Chapter 6 Classicized Hijazi: imposition of the hamzah

In: Quranic Arabic
Marijn van Putten
Search for other papers by Marijn van Putten in
Current site
Google Scholar
Open Access



6.1 Introduction

In 2020, Ahmad Al-Jallad put forward the bold hypothesis that Classical Arabic as we know it today is not a single linguistic system but rather the outcome of a complex interaction between Old Hijazi, i.e. the language of the Quran and early Islamic Arabic on the one hand and the poetic register of the Qaṣīdahs of the Maʕaddites on the other (Al-Jallad 2020b, 69 ff.). This suggestion is very much in line with what we have argued for so far in the previous chapters and previous studies (van Putten 2017a; 2017c; 2018; 2019b; van Putten and Stokes 2018).

While Quran today is read with a certain amount of linguistic variation, these reading traditions, despite their variation still agree on several central features such as the retention of the (in most environments), and a full case inflection with final short vowels and tanwīn both of which appear to have been absent in the original form of Quranic Arabic as reflected by the QCT. To get from the language of the QCT to the language(s) used in recitation, this language has to have been ‘classicized’ over time. This claim will, of course, bring to mind the work of Karl Vollers (1906), who famously claimed that the Quran was composed in the Hijazi vernacular and only later reworked by the Arab grammarians towards Classical Arabic. His hypothesis was criticized by many, and few authors have taken his book particularly seriously but even fewer have seriously answered his arguments in a coherent way.

Vollers (1906, § 39–43) considered the original Hijazi vernacular—and by extension the language of the Quran—to have lacked all forms of case inflection. This is not in keeping with the primary source material. First of all, the QCT very likely reached closure during the reign of ʕuṯmān, around 650 CE (Sidky 2021; Sinai 2014a; 2014b; van Putten 2019c). This is clearly before the development of grammatical theory, and reworking of the text by Arab grammarians towards a literary standard which gets established by the Arab grammarians over a century later is chronologically no longer defensible. Much of his argumentation requires us to assume that the consonantal text was changed in the decades after 650 CE. With this new material evidence, this part of his argumentation has lost most of its explanatory power. Likewise, evidence adduced from canonical and non-canonical readings alike only tells us something about the linguistic variation that was considered acceptable as part of the ʕarabiyyah, and nothing about the language of the Quran itself as it is reflected in the QCT. The only argument that relies on the rhyme of the Quran is his argument for the absence of the indefinite accusative (§ 42). He suggests that the indefinite accusative ending was invisible to rhyme, but this is clearly not the case and the presence of this vowel is in fact essential for the choice between certain otherwise identical formulae (e.g. ‮ان الله عليم حكيم‬‎ versus ‮ان الله كان عليما حكيما‬‎, see van Putten and Stokes 2018, 145 f.). Any expression of case that is explicitly present in the QCT is certainly part of the language that the QCT was written down in, and likely (and sometimes demonstrably so) present in the language of the original composition as evidenced by the rhyme.

However, those arguing against Vollers have frequently taken the argument to the opposite extreme: any sign of any case at all must mean that Quranic Arabic had full case inflection exactly how the Arab grammarians present it, with full case/mood inflection and tanwīn. But this conclusion is not borne out by the evidence either. Between a stage of full case inflection, which must certainly be reconstructed for Proto-Arabic (Al-Jallad and Putten 2017), and no case at all, there must be a whole spectrum of case systems that were in the process of losing it. Already in the pre-Islamic period there were clearly varieties that had lost their case inflection to various degrees.

While the loss of case and mood has often been seen as a catastrophic event that very rapidly, or instantly changed the language from its Old Arabic stages to its Neo-Arabic stage (e.g. Ferguson 1959; Blau 1977; Versteegh 1984), discoveries of the past decades in Arabic dialectology and especially of recent years in Old Arabic epigraphy have made it clear that such a simplified scenario cannot account for the variation that we see. Safaitic, for example, seems to have only marked the accusative case, while not marking tanwīn at all, centuries before the rise of Islam (Al-Jallad 2015, 69). Also, the case system of Nabataean Arabic, lacks tanwīn in the earliest period but still seems to have a tripartite case inflection. Only later this case distinction seems to be lost, exchanged with an invariable ⟨-w⟩ (“wawation”)—a trace of the original nominative (Al-Jallad forthcoming; Diem 1973). Likewise, the case system present in the Damascus psalm fragment is almost completely on its way out, despite being written around the same time as the activities of the early Arab grammarians (Al-Jallad 2020b). However, if we would follow the indications of the Arab grammarians, we would never know such varieties existed at all, as these clearly fell outside of their normative framework of proper Arabic.

It is now clear that certain forms of pre-Islamic Arabic with clearly archaic features—often more archaic than what we find in Classical Arabic—existed which, nevertheless had a different and frequently more reduced ʔiʕrāb/tanwīn system than Classical Arabic. It is therefore not a given that this system, whose linguistic reality is proven by the rhyme and meter in pre-Islamic poetry, could be imposed onto the language of the Quran, even if it is present in all the Quranic reading traditions.

Van Putten & Stokes (2018) argued that the Quran did not completely lack the Proto-Arabic case system as Vollers suggests, but rather had a transitional system where final short vowels and nunation had been lost (where an had become ā). Case expressed by long vowels was generally retained, as well as case vowels for nouns in construct. In other words, we have argued and adduced evidence that case was only retained in places where the QCT actually reflects it. Examples usually invoked to prove that the case system must have operated, tend to not counter such a system particularly well. In countering Vollers’ suggestion, for example Fück (1950, 2) cites the following examples which he considered ambiguous had case been lost. All of the examples would be unambiguous in the case system that Van Putten & Stokes reconstruct. The examples of Fück are given below along with the likely form they may have taken in Quranic Arabic as I would reconstruct it:

‮انما يخشى الله من عباده العلموا‬‎ /innamā yaxšē ḷḷāh min ʕibādi-h al-ʕulamō1/ (Q35:28)

‘Only the knowledgeable among his servants fear God’

‮واذا حضر القسمه اولوا القوربى‬‎ /wa-iḏā ḥaḍar al-qismah ulū l-qurbē/ (Q4:8)

‘And when relatives, at the time of division, are present’

‮ان الله برى من المشركين ورسوله‬‎ /anna ḷḷāh bariyy min al-mušrikīna wa-rasūlu-h/ (Q9:3)

‘That God is free from the idolaters, as is his messenger’

‮واذ ابتلى ابرهم ربه‬‎ /wa-iḏ ibtalē ibrāhām rabbu-h/ (Q2:124)

‘And when his lord tried Abraham’

The first two of these are in fact distinguished by the rasm, and certainly did not present any problem to the understanding. The other two would not be ambiguous if, as we have argued, case vowels had been retained in construct. One should note, however, that even if such phrases would be fully ambiguous, pragmatically such phrases hardly ever pose true ambiguity—it is unlikely that anyone would think that it is God who fears the servant in Q35:28, for example. The very fact that Classical Arabic writing manages to communicate the intended meanings with a writing system that generally does not express case, should make it clear that such ambiguities can be resolved to a large extent through pragmatic considerations.

Moreover, many extra-linguistic hints such as intonation and pause, which are likely to have played a role in the original composition of the Quran, are systematically erased almost completely in Quranic recitation. These too would have helped with the resolution of ambiguities, even with a strongly reduced case system. It is therefore difficult to accept unusual word order to hold much weight as an argument for a full case system, and evidence for its presence or absence needs to be found elsewhere.

Starting from the linguistic situation where and final short vowels and nunation were lost completely, one would naturally expect that at times the Quran had been imperfectly classicized towards a variety that did have these features. It has, on multiple occasions, been claimed that the Quran cannot have been classicized for the exact reason that there are no such traces of imperfect classicization, as exemplified, for example by Blau saying that “the total lack of Neoarabic and pseudo-correct features in the Koran establishes a linguistic situation in which the differences between the literary and spoken language could not have been too far-reaching” (Blau 1977, 15). I agree with Blau that the Arabic of the Quran was probably close to the vernacular of the Hijaz, and that little to no reworking has been undertaken on the consonantal text. However, this implies that the language of the Quran did not have hamzah, and indeed that it had a reduced case/tanwīn system. Blau seems to admit the possibility that Quranic Arabic had a somewhat reduced system that had lost (at least) word final -i (pg. 15 f.), but does not commit to a strong opinion on what this system may have looked like.

However, we frequently find evidence in the Quranic reading traditions that these texts have been grammatically reworked by its reciters. Pseudo-correct features that clearly point to conscious and artificial tampering with the language of recitation frequently appears in them. In fact, Vollers found many examples of this in his magnum opus, although many of his critics seem to have missed these points, and have rather chosen to attack his admittedly much weaker argumentation in favour of his ‘caseless Quran’.

As I see it, there are three main systematic features that differentiate Quranic Arabic as it can be reconstructed from the QCT, and how it appears in the Quranic reading traditions. The first, and the most widely admitted difference is that Quranic Arabic seems to have lost the hamzah entirely, something that is obvious from the orthography and can be clearly demonstrated from Quranic rhyme (van Putten 2018). In this chapter we will show that the pattern of both the pseudocorrect presence and absence of hamzah frequently occurs in the Quranic readings, clearly showing that later philologists have inserted the hamzah into the recitation of the Quran and were not always successful in doing so with regard to the placement that would be expected from its etymology.

The second feature, is the quintessentially Classical Arabic feature, namely the system of ʔiʕrāb and tanwīn, which the language of the QCT appears to have largely lost (van Putten and Stokes 2018). In chapter 7, I will show that to the Quranic reciters, placement of ʔiʕrāb and tanwīn was a highly theoretical undertaking, not one that unambiguously stemmed from its prototypical recitation and composition. Within this theoretical framework, there are also occasional cases where the reciters fail to fully apply the final short vowels in a manner that would be expected, yielding forms without final short vowels, where we would have expected them.

The third feature, is the retention of a phonemic distinction between the two ʔalif maqṣūrahs, the one written with ʔalif reflecting /ā/ and the one written with yāʔ reflecting /ē/, a distinction clearly reflected in the Quranic rhyme (van Putten 2017a). This feature is different from the previous two. While all readings have, to a greater or lesser extent, retained a good number of cases of etymological hamzah, and all of them in principle reflect the Classical Arabic system with ʔiʕrāb and tanwīn, this last feature is a topic of disagreement among the canonical readers. While normative Classical Arabic eventually opts for a merger of these two sounds, the Quranic reading traditions give ample evidence for an original distinction between the two sounds. This is found regularly in the readings of Ḥamzah, al-Kisāʔī, Xalaf and Warš ʕan Nāfiʕ. It is self-evident that not both retention and loss can be true simultaneously for the original composition of the Quran, and rhyme clearly favours the readings that retain this distinction. I will therefore not discuss this feature in more detail in these chapters.

6.2 Pseudocorrect Hamzah

In § 5.2 and Van Putten (2018) we have argued that the language of the QCT lacked a hamzah altogether and that the reading traditions eventually classicized Quranic Arabic. Van Putten (2018, 98–101) showed already that the reading traditions treat the hamzah rather inconsistently. In phonetically identical environments sometimes the hamzah is lost while other times it is not, occasionally based on grammatical principles, other times seemingly by rhyme. The fact that the Quranic readings fail to undergo regular sound changes clearly suggests that the readings are not natural language, but rather a mixed literary register (see § 3.3).

Evidence for a transition from a Hijazi hamzah-less pronunciation of the Quran, as confirmed by the rhyme and orthography, towards a more classical system can be seen by the presence of pseudocorrection of the hamzah in the Quranic reading traditions. Indeed, we would expect to see the application of hamzah where it should have never appeared etymologically, and likewise failure to insert the hamzah where we would etymologically expect it. Cases of both types of pseudocrrection can indeed be found in the reading traditions (as well as in Classical Arabic). This is a strong indication that Quranic Arabic originally lacked the hamzah and that it was only later artificially inserted, as it became fashionable for proper Arabic to have a hamzah.

There appears to be a historical memory of this transition taking place in the beginning of the second Islamic century, at least for Medina, as Ibn Muǧāhid (60) reports that Qālūn said: kāna ʔahlu l-madīnati lā yahmizūna ḥattā hamaza bnu ǧundabin, fa-hamazū mustahziʔūna, wa-stahziʔ “The people from Madīnah used to not apply the hamzah until [Muslim] Ibn Ǧundab (d. 130 AH/747 AH) applied the hamzah. From then on they applied the hamzah to mustahziʔūna and istahziʔ2

6.3 Hamzah among the Quranic Readers

Before we discuss the cases of pseudocorrect hamzah it is worth discussing the generalizable treatment of the hamzah in the different canonical reading traditions, as this way we are better able to appreciate the instances when readers deviate not just from the imagined classical Arabic norm, but also from their own norms.

The majority of the Quranic readers regularly retain the hamzah in most environments. Readers such as ʕāṣim, Ibn Ḏakwān ʕan ibn ʕāmir, Qālūn ʕan Nāfiʕ, Ibn Kaṯīr, al-Kisāʔī, Xalaf and Yaʕqūb by and large retain the hamzah in all positions. That is: in pre-consonantal position, post-consonantal position and in intervocalic positions. The only position where all readers agree that etymological hamzah is to be dropped is in sequences of two hamzahs, where the first one is followed by a vowel and the second by a consonant, within a single word, e.g. ʔaʔmuru-hūʔāmuru-hū ‘I order him’ (Q12:32).

The remaining readers adhere to several general principles of the dropping of the hamzah. In the following discussion we will only discuss cases of pseudocorrect hamzah that cannot be explained by the general rules of the readings.

ʔabū Ǧaʕfar drops each pre-consonantal hamzah, with compensatory lengthening, e.g. muʔminūnamūminūna, šiʔtašīta, yaʔkuluyākulu. He likewise does the same for word-final vowel + ʔ sequences: iqraʔiqrā, nabbiʔnabbī (Ibn al-Ǧazarī, § 1466). Besides this he also regularly shifts the sequences iʔūna, iʔīna and iʔū to ūna and ī(na) respectively, e.g. mustahzūna (Q2:14; Q15:95), ʔa-tunabbūna (Q10:18), muttakūna (Q36:56), fa-mālūna (Q37:66; Q56:53), al-munšūna (Q56:72), al-xāṭūna (Q69:37), al-xāṭīna (Q12:29, 91, 97), al-mustahzīna (Q15:95), muttakīna (Q18:31), yuṭfū (Q9:32). He has a single exception to this: xāsiʔīna (Q2:65; Q7:166) (Ibn al-Ǧazarī, § 1496). He would also drop the hamzah whenever it stood in the sequence uʔa, where ʔ was the first root consonant, e.g. yuʔaddihīyuwaddihī, yuʔāxiḏuyuwāxiḏu (Ibn al-Ǧazarī, § 1485).

ʔabū ʕamr has the option to drop prescononsantal hamzah, or to conservatively keep it (Ibn al-Ǧazarī, § 1472–1474). However, even with the option to drop the hamzah, ʔabū ʕamr would not drop it if hamzah was root-final, and in the apocopate or imperative. This is not just in word-final position such as našaʔ and tasuʔ, but also on morpheme boundaries such as ʔanbiʔ-hum and ʔarǧiʔ-hu, where the hamzah is pre-consonantal within the same word (Ibn al-Ǧazarī, § 1475).

Warš ʕan Nāfiʕ has two main treatments. In the transmission path of al-ʔazraq the rule is that Warš drops pre-consonantal and intervocalic hamzah, but only if it is the first root consonant. Hence: muʔminunmūminun and yaʔxiḏuyāxiḏu, and yuʔaxxiruyuwaxxiru but not biʔsa, ǧiʔta, or yašaʔ (Ibn al-Ǧazarī, § 1471).

The other path of transmission of Warš, that of al-ʔaṣbahānī, has a principle that is closer to that of ʔabū Ǧaʕfar. He drops any preconsonantal hamzah, regardless of the position in the root. He, however, has a list of exceptions to this general rule, causing him to retain significantly more hamzahs than ʔabū Ǧaʕfar. These exceptions are: baʔs, baʔsāʔ, (al)-luʔluʔ, riʔyan, kaʔs, ar-raʔs, ǧiʔta (and other forms of the verb such as ǧiʔnā-hum), nabbiʔ (and other apocopates derive from that root), qaraʔta (and other suffixed forms of the verb); hayyiʔ/yuhayyiʔ and tuʔwī/tuʔwī-hi (Ibn al-Ǧazarī, § 1469). Like Warš in the path of al-ʔazraq, he also drops any word-internal intervocalic hamzah when it is the first root consonant (Ibn al-Ǧazarī, § 1485).

Both transmissions of Warš are in agreement that post-consonantal hamzah is dropped if there is a word boundary between the word-final consonant and the next word, or if the word preceding the hamzah is the definite article. Thus, qad ʔaflaḥaqadaflaḥa and al-ʔarḍualarḍu (Ibn al-Ǧazarī, § 1541).

Ḥamzah and Hišām ʕan ibn ʕāmir both have conservative hamzah treatment, but make a special exception in pause. Hišām drops all word final hamzahs in pause (after dropping the final short vowels), whereas Ḥamzah drops all hamzahs in pause. That is, words like yaʔkulu, yasʔalu, al-luʔluʔi, as-samāʔu and al-ʔarḍu would be pronounces yākul, yasal, as-samā,3 al-lūlū and alarḍ in pause (Ibn al-Ǧazarī, § 1541).

6.4 Pseudocorrect Presence of Hamzah

In several cases throughout the Quran, we find examples where readers have a hamzah where clearly none was ever present etymologically. Such pseudocorrections fall into three types. First, some words can be shown to behave irregularly within the system of the ʕarabiyyah in the appearance of the hamzah. Second, some words are loanwords from Hebrew or Aramaic where the hamzah is absent, but has been inserted into the Arabic form. Finally, there are several inherited Semitic words which on comparative Semitic grounds can be shown to have never had a hamzah in their stem but have acquired them in the readings.

6.4.1 Ḍiyāʔ → ḍiʔāʔ

Qunbul ʕan Ibn Kaṯīr pronounces the verbal noun of ḍāʔa (√ḍwʔ) not as ḍiyāʔ, as one would expect for a fiʕāl pattern of such a root, but as ḍiʔāʔ (Ibn al-Ǧazarī § 1534). This is clearly pseudocorrect: the root consonant w (which appears in the also Quranic ḍawʔ) is simply expected to shift to y after i (for example II-w roots with CiCāC plural like, diyār ‘dwellings’, or verbal nouns like qiyāmah ‘resurrection’).4

Ibn Muǧāhid (323), who was a direct student of Qunbul, was clearly bothered by this reading. He reports that Ibn Kaṯīr read it as such and that that is how he learned it from Qunbul. However, he brings transmissions of not just al-Bazzī, one of the transmitters he also reports in his discussion of his ʔisnāds of Ibn Kaṯīr but also Ibn Fulayḥ, that they rejected the reading and that Ibn Kaṯīr only read with one hamzah. He discusses this word again at Q28:71, where Ibn Muǧāhid (495) says: “Only Ibn Kāṯīr read bi-ḍiʔāʔin with two hamzahs. And I learned it thus from Qunbul, but he was wrong (ġalaṭa).”5

6.4.2 Mūṣadah → muʔṣadah

The C-stem passive participle written as ‮موصده‬‎ ‘closed’ (Q90:20; Q104:8) is read by the majority of the readers as mūṣadah. However, Ḥafṣ ʕan ʕāṣim, ʔabū ʕamr and Ḥamzah read it as muʔṣadah (Ibn al-Ǧazarī, § 1484). This variant is a clear pseudocorrection. The verb ʔawṣada ‘to close’ (√wṣd) is also recorded as ʔāṣada (√ʔṣd) in classical lexicons, but within the QCT the verb clearly has √wṣd as its root, as is confirmed by waṣīd ‘threshold, doorstep’ derived from the same root and also attested in the Quran ‮الوصيد‬‎ (Q18:18). As the root is √wṣd in Quranic Arabic, mūṣadah is the expected form and muʔṣadah the pseudocorrection resulting from the ambiguity of the hamzah-less Quranic Arabic, where C-stem participles (and imperfect) of I-ʔ regularly merge with I-w roots.

This pseudocorrection did not go unnoticed by classical authors either. Al-Zamaxšarī (d. 538 AH/1144 CE) in his al-Kaššāf (IV, 257) brings a report (without ʔisnād) on the authority of Šuʕbah—who read mūṣadah—saying: “our Imam [i.e. ʕāṣim] would apply the hamzah to ‮موصده‬‎; and I wanted to plug my ears whenever I would hear it.” This story may be apocryphal, designed to explain the difference between Ḥafṣ and Šuʕbah in their transmission of ʕāṣim. Nevertheless, it highlights that clearly this reading was disturbing enough to the grammarian and theologian al-Zamaxšarī, that it was worth relating it.

The confusion between I-ʔ and I-w roots is well-known for the D-, L- and C-stems in Middle Arabic (Blau 1967, § 72.1), and is the result of subsequent analogies after the loss of the hamzah, which is common to Quranic Arabic and Middle Arabic alike. This results in a merger of the two root types in the prefix conjugation and participial derivations. This leads to the frequent appearance of pairs of I-ʔ and I-w verbs with identical meaning. We find a similar case of such a development for ‮توكيدها‬‎ ‘their affirmation’ (Q16:91), which looks like the verbal noun of wakkada ‘to affirm’, but Classical Arabic lexicons also record taʔkīd and ʔakkada with the same meaning. When cognates in other Semitic languages are lacking, it is often difficult to recover whether the I-ʔ form was originally a pseudocorrection, or that the I-w is simply a generalized form from an original I-ʔ verb in a dialect that has lost the hamzah.

6.4.3 Ḍiʔzā

Another case of pseudocorrection in Ibn Kaṯīr’s reading is the word ‮ضيزى‬‎ ‘most unfair’ (Q53:22) as he reads it as ḍiʔzā rather than ḍīzā/ḍīzē (Ibn al-Ǧazarī, § 1484). While this word seems to be basically only known to the Arabic lexicographers and grammarians from its Quranic context, its morphology is transparent: it must be a feminine elative, as there are no other feminine adjectives that end in ʔalif maqṣūrah. It being an elative, one would expect the pattern to be CuCCā, had the noun indeed been derived from a root √ḍʔz, then we would expect ḍuʔzā, not ḍiʔzā, which rather is a pseudocorrect insertion of hamzah on the vocalic pattern of a √ḍyz root, cf. CAr. ʔaṭyab f. ṭībā ‘better’, CAr. ʔarʔaz f. ruʔzā ‘more roaring’.6

6.4.4 Manōh → manāʔah

The majority of the readers read the name of the pre-Islamic goddess Manāt as manāh. But Ibn Kaṯīr reads this as manāʔah. The goddess Manāt is a personification of Fate, whose name is deribed from the root √mnw alternating with √mny. This root is well-attested in Pre-Islamic Arabic, the deity Manāt is spelled mnwt in Nabataean, and the fates are also an often invoked in Safaitic as mny /manāy/ (Al-Jallad and Jaworska 2019). The insertion of the hamzah by Ibn Kaṯīr cannot be seen as anything but a pseudocorrect reading.

6.4.5 ʕādan l-ʔūlā

Q53:50 contains a unique sequence in the Quran, the only place where a word ending in tanwīn is followed by the definite article, which is followed by a word that starts with a hamzah. This sequence yields a cluster of three consonants /ʕādan l-ʔūlā/, which is resolved differently by different readers (Ibn al-Ǧazarī, § 1547–1557s). Normally, in the case of a clash of nunation with the definite article, an epenthetic i is inserted, and that is the reading of the majority of the canonical readers: [ʕādani l-ʔūlā]. Warš ʕan Nāfiʕ and ʔabū ʕamr, however, resolve this cluster differently in this specific case. The tanwīn is assimilated to the lām and the subsequence /llʔ/ is resolved by eliding the glottal stop of the word, yielding [ʕādal-l-ūlǟ].

Qālūn ʕan Nāfiʕ however, applies yet another development and reads the [ʕādal-l-uʔlā]. Qālūn must have interpreted the feminine elative as being phonemically /ʔuʔlā/, after the application of the regular development ʔvʔC > ʔv̄C, as seen for example in *ʔaʔkulu > ʔākulu ‘I eat’, *ʔuʔtiya > ʔūtiya ‘it was given’. With the loss of the initial hamzah, the second hamzah gets a chance to reappear, a phenomenon we mostly see in imperatives such as ʔīti /(i)ʔti/ but wa-ʔti /wa-ʔti/. The problem here, however, is that the interpretation of ʔūlā as /ʔuʔlā/, is clearly pseudocorrect due to the inherent ambiguity of the surface form. The root of this form is √ʔwl, and thus the underlying form is not **/ʔuʔlā/ but /ʔuwlā/.

6.4.6 Durriyy → du/irrīʔ

‮كوكب درى‬‎ “a brilliant star” (Q24:35) is read by the majority of the readers as kawkabun durriyy, where the latter word is clearly to be understood as a denominal adjective of durr ‘pearls’ followed by the nisbah-ending. However, ʔabū ʕamr and al-Kisāʔī read this word as dirrīʔ and Šuʕbah ʕan ʕāṣim and Ḥamzah as durrīʔ (Ibn al-Ǧazarī, § 3731).

Ibn Xālawayh (ḥujjah, 262) explains that the reading dirrīʔ should be understood as an intensive adjective (like sikkīt ‘intensely silent’) of the root drʔ ‘to avert; rush out (said of a torrent)’, hence ‘rushing out intensely’ likening the rushing out to the intensity of the light. This explanation is probably a post hoc rationalization of a reading with a pseudocorrect hamzah. Ibn Xālawayh suggests that durrīʔ has the same meaning as dirrīʔ, but fuʕʕīl nouns like this otherwise do not exist in Arabic, so such an explanation is not particularly convincing.

6.4.7 Maʕāyiš

An interesting point where what is considered correct and what is transmitted comes into conflict is in the plural of maʕīšah ‘livelihood’, which in the reading traditions today is maʕāyiš (Q7:10, Q15:20). The use of a yāʔ in this case is surprising, other nouns with a similar structure consistently have a hamzah in this position, e.g. madīnah pl. madāʔin ‘town’; ḥadīqah pl. ḥadāʔiq ‘garden’; xalīfah pl. xalāʔif ‘successor’; qabīlah pl. qabāʔil ‘tribe’, etc.

The shift of āyi, āwi to āʔi is essentially a regular development, and we find it not just in the broken plural pattern here, but also active participles of hollow roots, e.g. qāʔim ‘standing’, and this development may also be the origin of word final āʔ such as in samāʔ ‘sky’.7

The only place in Classical Arabic where both *y and *w are retained after ā and before i is in the verbal system, the L-stems retain the root consonant in the imperfective, even though the regular development would require a shift to āʔi. This, however, can be easily explained as the result of analogy. The perfective form regularly retains the root consonants, and this is simply expanded to the imperfect, where it would have regularly been lost. This development can be seen as a three-stage development as follows:

1. Proto-Arabic

2. *āyi, *āwi > āʔi

3. Analogical levelling



qāwama/yuqāʔimu >> yuqāwimu



ʕāyaša/yuʕāʔišu >> yuʕāyišu

For the plural maʕāyiš no analogical basis to which the *y could be restored can be found. As such, maʕāyiš is a deviation from what we would expect a form of Arabic that underwent the *āyi, *āwi > āʔi shift to produce. It turns out that in the reading traditions, the form maʕāʔiš is in fact known.8 Ibn Muǧāhid, who does not usually spend time discussing šāḏḏ readings in his Sabʕah fī al-Qirāʔāt discusses this form and is curt about it:

The word ‮معيش‬‎: All of them read maʕāyiš without the hamzah. But Xāriǧah, on the authority of Nāfiʕ transmits maʕāʔiš with an overlong vowel and a hamzah. And ʔabū Bakr [ibn Muǧāhid] said: this is a mistake.9

Sadly, Ibn Muǧāhid does not elaborate on why he considers it a mistake. An answer is found in Sībawayh’s al-Kitāb however, who is in agreement with Ibn Muǧāhid that this word should not have a hamzah. He argues that, because this word is derived from a root where the yāʔ is part of the root √ʕyš, this yāʔ should be retained (Sībawayh, IV, 354–357). Sībawayh is right to observe that this makes the word objectively different from the other words cited so far, where the ī of the singular formation is part of the pattern CaCīCah, rather than part of the root, e.g. madīnah has √mdn10 and ṭarīqah has √ṭrq.

In this argument, Sībawayh is undoubtedly thinking of words such as the imperfective L-stem verbs such as yuʕāyišu where the root consonant is retained as well. However, we must conclude that this is a post hoc argumentation. First of all, we cannot assume that speakers of Arabic were themselves grammarians like Sībawayh, and therefore a sound law that would only apply to CaCīCah nouns, when the ī happened to not be the result of a root consonant, is not something that is likely to have occurred in natural language, as it requires a highly abstract model of formal grammatical thinking. Second, the argument that because the yāʔ is part of the root it could not undergo the *āyi > āʔi shift clearly breaks down in other derivational forms. The active participle of ‘to live’, after all is ʕāʔiš, not ʕāyiš, nor is ‘bird’ ṭāyir, but rather ṭāʔir. Sībawayh’s opinion, which Ibn Muǧāhid upholds as the status quo, therefore cannot be seen as anything other than rationalization for his choice to prefer maʕāyiš over maʕāʔiš when he was confronted with the choice between the two.

While later scholars of the qirāʔāt, such as Al-Dānī (ǧāmiʕ, 511), simply fell in line with Ibn Muǧāhid’s opinion, not all scholars found themselves in agreement with his judgment. ʔabū Ḥayyān al-Andalusī (d. 754 AH) in his al-Baḥr al-Muḥīṭ (V, 15) brings forth a rather spirited argument in favour of maʕāʔiš as a correct and acceptable reading.11

The general public reads maʕāyiš with the yāʔ, this is an analogy (qiyās), because the yāʔ in the singular is part of the root, and not an extra letter to the pattern so that it receives a hamzah. When it is an extra letter of the pattern, they add the hamzah, for example saḥāʔif of ṣaḥīfah. Al-ʔaʕraǧ and Zayd b. ʕalī and al-ʔaʕmaš and Xāriǧah, on the authority of Nāfiʕ and Ibn ʕāmir in their (respective) transmission read maʕāʔiš with a hamz. This is not analogy (qiyās), because they reported it, and they were trustworthy, so it is necessary to accept it (as a valid reading). This hamzah is irregular in the same way as it is irregular in manāʔir, the plural of manārah—it is originally manwarah—and [it is irregular in the same way as it is irregular] in maṣāʔib, the plural of maṣībah—it is originally maṣwibah. Manāwir and maṣāwib are analogies as they would say maṣāwib on the basis of the root, in the same way that they say the plural of maqāmah as maqāwim; [the plural of] maʕūnah as maʕāwin.

Al-Zaǧǧāǧ said “all of the Basran grammarians decided that adding a hamzah is a mistake, but I know nothing of this perspective; [I know] only that [adding hamzah makes] it similar to ṣaḥīfah, ṣaḥāʔif and it is not proper to rely on this reading [i.e. maʕāʔiš].”

Al-Māzinī said: “The origin of the dispute of this reading is on the authority of Nāfiʕ, but he did not know what the ʕarabiyyah was, and the speech of the Arabs [i.e. correct Arabic] is to correct it [i.e. towards maʕāyiš] in such cases.”

But we are not worshippers of the opinions of the grammarians of Basra! (lasnā mutaʕabiddīna bi-ʔaqwāli nuḥāti l-baṣrah).

Al-Farrāʔ said: “sometimes the Arabs added a hamzah to this and what is like it, supposing that it is a faʕīlah, and they liken mafʕilah to faʕīlah”.12 So, this is an account from al-Farrāʔ on the authority of the Arabs that they would sometimes add a hamzah to this and what is like it.

He brought an account of the reading of trustworthy people: Ibn ʕāmir, he is a pure Arab, and he received the Quran from ʕuṯmān before corruption [of the Arabic language] manifested itself. As for al-ʔaʕraǧ, he was among the greats of the readers of the followers [of the companions of the prophets]. Zayd b. ʕalī, with regard to eloquence and knowledge and cases one seldomly encounters, in that [more than] anyone. As for Al-ʔaʕmaš, he was, with regard to precision, perfection, memory and trustworthiness of high status. As for Nāfiʕ, he was taught by 70 of the followers [of the companions of the prophet] and with regard to eloquence, precision and trustworthiness he was of high status, as he was not ignorant. Therefore, it is necessary that we accept what they relate to us, and [we should] not pay heed to the disagreement of the grammarians of Basra in this example.

As for the words of al-Mazānī “The origin of the dispute of this reading is on the authority of Nāfiʕ”, this is incorrect, because it is (also) reported on the authority of ibn ʕāmir and on the authority of al-ʔaʕraǧ, Zayd b. ʕalī and al-ʔaʕmaš; As for the words “Nāfiʕ did not know what the ʕarabiyyah is”, this is the evidence for the rebuttal: If we suppose that he did not know what the ʕarabiyyah was; is it this skill [i.e. knowing what the ʕarabiyyah is] which gives him access to speaking the language of the Arabs? He does not have to [know what the ʕarabiyyah is] to do that [speaking the ʕarabiyyah]! For he is eloquent of speaking the ʕarabiyyah, as he is a transmitter of the reading on the authority of the eloquent Arabs. And many among those grammarians think badly of the readers, but it is not correct of them [to do] that.

This account clearly shows that, despite the objections of the Basran grammarians, such forms were known and at least allowed by some, and may have indeed been the regular outcome in the dialects that gave rise to the CaCāʔiC style plurals.

6.4.8 Måḡoḡ → Maʔǧūǧ

ʕāṣim is the only reader who reads the names of Gog and Magog as yaʔǧūǧ and maʔǧūǧ, whereas the other readers read yāǧūǧ and māǧūǧ (Ibn al-Ǧazarī, § 1484). As these names are clearly borrowed from the Hebrew ‮גוג ומגוג‬‎ goḡ u-måḡoḡ, which do not have a hamzah in either word, ʕāṣim’s reading is an innovation from its original source.

6.4.9 Zakariyyā → Zakariyyāʔ

Most readers are in agreement that the Biblical name Zachariah in Arabic is supposed to end in a hamzah, i.e. zakariyyāʔ, this despite the fact that the Quranic rhyme in Q19:3 clearly suggests the name was pronounced /zakariyyā/ in Quranic Arabic. Only Ḥafṣ ʕan ʕāṣim, al-Kisāʔī, Ḥamzah and Xalaf lack this hamzah (Ibn al-Ǧazarī, § 2840). Considering that the Hebrew name is ‮זכריה‬‎ Zəḵaryå, (or Greek Ζαχαρίας) without a final glottal stop, we must conclude that the majority of the readers are pronouncing the name with a pseudocorrect hamzah.13

6.4.10 Sāq, sāqay-hā, sūq → saʔq, saʔqay-hā, suʔq/suʔūq

Another case of pseudocorrection is found in the plural and dual of sāq ‘thigh, shank’ in the canonical reading traditions. While in Classical Arabic this word is pronounced sāq pl. sūq, Ibn al-Ǧazarī (§ 3810) reports that Qunbul ʕan Ibn Kaṯīr read ‮بالسوق‬‎ (Q38:33) and ‮سوقه‬‎ (Q48:29) with a hamzah (= bi-s-suʔq or alternatively bi-s-suʔūq), but his transmitter al-Bazzī read it without a hamzah. He also reports the presence of the hamzah for the dual ‮ساقيها‬‎ (Q27:44), i.e. saʔqay-hā ‘her two shins’.

Ibn Muǧāhid (483) explicitly points out that the singular sāqin (Q68:42, and by extension presumably its other attestation in Q75:29) was not pronounced with a hamzah. Eventually the form without the hamzah wins out in the classical norm, and it is clear that even by Ibn Muǧāhid’s time this was the norm, but it is also clear that the form with hamzah was a serious contender at least in the tradition that sprouted from Ibn Kaṯīr. For the plural, the forms suʔq and suʔūq have become canonical in Qunbul’s transmission, rather than the expected form sūq.

Unease with these forms used by Ibn Kaṯīr can also be gleaned in the discussion of ʔabū Ḥayyān (VIII, 244 and also IX, 155), who quotes ʔabū ʕaliyy14 as saying that forms like saʔq, saʔqay-hā and suʔq are weak, and that it is based on a ‘well-known linguistic practice’ (luġah mašhūrah) to apply the hamzah to a unvowelled wāw when a ḍammah precedes, citing a piece of poetry from ʔabū Ḥibbah al-Numayrī: ʔaḥabbu l-muʔqidīna ʔilayya muʔsā ‘Moses is the most beloved of kindlers15 to me’. This explanation fails to account for the presence of the hamzah in the dual saʔqay-hā, and presumably for that reason ʔabū Ḥayyān disagrees. He says that the form is acceptable because there is a hamzah in the root, clearly showing that as late as his lifetime there still had not developed a complete consensus as to whether the root of sāq should be understood to be √sʔq or √swq.

The Arab grammarians were unable to resolve the question as to whether the root was supposed to contain a hamzah or not. But from a comparative linguistic perspective it is clear that the hamzah in the word is pseudocorrect. Other Semitic languages show no sign of the in this word. Aramaic has šāq, but the sequence *aʔC should yield ēC in Aramaic. This is clear from the verbal system, e.g. yēmar ‘he says’ < *yaʔmuru and also from other words of the shape CaʔC, e.g. rēš ‘head’ (cf. Ar. raʔs, Hebr. roš spelled etymologically as ‮ראש‬‎), kēḇ ‘grief’ (cf. Ar. kaʔb ‘id.’ and Hebr. kʔeḇ ‘id.’).16 Hebrew has šoq, spelled ‮שק‬‎; this points to the absence of the as Hebrew usually retains the spelling of the spelled with ʾålɛp̄ in the consonantal text, but šoq is not spelled **‮שאק‬‎. Finally, in Ugaritic, which retains the Proto-Semitic with a variety of signs, lacks it completely in this word ⟨šq⟩ (not the expected **⟨šˀiq⟩). This evidence leads us to an unambiguous reconstruction of this word for Proto-West Semitic as *sāḳ, without a glottal stop. The form sāq is therefore etymological and forms with a hamzah are pseudocorrect.17

On the discussion of sāq, Ibn Xālawayh (ʔiʕrāb, II, 152f.) explicitly calls out ‘Arabs’ for placing the hamzah in places where it is incorrect.

Others said: sāq is like bāb, because the root is s-w-q, and the wāw is changed to an ʔalif, so it is incorrect to give it a hamzah. This is what is among the things in which the Arabs make mistakes, so they do apply the hamzah on what does not have a hamzah, and similarly with what has a hamzah they do not give it the hamzah, so kaʔs and raʔs and sāq their stem shape (waznuhā) is the same (i.e. as CāC), so they make them similar to one-another, yes, he has seen that Arabs say: ḥallaʔtu s-sawīqa, but originally it is ḥallaytu, and likewise, with ḥallaʔtu l-ʔinsāna ʕani l-māʔi wa-l-ʔibili. However, the plural of sāq, through replacement (qalb) (of the hamzah) is ʔaswuq without hamzah and if you wish (can be) ʔasʔuq with hamzah.18

6.4.11 Kās → kaʔs

As already noticed by Blau (1970, 56), much like the case of saʔq discussed above, comparative Semitic evidence shows that kaʔs ‘cup’ must have a pseudocorrect hamzah in Arabic. The reflexes in Hebrew ‮כוס‬‎ kos (spelled without ʾålɛp̄) and Aramaic kās as well as Ugaritic ⟨ks⟩ leave no doubt that the reconstruction of this noun in Proto-West Semitic is *kāts and the hamzah in the Quranic reading traditions must be pseudocorrect. What is different from the case of saʔq, however, is that this word is read with hamzah universally by all the canonical readers. Moreover, this pronunciation has become the de facto standard in Classical Arabic, although the form kās is known to exist among the lexicographers (Lane 2639c; Lisān 3802c).

6.4.12 Yuḍāhūna → yuḍāhiʔūna

ʕāṣim is unique in reading ‮يضهون‬‎ ‘they imitate’ (Q9:30) as a III-ʔ root yuḍāhiʔūna. All other readers treat the verb as a III-w/y verb, reading yuḍāhūna (Ibn al-Ǧazarī, § 1532). This verb is attested in an Old Arabic inscription in Safaitic script as ḍhw ‘to copy’ (Al-Jallad and Jaworska 2019). As Safaitic regularly retains the hamzah (Al-Jallad 2015, 45, 53), ʕāṣim’s reading is evidently pseudocorrect here, and the majority reading is the original.

6.4.13 Aṣ-ṣābūna → aṣ-ṣābiʔūna

There is disagreement among the readers on how to read ‮الصبين/الصبون‬‎ ‘the Sabians’ (Q2:62, Q5:69, Q22:17), which is variously read as aṣ-ṣābū/īna (Nāfiʕ19) an aṣ-ṣābiʔū/īna (the others) (Ibn al-Ǧazarī § 1496). That is, either as an active participle from a root √ṣbw/y or from a root √ṣbʔ.

Neither the root √ṣbw/y nor √ṣbʔ is attested in Arabic in a meaning that would elucidate the meaning of the word Sabians as an Arabic word; hence it is usually taken to be a loanword. If Wellhausen (1897, 237) is correct to identify this word as a plural active participle derived from the Mandaic verb ṣḇā ‘to baptize’ then we must conclude that the hamzah is a pseudocorrection. As the Mandaic form is a final weak verb, we would expect the plural active participle to simply be aṣ-ṣābūna.20

6.4.14 Conclusion

While the cases where hamzah is applied to a word which etymologically never had it is relatively rare, it is common enough to show that there was a real attempt to classicize the readings towards an ideal that included hamzah, by people to whom it was not necessarily obvious which words were supposed to have a hamzah or not. This is certainly consistent with what we would expect to find, considering that rhyme evidence shows that the Quran was originally composed in the Hijazi dialect without a hamzah.

It is remarkable that a good number of these pseudocorrections are found with Ibn Kaṯīr, the Meccan reciter. Ibn Kaṯīr, despite being a Hijazi, has a remarkably conservative use of hamzah in his recitation. Considering how the Hijazi vernacular appears to have mostly lost the hamzah, it is by no means surprising that it is exactly this reader that is most prone to pseudocorrection. It should be noted, however, that pseudocorrections are also found with other readers. All readers read kaʔs, and the Kufans have several forms with pseudocorrect presence of hamzah as well.

6.5 Failure to Insert hamzah

While the amount of pseudocorrect insertions of hamzah in places where the word historically lacked the hamzah is a fairly rare occurrence among the readers, failure to insert the hamzah is more common. The tradition also explicitly acknowledges this: the dropping of hamzah is part of the ʕarabiyyah (Sībawayh, III, 541 ff.) and therefore grammarians did not see it as a problem to, in general, retain hamzah, but in cases that one was uncertain whether the root had a hamzah, to opt for the hamzahless form instead. However, the pseudocorrect application of hamzah was considered something to be avoided. This can be seen, for example in ʔabū ʕamr’s statement concerning his reading of minsaʔata-hū as minsāta-hū (Ibn al-Ǧazarī, § 3962). On this topic al-Farrāʔ (Maʕānī II, 356 f.) reports:

ʔabū Jaʕfar al-Ruʔāsī (d. 187 AH) declared to me (al-Farrāʔ, d. 208 AH) that he asked ʔabū ʕamr (d. 154 AH) about it [i.e. the pronunciation of ‮منسٰته‬‎], and (ʔabū ʕamr) said: “minsāta-hū is without hamzah”, and he also said: “Because I do not know it, I remove its hamzah.”

A slightly more expanded version of this account is related by ʔabū Ḥayyān (VIII, 531):

ʔabū ʕamr said: “I do not apply hamza to it, because I do not know its derivation; If it was among those (roots) that are not hamzated, I have been (sufficiently) cautious (iḥtaṭtu), and if it was hamzated, then it would be permissible for me to take away the hamzah in what contains a hamzah.”

This account clearly shows that the leaving out the hamzah in places where it is etymologically present was not considered a mistake, while adding it where it should have been was. Moreover, it shows that adding the hamzah was a rational and theoretical endeavour by the readers, and in case of uncertainty they could decide to leave it out.

The reading of Ibn Ḏakwān of this word is minsaʔta-hū (sic!). This is evidently ungrammatical as it suggests a miCCaCt stem formation, something that does not occur in any form of Arabic. It rather seems like an attempt at inserting the hamzah into a word that he originally learned to recite as minsāta-hū. If one disregards any forms of grammar, there is no way to decide whether a base for minsāta-hū is to be pronounced minsaʔta-hū or minsaʔata-hū.

Al-Dānī (Ǧāmiʕ, 680) points out that Ibn Ḏakwān’s reading was considered extremely weak by the Arab grammarians in general, because the feminine ending should always be preceded by -a- or an ʔalif. But, he says, there is a line of poetry, transmitted by al-ʔaxfaš (the same person who transmits this reading for Ibn Ḏakwān), which serves as evidence that the form minsaʔt- exists:

× × ⏑ – | × × ⏑ – | × × ⏑ – ||

× × ⏑ – | × × ⏑ – | × × ⏑ – ||

ṣarīʕu xamrin qāma min wakaʔti-hī

ka-qawmati š-šayxi ʔilā minsaʔti-hī

‘A drunk stood up from his reclination,

like the standing up of an old man on his stick.’

The problem with this poetic evidence is that minsaʔti-hī is metrically identical to minsāti-hī, and therefore this poem can hardly be used as evidence for it. This is assuming that this line of poetry is not an outright fabrication, which seems more likely in this case. This anonymous line of poetry is only ever cited to explain Ibn Ḏakwān’s reading, and al-ʔaxfaš seems to be the originator of the line.

The contemplative and theoretical nature of the reading with or without hamzah is also displayed in a colourful exchange between al-Kisāʔī and Ḥamzah on the discussion of al-Kisāʔī reading ḏiʔb, as ḏīb (Xalaf, Warš and regularly by his principles ʔabū Ǧaʕfar follow him in this exceptional reading, see Ibn al-Ǧazarī § 1472). This is related in several Ṭabaqāt works, such as the one of al-Ḏahabī (153 f.):

[…] Muḥammad b. ʕalī b. Sulaymān al-Marwazī said: I asked Xalaf b. Hišām: why is al-Kisāʔī called al-Kisāʔī? And he said: al-Kisāʔī entered Kufa and came to the as-Sabīʕī mosque where Ḥamzah was teaching recitation, and al-Kisāʔī came forward and he was wrapped in a black robe. When Ḥamzah was done praying he said: who goes first? And it was said: “al-Kisāʔī”, and they meant the guy in the (black) robe, and they turned their gaze to him and said: if you are a weaver,21 you will recite Sūrat Yūsuf and if you are a salt vendor (or sailor (?), mallāḥ), you will recite Sūrat Ṭāhā. So, he heard them and started to recite Sūrat Yūsuf, and when he arrived at the pericope of the wolf, he recited it without hamzah (i.e. aḏ-ḏīb). So, Ḥamzah said: “ad-ḏiʔb is with hamzah.” So, al-Kisāʔī replied: “So should I apply the hamzah like that in al-ḥuʔt (for al-ḥūt ‘the whale’) as well?”—this is about (the verses) fa-ʔakala-hu ḏ-ḏiʔb (Q12:17) and fa-ltaqama-hu l-ḥūt (Q37:142). Ḥamzah looked to Xallād the cross-eyed, and they argued as a group, but nobody was able to (answer him). Then they said (to al-Kisāʔī): “liberate us, please!” Then (al-Kisāʔī) said: “Learn from what this weaver has to say! When you compare a man to a wolf, you say qad istaḏaʔaba r-raǧul ‘the man was fierce like a wolf’, and if you would say istaḏāba without hamzah, then it is as if you attribute to him emaciation (huzāl) [because ḏāb means ‘vice, fault, defect’]. But when you liken him to a whale, you say: istaḥāta r-raǧul”—which means he eats a lot, because a whale eats a lot—and then he recited:

ʔayyuhā ḏ-ḏību wa-bnu-hū wa-ʔabū-hū
ʔanta ʕindī min ʔaḏʔubin ḍāriyātī
‘O wolf, and his son, and his father!
You are to me among the voracious wolves!’

And he is known as al-Kisāʔī ever since that day.

This account once again shows that, while eliding the hamzah is considered acceptable—after all that is how the star of the story recites it—it is not allowed to pseudocorrectly apply the hamzah to words that do not have it in their root.22

It is worth mentioning here a not quite as colourful, but related account on the authority of Nāfiʕ related by al-ʔaṣmaʕī → ʔabū Saʕīd al-Ḥāriṯī → Ibn Muǧāhid: “I asked Nāfiʕ about ‮الذيب‬‎ and ‮البير‬‎, and he said: ‘If the Arabs provide a hamzah to them, then provide them with a hamzah’ ” (Ibn Muǧāhid, 346). This quote is related in the context of disagreement among the transmitters of Nāfiʕ on these words. While most transmitters are in agreement that he read these words with hamzah, Warš and the non-canonical transmitter Ibn Ǧammāz read them as al-bīr and aḏ-ḏīb, a practice that Ibn Muǧāhid considered mistaken (wa-hāḏā wahm). What this quote illustrates is the rather practical nature of reading with or without hamzah. Nāfiʕ gives a rather non-committal answer to the question, telling the readers to follow what they believe what “the Arabs” do.23

Once we look closer among the canonical readers, we find numerous examples where there is uncertainty on whether a word is supposed to carry a hamzah or not, several readers opt for hamzah-less forms where according to their general principles of recitation we would expect them to have been retained. In the following section, we will examine the many cases of incomplete application of the hamzah as they occur among the readings.

All of this uncertainty about where the hamzah should go is difficult to understand, if we assume that the language of the Quran was indeed pronounced and transmitted with a hamzah from the very start. On the other hand, such discussions make perfect sense if the Quranic language was—as is admitted for Hijazi Arabic—without the hamzah, and as a new linguistic ideal of the classical poem gained prominence, reciters started adapting features, including the use of the hamzah, into their recitation.

6.5.1 Long Vowels Followed by Hamzah Nabīʔ, nabīʔīn, ʔanbiʔāʔ, nubūʔah

The majority of the Quranic reciters do not pronounce the hamzah in the word ‮نبى, النبى‬‎ ‘prophet’ or its plurals ‮النبين, انبيا‬‎ nor ‮النبوه‬‎ ‘prophecy’. The Medinan reciter Nāfiʕ, however is an exception to this, as he consistently recites these words as nabīʔ, an-nabīʔīn, ʔanbiʔāʔ and an-nubūʔah (Ibn al-Ǧazarī, § 1531).

This has frequently been construed as pseudocorrect application of the hamzah (e.g. Vollers 1906, 95; Rabin 1951, 131–133; Fischer 2002, 26), where Zwettler (1978, 179 f., n. 71) even claims that it was never part of the ʕarabiyyah. From an etymological perspective it is not clear that this is correct. This word is ultimately a loanword from Aramaic or perhaps Hebrew, and while in later forms of both Aramaic and Hebrew the is lost, Biblical Hebrew and Aramaic still retain the ʾålɛp̄ spelling, suggesting its original presence and pronunciation in these respective corpora, cf. Biblical Aramaic Ktiv ‮נביאה‬‎ */nabīʔā/; Qre nḇiyy-å pl. Ktiv ‮נביאיא‬‎ *nabīʔ-ayyā Qre nḇiyyayyå (Ezra 5:1); Hebrew ‮נביא‬‎ nåḇi < *nabīʔ pl. ‮נביאים‬‎ nḇiʾim < *nabīʔīm. Cf. also Hebrew ‮נבואה‬‎ nḇuʔå ‘prophecy’ and Biblical Aramaic Ktiv ‮בנבואת חגי‬‎ /*bi-nabūʔat/ Qre bi-nḇuʔaṯ ‘the prophecy of (Haggai)’ (Ezra 6:14). As the Aramaic loanwords in the Quran consistently show exceptionally archaic phonology (see van Putten 2020a, 69 ff.), it is not a priori obvious that the presence of the hamzah in these words was never part of the Classical language.

The belief that this must be a pseudocorrection seems to be based on the fact that Sībawayh (III, 555) expresses a clear normative bias against pronouncing the word as nabīʔ (and idem for barīʔah for bariyyah see the discussion below), saying that this is the manner of pronunciation of the people of the Hijaz who pronounce the hamzah, calling it rare and abhorrent (qalīl radīʔ). But while this is the case, he clearly considers the base of this word and barīʔah to contain a hamzah, as he discusses it as part of the shift of īʔ, ūʔ, ayʔiyy, uww, ayy including words which in Classical Arabic are usually realized with the hamzah, e.g. xaṭīʔah → xaṭiyyah ‘sin’, and maqrūʔmaqruww ‘readable’. For the formation of diminutives Sībawayh (III, 547) explicitly allows both nubayy and nubayyiʔ, but the diminutives of barīʔah/bariyyah and nubūʔah/nubuwwah he only endorses the forms with hamzah, i.e. burayyiʔah and nubayyiʔah. So, while he has a normative opinion for the dropping of hamzah, he clearly considers the ʔaṣl of the word to have had the hamzah. We cannot conclude from this that nabīʔ or barīʔah are pseudocorrect, but only that the now normative form without hamzah had gained enough ground in Basra in Sībawayh’s time that it was considered normative despite being exceptional among the people that usually preserve the hamzah. But Nāfiʕ is Medinan and a contemporary of Sībawayh’s teacher al-Xalīl b. ʔaḥmad, so clearly it was still part of the ʕarabiyyah at that time despite Sībawayh’s misgivings.

Despite the archaic nature of Nāfiʕ’s reading, it is quite clear that this was not the reading that belonged to the language of the QCT. The broken plural pattern the QCT uses (ʔaCCiCāʔ) is almost exclusively applied to final weak and geminate roots only a few sound roots have this pattern, e.g. ġaniyyʔaġniyāʔ ‘rich’ and šadīdʔašiddāʔ ‘strong’, qarībʔaqribāʔ ‘relative’ (van Putten 2020a, 64). Had the Quranic Arabic form indeed been nabīʔ, we would have rather expected a plural nubaʔāʔ.24 This plural pattern therefore suggests that in Quranic Arabic, as would be expected in Hijazi Arabic the final hamzah had been lost and the word was indeed pronounced as the majority of the readers read it.

Nāfiʕ’s reading in this case is therefore an archaism, and one that was not considered proper by everyone. A commonly cited prophetic Hadith has someone address the prophet by yā nabīʔa ḷḷāh, which is promptly denounced by the prophet. This tradition is explicitly invoked as one of the reasons why a reciter might read nabiyy instead of nabīʔ by Ibn Xālawayh (Ḥuǧǧah, 80 f.): “the first reason is that applying the hamzah is heavy on their speech, and the evidence for this is his speech (PBUH): I am not the prophe’ of God (lastu nabīʔa ḷḷāh); it is as if he disliked applying the hamzah because he was of Qurayš who do not apply the hamzah”. Barīʔah/bariyyah

Another loanword from Aramaic or Hebrew is ‮البريه‬‎ ‘creature’ (Q98:6,7), which like nabīʔ, is read as al-barīʔah by Nāfiʕ, but in this case Ibn Ḏakwān ʕan Ibn ʕāmir joins him in this reading, other reciters read al-bariyyah (Ibn al-Ǧazarī, § 1536).

Here too we are likely dealing not with a pseudocorrection, but an accurate transmission of the ancient pronunciation of an original in Hebrew ‮בריאה‬‎ briʾå and/or Aramaic (Jeffery 2007, 76), which is a derivation from the verb ‮ברא‬‎ ‘to create’ (Hebrew bårå), which likewise was borrowed into Arabic as baraʔa. Nasīʔ

‮النسى‬‎ (Q9:37) ‘the postponement’ is read by Warš ʕan Nāfiʕ (in the path of al-ʔazraq) and ʔabū Ǧaʕfar as an-nasiyy while the other readers read it as an-nasīʔ (Ibn al-Ǧazarī § 1525). In the context, it seems quite clear that we should derive this word from the root nasaʔa ‘to postpone; to drive’, and not from nasā ‘to forget’, where an-nasiyy would end up meaning ‘the forgotten one; that which is to be forgotten’. Note that minsaʔah, also a word derived from this root, likewise yielded uncertainty among the readers as to whether or not it should have the hamzah (see § 6.5 above).

Ibn Muǧāhid (314) reports several other readings. In non-canonical transmission paths of Ibn Kaṯīr we find an-nasʔ (ʕuqayl ← Šibl ← Ibn Kaṯīr)—a reading that seems to disagree with the rasm—and an-nasiyy (ʕubayd ← Šibl ← Ibn Kaṯīr). He also reports an-nasy on Ibn Kaṯīr’s authority, but without ʔisnād. Xaṭīʔah pl. xaṭāyā ‘sin’

A clear example of failure to apply hamzah in the QCT which has subsequently made it into the Classical Arabic language is the plural formation of xaṭīʔah ‘sin’, its plural, xaṭāyā, not only lacks the expected hamzah altogether, it could never have even had this hamzah, as the plural formation it employs is one typical of final-weak roots. Fischer (2002, § 99b) cites as examples hadiyyah pl. hadāyā ‘gift’, hirāwah pl. harāwā ‘club’ and zāwiyah pl. zawāyā ‘corner, angle’.

There are some other contexts in which the CaCāCā plural appears, but none of them apply to xaṭīʔah.25 There are a few isolated lexical items that take this plural of sound roots, for example yatīm pl. yatāmā. Note, however, that if xaṭīʔah would belong to this group of nouns we would have expected **xaṭāʔā rather than the inexplicable xaṭāyā.

As such we would expect the original singular of this noun in Quranic Arabic to have been the hamzahless form xaṭiyyah, following the same derivation as hadiyyah. The reading xaṭiyyah was subsequently classicized to xaṭīʔah by all readers, while failing to classicize the plural formation along with it.26 The issue of this specific broken plural pattern associated with this singular was not lost on the Arab grammarians, and Lisān al-ʕarab has a lengthy discussion on what was evidently perceived as a problematic plural. The regular plural of CaCīCah nouns is CaCāʔiC (cf. ḥadīqah pl. ḥadāʔiq; madīnah pl. madāʔin), and as such the expected plural is al-xaṭāʔiʔu,27 due to the regular elision of the second hamzah when two hamzahs follow in a row, this should have become al-xaṭāʔī, in the same way that the active participle of ‘to come’ turns from al-ǧāʔiʔu into al-ǧāʔī, and it would therefore be expected to have merged with the faʕālin type plurals. Sībawayh (III, 552 f.) starts his discussion of this plural with: “it is as if [the hamzah] was turned into a yāʔ and the end of xaṭāyā (i.e. the yāʔ of xaṭāʔī) was replaced with an ʔalif”. He then commences to explain how one could get from a singular xaṭīʔah to the plural xaṭāyā without having to assume a singular base xaṭiyyah. He likens the replacement of the final yāʔ of the hypothetical *xaṭāʔī (from earlier xaṭāʔiʔ) with ʔalif to this happening in the final weak plural maṭāyā (plural of maṭiyyah ‘mount’), it is striking here that Sībawayh has to draw an analogy with a CaCiyyah noun, to be able to explain the presence of this plural pattern, while the discussion seems explicitly aimed to avoid this. This brings him to an intermediary form *xaṭāʔā. The hamzah of *xaṭāʔā is subsequently replaced with a yāʔ because it stands between two ʔalifs. While hamzah as a root consonant can stand between two ʔalifs such as in kisāʔāni, kisāʔā, hanāʔā, this is not the case for *xaṭāʔā because its hamzah is not a root consonant, but part of the plural pattern (CaCāʔiC), therefore it is weakened to a yāʔ instead, yielding xaṭāyā. The change from īā is, of course, ad hoc, as is the rule for replacing the hamzah with a yāʔ to go from xaṭāʔā to xaṭāyā, which as far as I can tell is not applied to any other word in the lexicon.

The complexity of discussion ultimately comes down to the fact that Sībawayh, and grammarians after him (see the discussions in Lisān, 1193, for example) refuse to use a surface form like xaṭiyyah—a form explicitly considered to be allowed—for the derivation of the plural. This constraint that the grammarians imposed upon themselves does not lead to a convincing explanation, and that does not seem to have been the point. The grammarians were simply trying to find an explanation of how one could hypothetically come from the idealized source form xaṭīʔah to xaṭāyā without having to assume the loss of hamzah as the basis. The self-evident explanation for the plural xaṭāyā is that it was formed upon the form xaṭiyyah, not xaṭīʔah, thus betraying an original hamzahless form, despite its absence in the canonical readings.

6.5.2 Post-consonantal Hamzah

Above, we discussed a class of words with the lack of the expected hamzah when it occurs after a long vowel. But this is not the only position where we find that readers irregularly lose the hamzah. We also find it in post-consonantal position.

The very name of the Quran itself is one of these cases where the presence of the hamzah is disagreed upon. The word is spelled both defectively ‮قرن‬‎ and plene ‮قران‬‎ in early manuscripts, and it is usually read as qurʔān, however Ibn Kaṯīr reads it as qurān (Ibn al-Ǧazari, § 1571). There can be no doubt that the root of qurʔān ‘recital’ is qaraʔa ‘to recite’, and thus in qurān the expected hamzah is missing. Attempts of Arab philologers to see Ibn Kaṯīr’s qurān as a derivation from qarana ‘to bring together’ are obviously not very satisfying (Jeffery 2007, 233).

Nāfiʕ treats two CiCC verbal nouns derived from III-ʔ roots as CiC stems, ‮ردا‬‎ ridan (versus the other readers ridʔan) ‘as help’ (Q28:34; Ibn al-Ǧazarī, § 1559) and ‮مل‬‎ milu ‘fullness’ (Q3:91; Ibn al-Ǧazarī, § 1560, only in the path of al-ʔaṣbahānī from Warš, and with disagreement among his transmitters). Thus here, like our previous word, post-consonantal hamzah was incompletely re-inserted in this reading.

The QCT of the Quran makes it clear that there was no hamzah in the imperative saʔala ‘to ask’, as it is consistently spelled ‮سل‬‎. Had this word had a medial hamzah, we would have expected a prothetic ʔalif in the imperative **‮اسل‬‎ for isʔal. As such, to agree with the rasm, readers have to read sal if nothing is prefixed to the word. However, whenever wa- or fa- precede the imperative, readers generally include the hamza, as now the rasm allows the correct syllable structure, hence: wa-sʔal al-qaryah ‘ask the village’ (Q12:82) and fa-sʔalū ʔahla ḏ-ḏikr ‘ask the people of remembrance’ (Q16:43). Ibn Kaṯīr, al-Kisāʔī and Xalaf, however, always read the hamzahless form regardless of context (Ibn al-Ǧazari, § 1562). Either reading is, of course, irregular as the imperfect forms of this verb have the same phonetic context but are invariably read as yasʔalu etc.

There are several words that are expected to have a post-consonantal hamzah on comparative grounds, but where all readers are in agreement to not read the hamzah. The most obvious of these is the word ‮ملك‬‎ malak ‘angel’. This word is generally taken to be a loanword from Gəʕəz mălʔăk pl. mălaʔəkt ‘id.’, mostly because it shares the same plural formation as the Arabic ‮مليكه‬‎ malāʔikah, which is a plural formation that is rare, and mostly restricted to loanwords (van Putten 2020a, 66). The Gəʕəz form itself is, of course ultimately derived from the Hebrew ‮מלאך‬‎ malʾåḵ ‘id.’ or Aramaic malʾaḵā ‘id.’

The Arabic plural itself clearly points to a missing postconsonantal hamzah; there are no other CaCaC nouns that have such a quadriradical plural formation (or more common formations like CaCāʔiC). The lexicographical tradition does in fact record the expected form malʔak (Lisān 4269b), but the canonical readers are in agreement that the form is malak, despite this being an irregular outcome within the phonologies of these reading traditions.

The imperfect of the verb raʔā forms a surprising exception to the retention of postconsonantal hamzah, as it is not yarʔā but yarā/yarē among all the canonical readers. The irregular behaviour of this verb seems to have already been a feature of the ʕarabiyyah by the time of Sībawayh (III, 546), as he explicitly mentions the exceptional nature of this word: “all Arabs agree on the dropping of it (the hamzah in forms like ʔarā, tarā, yarā, narā) because of its frequent use”, but he adds: “ʔabū al-Xaṭṭāb told me that he has heard one say qad ʔarʔā-hum bringing the verb in its original form raʔaytu, among the trustworthy Arabs.” Al-Farrāʔ (Luġāt, 165) also says that all Arabs agree on dropping the hamzah with the exception of the Banū ʔasad and Taym al-Rabāb.28

6.5.3 Intervocalic Hamzah Riʔāʔa n-nās → riyāʔa n-nās

‮ريا الناس‬‎ ‘to be seen by men’ (Q2:264; Q4:38; Q8:47) is read by most readers as riʔāʔa n-nās, the regular outcome of a fiʕāl stem of the root √rʔy, but ʔabū Ǧaʕfar has irregularly shifted the medial hamzah to yāʔ, yielding riyāʔa n-nās (Ibn al-Ǧazarī § 1490).29 Liʔallā → liyallā

Warš ʕan Nāfiʕ in the path of al-ʔazraq reads ‮ليلا‬‎ as liyallā ‘so that not’ (Q2:150; Q4:165; Q57:29), while the rest of the Quranic readers read li-ʔallā (Ibn al-Ǧazarī, § 1495). This is irregular behaviour in the reading of Warš, which otherwise retains the hamzah in such environments. Kufuʔan, huzuʔan → kufuwan, huzuwan

While Ḥafṣ is generally very conservative in the retention of the hamzah, he is unique in dropping the hamzah in ‮كفوا‬‎ ‘an equal’ and ‮هزوا‬‎ ‘contempt’, reading them as kufuwan and huzuwan respectively, while the other readers read these words either as kufuʔan/huzuʔan (the majority reading) or kufʔan/huzʔan (Ḥamzah) (Ibn al-Ǧazarī § 2668). Bādiya r-raʔyi → bādiʔa r-raʔyi

An interesting point of disagreement among the readers on the placement of the hamzah occurs in the phrase ‮بادى الراى‬‎ (Q11:27). The majority of the readers reads ‮بادى الراى‬‎ as bādiya r-raʔyi, only ʔabū ʕamr reads it with hamzah, i.e. bādiʔa r-raʔyi (Ibn al-Ǧazarī, § 1535). However, it is not entirely clear that we are dealing with a pseudocorrection or irregular absence of hamzah.

Ibn Xālawayh (Ḥuǧǧah, 186) takes ʔabū ʕamr’s reading as primary, saying that whoever reads it as bādiya is deriving it from the verb badaʔa/yabdaʔu ‘to begin’ and is dropping the hamzah thus understanding the phrase as “beginning in opinion”. If this interpretation is correct, we are indeed dealing with the absence of the expected hamzah which is irregular among each of the readers that reads it thus.

However, al-Farrāʔ (Maʕānī, II, 11) clearly has a different opinion and views bādiya and bādiʔa as two separate lexical items. He tells us: “you should not apply the hamzah to bādiya, because the meaning yabdū ‘it is obvious’ seems more obvious to us [i.e. “obvious in opinion”]; if you were to apply the hamzah to it, then you would intend the meaning ʔawwal al-raʔy “first/beginning in opinion.” ” If al-Farrāʔ is correct to see the two readings as intending two different meanings, this obviously still stems from an ambiguity of the text which only became ambiguous when readers started to add the hamzah to the recitation of the Quran.

6.5.4 Pre-consonantal Hamzah

Among the canonical readers, the dropping of hamzah in pre-consonantal position is by far the most common, because it is a regular practice in a restricted form with Warš, and mostly unrestricted for ʔabū Ǧaʕfar and ʔabū ʕamr (optional for the latter). For the other readers, however, such dropping of the hamzah is not regular, but despite that, it is occasionally attested in isolated words among the other canonical readers.

At the start of this section (§ 6.5) we already mentioned that al-Kisāʔī read aḏ-ḏīb ‘the wolf’ without hamzah. Xalaf joins him in this reading. Warš likewise drops the hamzah in this word, but adds to it also bīr ‘well’, and bīsa, bīsa-mā ‘how bad!’ (Ibn al-Ǧazarī § 1471–1472). Also, the reduplicated noun luʔluʔ, Šuʕbah ʕan ʕāṣim goes against his general principles reads lūluʔ in all its attestations (Ibn al-Ǧazarī § 1482). Qālūn ʕan Nāfiʕ and Ibn Ḏakwān ʕan Ibn ʕāmir read ‮ريا‬‎ (Q19:74) as riyyan rather than riʔyan (Ibn al-Ǧazarī § 1483). Finally, Qālūn ʕan Nāfiʕ, with disagreement among his transmitters, reads al-mūtafikah (Q53:53) and al-mūtafikāt (Q9:70; Q69:9) ‘that which is overthrown’, rather than muʔtafikah/muʔtafikāt (Ibn al-Ǧazarī § 1482).

‮التين‬‎ ‘the fig’ (Q95:1) is an example where all readers are in agreement that the word is to be read as at-tīn, without hamzah, whereas from an etymological perspective, it seems that this word should have had a hamzah. Hebrew ‮תאנה‬‎30 and Syriac ‮ܬܐܬܐ‬‎ tettā both point to a reconstruction *tiʔn-(at-), which would be expected to just yield tiʔn in varieties that retained .

The suppletive imperative hātū ‘give!’ lacks a hamzah among all readers where it would be expected to exist. The verb is transparently historically derived from an imperative of the C-stem of √ʔty, i.e. ʔātā ‘to give’, and it still inflects as an imperative of this type in Classical Arabic hāti (; hātī (; hātiyā (du.); hātū (; hātīna ( In the Quran only the masculine plural hātū is attested (Q2:111; Q21:24; Q27:64; Q28:75).

The initial h is an ancient retention of the Central Semitic C-stem, which had an *h- as can be seen, for example in the Hebrew C-stem that has the shape hip̄ʕel. So, where the causative in Classical Arabic is expected to be ʔāti < *ʔaʔti, the form hāti developed from a form with retained *h-, i.e. *haʔti.31 While Classical Arabic undergoes a dissimilatory process of *ʔvʔ > ʔv̄ that can explain the lack of the glottal stop in the regular imperative ʔāti, this same sound law cannot be used to explain the absence of the glottal stop in hāti, which has irregularly lost the hamzah in Classical Arabic as well as all reading traditions. This form probably developed because the form hāt(i) was not transparently analysable to the speakers anymore as coming from the root √ʔty, and thus the hamzah could not be reinstated.

6.5.5 Interchange between III-w/y and III-ʔ Verbs

In most, if not all, modern Arabic dialects, III-w/y and III-ʔ merge completely. This merger is already well on its way in the language of the QCT. In the imperfect, the verbs appear to have been indistinguishable from final weak verbs, and in the imperative and jussive, etymological III-ʔ verbs behave as III-y verbs three of the seven times they occur (see Appendix A.4.13). The result of this partial merger has also led to disagreement between the readers as to whether a verb form should be treated as a III-ʔ or a III-y verb.

Most conspicuous of the verbs that show this disagreement is the verb ʔarǧaʔa/ʔarǧā ‘to postpone’ forms of which occur throughout the Quran, with clear disagreement between the readers (Ibn al-Ǧazarī, § 1229; § 1533).






















‮ارجه‬‎ (Q7:111; Q26:36)







‮مرجون‬‎ (Q9:106)




‮ترجى‬‎ (Q33:51)




Leaving the unusual treatment of the pronominal suffix of ‮ارجه‬‎ aside for now (for a discussion on that see § 7.1.8), there is a mostly regular split: The Damascene, Meccan and Basran readers treat the verb as a III-ʔ root, whereas the Medinans and Kufans treat it as a III-y root, with the exception of Šuʕbah ʕan ʕāsim who has a mixed paradigm where the imperative is III-y and the other forms III-ʔ. As I have found no cognates of this verb in other Semitic languages, it is difficult to be sure whether the form with the hamzah is the original form, or a pseudocorrection.

Another verb that shows disagreement between the readers are derivations from waṭiʔa ‘to tread, step on’. ʔabū Ǧaʕfar in accordance with a general rule of his reads ‮يواطوا‬‎ (Q9:37) as yuwāṭū rather than yuwāṭiʔū as the rest. However, the forms of the G-stem ‮يطون‬‎ (Q9:120), ‮تطوها‬‎ (Q33:27), ‮تطوهم‬‎ (Q48:25) he reads as yaṭawna, taṭaw-hā and taṭaw-humū respectively, where the other readers read yaṭaʔūna, taṭaʔū-hā and taṭaʔū-hum(ū). These forms are not the regular outcome of his general hamzah loss rules. Other verbs of this type simply retain the hamzah, e.g. ‮يقرون‬‎ (Q10:94; Q17:71) as yaqraʔūna. He also reads ‮موطيا‬‎ ‘step’ (Q9:120) as mawṭiyan rather than mawṭiʾan. ʔabū Ǧaʕfar is inconsistent on the treatment of this sequence, cf. ‮سييا‬‎ sayyiʔan (Q9:102) but ‮خاسيا‬‎ xāsiyā (Q67:4).

However, he treats ‮وطا‬‎ ‘impression’ (Q73:6) as a III-ʔ stem, reading waṭʔan. This is rather surprising as wiṭāʔan is also consistent with the rasm and would have matched the treatment of this root as both III-y and III-ʔ, and this is in fact how ʔabū ʕamr and Ibn ʕāmir read (Ibn al-Ǧazarī, § 4467).

6.5.6 Sāla for saʔala (Q70:1)

Nāfiʕ, ʔabū Ǧaʕfar and Ibn ʕāmir read ‮سال‬‎ in Q70:1 (and only there) as sāla, with the expected hamzah not reinstated, which they do have elsewhere in their reading. Thus, for both of them ‮سالك‬‎ is read as saʔala-ka in Q2:186, for example (Ibn al-Ǧazarī, § 4441).

6.5.7 Šurakā-ya (Q16:27) for al-Bazzī ʕan Ibn Kaṯīr

Most readers are in agreement that ‮شركاى‬‎ (Q16:27) should be read with hamzah, šurakāʔ-iya ‘my partners’, but al-Bazzī ʕan Ibn Kaṯīr (with disagreement among his transmitters) reads it as šurakā-ya, treating this plural as a ʔalif maqṣūrah rather than an ʔalif mamdūdah (according to Ibn Muǧāhid, 371, and al-Dānī al-taysīr, 137, but not according to Ibn al-Ǧazarī, § 3417). This is not the regular behaviour of al-Bazzī with this noun. In fact, even the other cases of the phrase ‮شركاى‬‎ ‘my partners’ (Q18:52; Q28:62, 74; Q41:47) are read by al-Bazzī as šurakāʔ-iya.

6.6 Conclusion

In the above sections we have examined the position of the hamzah among the canonical readers. As is clear from this discussion we can find ample examples both of the application of hamzah where it is evidently pseudocorrect and places where the reading traditions lack hamzah where their regular rules would not predict it. These findings show that Blau’s assertion that there is no trace of pseudocorrection in the Quran is incorrect. Besides a good number of pseudocorrect hamzahs, we also find many examples where the readers fail to insert the hamzah where we would expect it. This combined with reports of introduction of hamzah in the second century (at least in Medina) suggests that application of the hamzah into the text was part of the goals of the Quranic readers. These readers would not always have the means to do this correctly, sometimes overzealously applied it to words that certainly never had it, and in other cases refrained due to uncertainty.

Of course, this does not show that the language was composed without hamzah, that evidence can only be retrieved from Quranic rhyme and orthography. What it does show is that the Quranic reading traditions cannot be taken as a reliable guide for the language of the Quran in this regard. The readers were actively trying to apply hamzah in what they considered to be the correct way (mistaken or not), and there is no indication that these attempts had anything to do with what the situation was in the original composition. As such, the presence and pervasive use of hamzah in the Quranic reading traditions cannot inform us as to what the treatment of the hamzah in the original language of the Quran was.


The interpretation of the final wāw-ʔalif sequence in what in Classical Arabic is pronounced al-ʕulamāʔu remains somewhat difficult to determine. It seems fairly clear that it does not represent āwu or āw. ō seems like a reasonable option. See Appendix A.2.3.6 for a discussion.


See also al-Ḏahabī (I, 59); Ibn al-Ǧazarī (al-Ġāyah, II, 260).


Optionally with an overlong vowel triggered by the following, now dropped hamzah, or without the length.


Vollers (1906, 95) sees the hamzah as the transitional stage between an original *ḍiwāʔ and the form ḍiyāʔ. There is no reason to assume nor is there evidence that such a transitional stage took place.


This line is missing in the first edition of this text, but the third edition has this line added.


Arabic lexicographers appear to have been aware of the weakness of this reading, as, for example Lisān (2540c) lists ḍuʔzā first, then ḍūzā (the expected form if one would drop the hamzah) and only then ḍiʔzā and ḍīzā respectively.


See Brockelmann (1908, 138 f.) for a discussion on this development, which has striking similarities with a development as found in Aramaic (see also van Putten 2020a, 61). Note however that this development cannot be reconstructed for Proto-Arabic, as varieties of Old Arabic still retain the glide in such places, e.g. Safaitic s¹my /samāy/ ‘sky’, ḫyṯ /xāyeṯ/ ‘travelling’, gyʿ /gāyeʕ/ ‘starving’. Moreover, the shift does not seem to have taken in several dialects of Yemen, where we find forms such as samāy ‘sky’, ʔalḥāy ‘jaws’ (pl. of liḥi) and ʕamyāy ‘blind’ (cf. CAr. samāʔ, ʔalḥāʔ and ʕamyāʔ) (Behnstedt 1987, 59–61).


Vollers (1906, 95) takes the plural maʕāʔiš as a pseudocorrection. Fück (1950, 39 f.), rather prescriptively, considers the reading maʕāʔiš a mistake and evidence that there was a lack of a developed grammatical school in Medina.


qawluhū “‮معيش‬‎”, kulluhum qaraʔamaʕāyišabi-ġayri hamz. Wa-rawā Xāriǧatu ʕan Nāfiʕinmaʕāʔišahmamdūdatan mahmūzah. Wa-Qāla ʔabū Bakrin: wa-huwa ġalaṭ. (Ibn Muǧāhid, 278).


Note that it is synchronically correct to consider this noun to be from a root √mdn in Arabic, as can also be seen from the other plural mudun, but ultimately in Aramaic, from which the word stems, mḏīnṯā ‘province, city’ is a noun of place of the root √dyn ‘to judge’ (Jeffery 2007, 260).


I thank Hythem Sidky for pointing me to this reference.


ʔabū Ḥayyān is citing al-Farrāʔ (Maʕānī, I, 373) whose wording is slightly different in the edition we have.


Larcher (2021, 49, n. 40) suggests that the “Classical Arabic” form of this name has the hamzah. This is a typical example of the imposition of modern norms onto the opinions of the Arab grammarians. Both Sībawayh (III, 394) and al-Farrāʔ (Luġāt, 47; Maʕānī, I, 208) explicitly state that this name may be pronounced Zakariyyā or Zakariyyāʔu with no normative preference for one over the other. Incidentally, as there is no reason to consider Zakariyyāʔu as more original, it is of course incorrect to take its appearance in Q19:3 in rhyme as evidence that word-final āʔ had lost its hamzah, in Quranic Arabic. It simply never had it, unlike the examples I adduce of āʔ that does seem to rhyme with words that end in a final consonant, and are words that derive from ancient *āy sequences that shifted to āʔ (van Putten 2018, 103–105).


Presumably ʔabū ʕaliyy al-Fārisī (d. 377 AH) a student of Ibn Muǧāhid (Ibn al-Ǧazarī al-ġāyah, I, 189). While ʔabū ʕaliyy discusses these variants in detail in his Ḥuǧǧah (IV, 109–111), nowhere does he call the hamzated forms weak.


In one of the two places that this line is cited, this form is vocalized al-muʔqidayni, but I would not know who these two kindlers would be.


An interesting exception appears to be Aramaic ʕānā ‘sheep’, which has lost the ʔ already in Official Aramaic times, spelled ‮קן‬‎ where Hebrew ṣōn ‮צאן‬‎, Arabic ḍaʔn and Akkadian ṣēn point to a reconstruction *ṣ́aʔn. This is probably the result of a dissimilation of the two guttural consonants occurring in a row.


As already recognized by Vollers (1906, 94). Vollers also noticed that such pseudocorrect forms entered the classical language through other channels than Quranic recitation, this is clear from the variable bāz/baʔz ‘falcon’, which, considering that it comes from Persian bāz, must certainly be considered a pseudocorrection as well.


This is the result of a fairly regular rule in the ʕarabiyyah that sequences of *wu or *wū become ʔu, therefore the plural ʔaswuq is expected to shift to ʔasʔuq. Note that this explanation is unable to make sense of the reading of the dual with hamzah, or in fact the other plural suʔq/suʔūq.


ʔabū Ǧaʕfar also reads as-ṣābū/īna, but this is part of his regular pattern of dropping the hamzah (see section § 5.2).


The Mandaic form is likely ultimately from a root *ṣbġ which yields Ar. ṣabaġa ‘to dye, baptize’ and Aramaic ṣḇaʕ with the typical loss of the gutturals of Mandaic.


Clearly meant as an insult, weavers were despised in medieval Islamic society, a position they share with the textor of Roman times (EI2 s.v. ḥāʾik).


Another humorous story about al-Kisāʔī’s reading of ḏiʔb as ḏīb exploits the polysemy of the verb hamaza which means both ‘to apply the hamzah’ and ‘to prod’. Someone asked al-Kisāʔī: lima lā tahmizu ḏ-ḏīb? “why do you not hamzate/prod the wolf?”. To this al-Kisāʔī answers: ʔaxāfu ʔan yaʔkula-nī! “I am afraid that it would eat me!” (al-Ḏahabī, 300) where al-Kisāʔī is playfully riffing on the verse in which his reading aḏ-ḏīb occurs: wa-ʔaxāfu ʔan yaʔkulahu ḏ-ḏīb “I fear that a wolf will eat him” (Q12:13).


Ibn Muǧāhid seems to have understood this quote as meaning that one is indeed to pronounce these words as biʔr and ḏiʔb, which reveals a significantly developed view of what ‘Classical Arabic’ is in the late third/early fourth century AH. To him what ‘Arabs’ say is clearly the form with hamzah. But one wonders if Nāfiʕ truly meant it in such a manner. To Sībawayh, for example, bīr and ḏīb are acceptable and certainly also something that ‘the Arabs’ say (Sībawayh, III, 541 ff.).


This plural is in fact attested in a poem by al-ʕabbās b. Mirdās (d. ca. 18–35 AH) starting with yā xātama n-nubaʔāʔi ʔinnaka mursalun “O seal of the prophets, you are sent” (al-Ǧabbūrī 1968, 95), another piece of evidence that the form nabīʔ pl. nubaʔāʔ indeed existed in the ʕarabiyyah, also outside of Quranic recitation.


For example, it is a regular plural (besides CaCāCin) for nouns that end in the feminine endings and -āʔ, e.g. fatwā ‘legal opinion’, pl. fatāwin, fatāwā and ʕaḏrāʔ ‘virgin’ pl. ʕaḏārin, ʕaḏārā (Fischer 2002, § 99a).


The expected for xaṭiyyah is attested in Classical Arabic lexicons (Lane 761c; Lisān 1193b).


Al-Zamaxšarī (mufaṣṣal, 167) cites ʔabū Zayd as having heard someone use this original plural in aḷḷāhumma ġfir lī xaṭāʔiʔī “O God, forgive me my sins.”


A few early manuscripts appear to give evidence that in earlier times such readings were more widespread. The vocalization of Arabe 334a’s ‮يروا‬‎ (Q36:31), places a fatḥah sign on the rāʔ which likely denotes the presence of a hamzah, thus suggesting yarʔaw, and the spelling ‮يراى‬‎ in DAM 01.29-1 leaves little doubt the scribe intended yarʔā/yarʔē, as this manuscript frequently employs the ʔalif to denote the presence of hamzah (van Putten 2019a, 370, n. 210).


This should not be considered part of the dissimilation of two consecutive hamzahs as in *ʔaʔimmah > ʔayimmah ‘Imāms’ and *ǧāʔiʔun > ǧāʔin ‘coming’ as suggested by Fischer (2002, § 41a). The dissimilation, at least as described by Sībawayh (III, 552) is always progressive, not regressive, and only occurs if a short vowel intervenes.


The pronunciation tʔenå of the Tiberian reading tradition is likely also pseudocorrect, as the glottal stop is expected to have been lost here.


For an account of the development of the C-stem from *s¹ to *h and ultimately to ʔ in Arabic, see Al-Jallad (yusapʕil). Al-Jallad argues that *s¹ regularly becomes h on word-boundaries in Proto-Central Semitic. For Arabic, *h becomes ʔ in pre-stress position, thus explaining the shift of *him ‘if’ and *hinna ‘verily’ to ʔin and ʔinna. He moreover argues that *hafʕala > ʔafʕala is the result of the same sound shift, drawing upon stress marking in the Damascus Psalm fragment to argue that the stress of C-stems was *ʔafʕála in Proto-Arabic, something that would be corroborated by Hebrew which likewise carries the stress in this position. If this is correct, it would seem that the imperative form of the verb of *haʔti carried the stress on the penultimate, i.e. *háʔti, which would explain the retention of the *h in this position.

  • Collapse
  • Expand