-Baṣrī are notable exceptions, but even they follow the ʿUṯmānic text more often than not.

²⁴ See for example Ibn Ḫālawayhi’s Muḫtaṣar which is in fact an edited version of marginal notes to his book al-Badīʿ on the canonical readings and the reading of Yaʿqūb. Ibn Ǧinnī’s work al-Muḥtasab is an example of another work dedicated to non-canonical readings.

²⁵ One example is Q. 35:28 where ʾAbū Ḥanīfa allegedly read: innamā yaḫšā ḷḷāhu min ʿibādi-hi l-ʿulamāʾa (God only fears the scholars among his servants) (Ḏahabī Ṭabaqāt 574 #435).

²⁶ Note that, unlike Sībawayhi’s description, syllables closed by a consonant always have short -hu among the Qurʾānic reading traditions.

²⁷ The common link of the transmission of his reading appears to be Ḥaywa b. Šurayḥ (d. 224/839) (Ḫuzāʿī Muntahā 231).

²⁸ Ḏahabī Ṭabaqāt 46 #25.

²⁹ Ḏahabī Ṭabaqāt 59 #34.

³⁰ Ḏahabī Ṭabaqāt 69 #35.

³¹ Right to left arrow indicates transmission from the individual on the right to the individual on the left.

³² This is based on the manuscript Ar. 3051, folio 7 v, situated at the Chester Beatty Library, a manuscript that was finished the year Ibn Ḫālawayhi died (d. 370/980) on which the only print edition is based. The edition changes this transmission to the canonical ʿalay-himū without comment (Ibn Ḫālawayhi Badīʿ 47).

³³ Ḏahabī Ṭabaqāt 89 #41.

³⁴ Not present in the edition but likely the case due to other mentions of Ibn ʾAbī Yazīd transmitting from Ibn Muḥayṣin.

³⁵ Reported as being read with harmony in Q. 70 (al-Maʿāriǧ), but this word does not occur in this Sūra. A very similar sequence of words occurs in Q. 80:34–37. This is likely a mistake on Rūḏsbārī’s end.

³⁶ Ḏahabī Ṭabaqāt 75 #36.

³⁷ Ḏahabī Ṭabaqāt 80 #37.

³⁸ Ḏahabī Ṭabaqāt 49 #29.

³⁹ Ḏahabī Ṭabaqāt 83 #39.

⁴⁰ Ḏahabī Ṭabaqāt 91 #42.

⁴¹ Ḏahabī Ṭabaqāt 112 #49.

⁴² Ḏahabī Ṭabaqāt 104 #45.

⁴³ Ibn Muǧāhid records that Qālūn said of Nāfiʾ that he ‘would not fault applying wāw to the [plural] mīm’. Ibn Muǧāhid takes this to mean that Nāfiʿ personally did not read with this length.

⁴⁴ Not the eponymous reader.

⁴⁵ Ḏahabī Ṭabaqāt 132 #59.

⁴⁶ Ḏahabī Ṭabaqāt 149 #65.

⁴⁷ Ḏahabī Ṭabaqāt 252 #145.

⁴⁸ Ibn Mihrān is not explicit in defining what constitutes a ‘long’ word.

⁴⁹ Ḏahabī Ṭabaqāt 196 #102.

⁵⁰ Ḏahabī Ṭabaqāt 175 #79.

⁵¹ Ibn Ḫālawayhi (Muḫtaṣar 47) is an earlier source, but is not very detailed. He only reports that Yaʿqūb reads ʿalayhum like Ḥamza with no further specification for either of the readers.

⁵² In his Mabsūṭ, Ibn Mihrān makes the distinction between mā bayna ʾaydī-hum which is read unharmonised, and min bayni ʾaydī-him which was harmonised. He then says that similar instances to the latter are read with kasr (Ibn Mihrān Mabsūṭ 87–8). The only apparent difference is the relative pronoun mā and preposition min change the syntax of bayn. This distinction does not make it into the canonical transmission.

⁵³ Ḏahabī Ṭabaqāt 199 #104.

⁵⁴ Ḏahabī Ṭabaqāt 245 #139.

⁵⁵ Ḏahabī Ṭabaqāt 258 #156.

⁵⁶ This is reported in al-Kāmil as فيسكن الهاء من غير إشباع. This is clearly a corruption of فيكسر since the comment regarding shortened vowel length (من غير إشباع) would make no sense otherwise. It is reported correctly in Nawzawāzī Muġnī 372–3.

⁵⁷ Ḏahabī Ṭabaqāt 42 #23.

⁵⁸ Ḏahabī Ṭabaqāt 128 #54.

⁵⁹ Ḏahabī Ṭabaqāt 81, #38.

⁶⁰ Ḏahabī Ṭabaqāt 126 #58.

⁶¹ In the manuscript of Ibn Ḫālawayhi’s Kitāb al-Bādīʿ with the collection of Šawāḏḏ written in the margin held at the Chester Beatty Library, written in Ibn Ḫālawayh’s death year (370 AH), the manuscript actually reads ʿalay-humi for both ʿAmr b. Fāʾid and Ḥasan al-Baṣrī. This is presumably a copyist error, but considering the age of the manuscript, we cannot dismiss it altogether.

⁶² Interestingly, the reading of Zuhrī, who was a Qurašī, is the closest (although not identical) to the description given by Farrāʾ, with unharmonised singulars, unharmonised plurals before consonant clusters, and a lack of harmony with the words ʿalay-hum, ʾilay-hum, and laday-hum.

⁶³ Implicit in this is that the remaining tribes, i.e. Qays and Tamīm, maintain harmony, i.e. ʿalay-himi l-malāʾikatu.

⁶⁴ It is worth nothing that ʿĀṣim, who was a mawlā of ʾAsad has a pronominal system that is identical to what is reported for ʾAsad.

⁶⁵ http://www.corpuscoranicum.de/.

⁶⁶ http://gallica.bnf.fr/. The naming conventions of the manuscripts in certain volumes is based on Déroche (1983).

⁶⁷ https://cudl.lib.cam.ac.uk/collections/islamic/1.

⁶⁸ https://viewer.cbl.ie/viewer/browse/-/1/SORT_TITLE/DC:islamiccollection/.

⁶⁹ For the locations of some of these libraries, we have used several abbreviations. BnF = Paris, Bibliothèque nationale de France. StaBi = Berlin, Staatsbibliothek, CBL = Chester Beatty Library, CUL = Cambridge University Library.

⁷⁰ There are a few manuscripts that use an elongated red line (e.g. Paris, BnF Arabe 329d) to mark the madd, that is the overlong articulation of a long vowel before a hamza. If such a vowel is written above a pronominal suffix, we can confirm the vowel is long. But examples of this vocalisation are rare in the Kufic manuscripts, and encountering the right phonetic environment for it is even rarer, so this is not usually informative.

⁷¹ For the classification of the Qurʾānic script styles see Déroche (1983).

⁷² For example, Paris, BnF, Arabe 349c has an unharmonised ʾilay-hu but harmonised ʾaḫīhi.

⁷³ The third person plural pronoun in harmonising environments is -humu, not -himi.

⁷⁴ Ḏahab#x12B; Ṭabaqāt 105.

⁷⁵ George also reports that ʾalif al-waṣl/ṣila in Medina and Maghreb/al-Andalus may be marked by a small red circle. However, this appears to be only a theoretical suggestion by Dānī who states ‘and if it was marked by a small circle, that would be good too because the circle is used by the Medinans and their dotters as a marker of sukūn and a letter which is dropped from speech. This is one of those subtleties that is lost on keen people, let alone idiots’ (Dānī Muḥkam 211).

⁷⁶ It is also worth appreciating that D.IV has an extremely high percentage of the Undescribed system, this pronominal system is almost as common in D.IV as it is in the B.II style. While these two styles are clearly distinct, it is striking to note that both styles typically have a similarly high number of lines to the page; these two features may suggest a closer connection between their production contexts.

⁷⁷ The style is so close to NS.I that Arberry in his catalogue of the manuscripts at the Chester Beatty Library in fact combines them both under the header of ‘Persian’ Kufic manuscripts. The later ‘Persian’ Kufic manuscripts, still are fairly distinct by having much more ornamental shapes of the vertical ascenders, and extremely high ascender height compared to the general letter shape height (Arberry 1967:10–4).

⁷⁸ The copyist’s nisba is Rūḏbārī, which is located in modern day Iran and is also suggestive of an Eastern provenance.

⁷⁹ The correlation between script and vocalisation would also seem to make intuitive sense of the rather surprising fact that latter Abbasid script styles consistently have less consonantal dotting than the earlier Hijazi and early Kufic styles. The lack of dotting seems to be explicitly intended to give room for the coloured diacritical dotting to remove these ambiguities, rather than leaving it to the consonantal skeleton to do so.