Al-ʿUṣūr al-Wusṭā 29 (2021): 1-19 The Digraph اى in the Quranic Consonantal Text and the Identification of a New Letter Shape for Final Hē in the 7th to 8th Century Arabic Script* AhmAd Al-JAllAd The Ohio State University (al-jallad.1@osu.edu) 1. Introduction The Quranic consonantal text1 (henceforth QCT) contains several layers of historical Arabic orthography. Its foundation lies in the orthographic principles of the imperial * I owe a great debt to Marijn van Putten for his help in locating examples of the orthographic peculiarities discussed in this paper and for finding supporting attestations in the manuscripts he is currently studying. I also thank Hythem Sidky for pointing me toward relevant examples in early Quranic codices. I made a short Facebook post on the identification of the new shape of the final hē in a Quranic manuscript on March 30, 2018, and I thank Yüsef Gürsey and Mila Neishtadt for discussing the further applications of this letter shape with me there. All errors are my own. Linguistic conventions: * = reconstructed pronunciation C = consonant; V = vowel / / = phonemic transcription Details of cited pre-Islamic Arabic-script inscriptions are given in an index following the main body of this article. 1. This term refers to the theory that the extant Qurans go back to a single written archetype, conventionally labeled the ʿ Uthmānic Codex, and that the text was composed in a single dialect of Arabic. For the full elaboration Abstract This paper proposes a hitherto unrecognized orthographic practice in the Quranic consonantal text: use of the digraph اى, that is, alif + denticle, to represent the noninitial glottal stop, most often adjacent to the high vowels i/ī and less commonly in other environments. This feature leads to the identification of a new letter shape for the final hē in the early Islamic Arabic hand, originating in the Nabataeo-Arabic script, which in turn can explain a number of previously enigmatic spellings in the Quranic consonantal text. © 2021 Ahmad Al-Jallad. This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License, which allows users to copy and distribute the material in any medium or format in unadapted form only, for noncommercial purposes only, and only so long as attribution is given to the original authors and source. mailto:al-jallad.1%40osu.edu?subject= 2 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) Aramaic writing tradition,2 inherited from Arabic’s Nabataean forebear. These were followed by a variety of innovative spelling strategies that emerged after the Nabataean script was fully applied to the representation of the Arabic language.3 These innovations do not necessarily originate in the same place and time. The ever-growing corpus of pre-Islamic Arabic-script inscriptions suggests that the Nabataean Aramaic script did not develop directly into a single Arabic script but rather produced several lineages of Arabic scripts with their own orthographic practices and, sometimes, letter shapes.4 The establishment of Arabic as the language of an empire in the seventh century produced a homogenizing bottleneck, narrowing much of the variation found in pre-Islamic sources. Older and newer orthographic practices exist side by side in the QCT and, to a lesser degree, in Classical Arabic orthography. The notation of internal long ā illustrates this phenomenon. There is no attempt to indicate the long vowel internally in the extant sixth-century Arabic-script inscriptions, and there was no mater lectionis for internal ā in the Nabataeo-Arabic and Nabataean scripts that preceded Arabic.5 The Quran in general agrees with this practice: long ā is rarely indicated word-internally, with the exception of words belonging to the CāC pattern.6 A newer, plene orthographic practice that indicates ā with alif in other environments seems to have emerged sometime in the seventh century. As van Putten has observed,7 the use of this new orthographic practice appears to be optional in the QCT: several words are written in both ways, sometimes according to the ancient orthography and other times with the innovative plene spelling. Certain core of this theory, see the introductions to M. van Putten, “The Development of the Triphthongs in Quranic and Classical Arabic,” Arabian Epigraphic Notes 3 (2017): 47–74; idem, “The Feminine Ending -at as a Diptote in the Qurʾānic Consonantal Text and Its Implications for Proto-Arabic and Proto-Semitic,” Arabica 64, nos. 5–6 (2017): 695–705; idem, “‘The Grace of God’ as Evidence for a Written Uthmanic Archetype: The Importance of Shared Orthographic Idiosyncrasies,” Bulletin of the School of Oriental and African Studies 82, no. 2 (2019): 271–88. 2. W. Diem, “Untersuchungen zur frühen Geschichte der arabischen Orthographie: I. Die Schreibung der Vokale,” Orientalia, n.s., 48 (1979): 207–57, at 209–10. 3. This second layer is often called Ḥigāzī–Meccan in the literature; see W. Diem, “Some Glimpses at the Rise and Early Development of Arabic Orthography.” Orientalia, n.s., 45 (1976): 251–61, at 255. However, as we shall see in the following discussion, the orthographic innovations of this layer do not stem from a single source or period. 4. This idea is developed in A. Al-Jallad, “‘Moge God Yazīd de Koning Indachtig Zijn’: Nadere Beschouwingen over de Yazīd-Inscriptie en de Ontwikkeling van de Arabische Schriften,” in Mohammad en de Late Oudheid, ed. J. van den Bent, F. van den Eijnde, and J. Weststeijn, 198–208 (Amsterdam: Verloren, 2018). 5. On the development of the Arabic script from its Nabataean forebear, see L. Nehmé, “A Glimpse of the Development of the Nabataean Script into Arabic Based on Old and New Epigraphic Material,” in The Development of Arabic as a Written Language, ed. M. C. A. Macdonald, 47–88 (Oxford: Archaeopress, 2010). On Nabataean orthography, see J. Cantineau, Le Nabatéen (Osnabrück: Zeller, 1978); and on the orthography of Arabic words, especially those in the inscriptions of Ḥegrā, see J. Healey, The Nabataean Tomb Inscriptions of Mada’in Saleh (Oxford: Oxford University Press, 1993). 6.  For example, the verb kāna is spelled كان more often than it appears, defectively, as كــن. Exceptions do exist: qāla is sometimes spelled قل in early manuscripts. 7.  For example, we find both عبده (Q 27:59) and عباده (Q 35:28) in the Cairo edition, and even greater variation once we compare spellings with internal ā across earlier manuscripts; see M. van Putten, Quranic Arabic: From its Hijazi Origins to its Classical Reading Traditions (Leiden: Brill, forthcoming). Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 3 vocabulary items of the writing tradition, however, are not affected by the new spelling and continue to be written according to the old orthography, even to the present day; examples include the demonstratives ‘this’ ــذا ــك ’hāḏā/, ‘that/ ه ــن ḏālika/, and the divine title/ ذل / رحم raḥmān/. This paper identifies a hitherto unrecognized orthographic practice in the QCT, which perhaps emerged in the period before the loss of the glottal stop8 and was phased out by the more phonetic writing principles of the main orthography of the Quran: use of the digraph that is, alif + denticle, to represent the noninitial glottal stop, most often adjacent to the ,اى high vowels i/ī and less commonly in other environments. This interpretation leads to the identification of a new letter shape for the hē in the early Islamic Arabic hand, originating in the Nabataeo-Arabic script, which in turn may explain a number of previously enigmatic spellings in the QCT. 2. Identifying the اى Digraph The following table presents all the examples known to me of the use of the digraph in the Cairo Edition and early Quranic manuscripts, along with the pronunciation of the words in which it occurs through various stages of the Arabic language. Table 1: The Attestations of the Digraph اى in the Cairo Edition & Early Quranic Manuscripts9 QCT spelling Proto-Arabic pronunciation Classical Arabic pronunciation Reconstructed QCT pronunciation10 جاىت11 *giʾtu ǧiʾtu ǵīt جاي12 *gīʾa ǧīʾa ǵī شاىت13 *śiʾtu šiʾtu śīt شاى14 *śVyʾun šayʾ śī 8.  Classical Nabataean orthography indicated the Arabic glottal stop with aleph, ʾ, regardless of the quality of the following vowel. The loss of the use of alif to represent the glottal stop, hamz, in the QCT results from the loss of that phoneme in its dialect; M. van Putten, “Hamza in the Quranic Consonantal Text,” Orientalia 86 no. 3 (2018): 93–120. As Diem (“Glimpses,” 254) has suggested, there must have been a local writing tradition of Arabic in the Ḥijāz that devised an orthography closer to the pronunciation of the local dialect, without the hamz. Nevertheless, these new spellings remained in some cases in competition with the older practice of representing the etymological glottal stop with alif. For example, the word ḍuʿafāʾu ‘weak’ (masculine plural) is spelled both according to the old orthography as ضعفا (Q 9:91) with the final etymological glottal stop represented by the alif and according to an innovative, phonetic spelling as ضعفوا (Q 14:25; 40:47) with the glide resulting from the loss of the glottal stop in pronunciation represented, /ḍuʿafāw(u)/. 9. Manuscript sigla follow https://corpuscoranicum.de/. 10. This reconstruction is based on van Putten, “‘Grace of God.’” 11. Q 19:27, in MS Tübingen, Universitätsbibliothek, Ma VI 165, fol. 12r. 12. Q 39:69; Q 89:23. 13. Q 24:62, in MS Tübingen, Universitätsbibliothek, Ma VI 165, fol. 37v. 14. Q 18:23. https://corpuscoranicum.de/ 4 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) QCT spelling Proto-Arabic pronunciation Classical Arabic pronunciation Reconstructed QCT pronunciation باىىته15 *biʾayāti-hu biʾayāti-hī biyayātVh الىىتنا16 *liʾayāti-nā li-ʾayāti-nā liyayāt(V)nā باىىم17 *biʾawyāmin biʾayyāmin biyayyām باىى18 *biʾayyi biʾayyi biyayy باىىد19 *biʾaydin biʾaydin biyayd ماىه20 *miʾata miʾatin mīyah باىىكم21 *bi-ʾayyi-kum(u) bi-ʾayyi-kum biyayyikum ساى22 *suyiʾa (or śīʾa) sīʾa sī (or siyy) سايل23 *suʾila sūʾila sīla فاين24 *fa-ʾin fa-ʾin fayin ياىس25 *yīʾasu yayʾasu yVyas تاىسوا26 *tīʾasū tayʾasū tVyasū السواى27 *sūʾan as-sūʾa as-sū راىى28 *ruʾyayi-ya ruʾyā-ya riyyāy اليلف29 *li-ʾilāfi li-ʾilāfi (or li-ʾīlāfi) liyilāfi 30 مالىهم31، مالىه *malaʾi-hū̆, *malaʾi-hum malaʾi-hī, malaʾi-him malayi-h/hum نباى32 *nabaʾi nabaʾi nabay(i) 15. Q 6:35. 16. Saray Medina 1a; Großer Korankodex Q 74 :16. 17. Q 14:5. 18. Passim, Surat al-Raḥmān, MS Paris, BnF, Arabe 331, fol. 48v. 19. Q 51:47. 20.  Passim in the Cairo Edition, but sometimes spelled مىه in other manuscripts, e.g., Samarqand Q 8:66. 21. Q 68:6. 22.  Samarqand Q 11:77; compare with the Cairo Edition: سى. 23.  Samarqand Q 2:108; compare with the Cairo Edition: سىل. 24. Q 3:144; 21:34. 25. Q 12:87. 26. Q 12:87. 27. Q 30:10. 28. Passim, Q 12:43, 100, MS London, British Library, Or. 2165, fols. 25r, 27r. 29. Q 106:1; for this analysis, see van Putten, “Hamza,” 110. 30. Q 7:103; 10:75; 11:97; 23:46; 28:32; 43:46. 31. Q 10:83. 32. Q 6:34. Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 5 3. Discussion This section describes the distribution of digraph اى. Its relatively limited use suggests that it is an orthographic relic rather than a productive feature of the spelling system used to write the earliest Qurans. śiʾtV33* = شاىت ;giʾtV* = جاىت .3.1 The third-person masculine singular of both of these verbs terminates in an alif in the QCT, جــا reflecting ǵā from *gāʾa and ســا reflecting śā, from *śāʾa, respectively. The first/ second-person form of these verbs in the synchronic dialect of the Quran, which had lost the glottal stop, must have been realized as ǵīt and śīt, respectively, as in many modern dialects. The marginal spelling of these verbs with the digraph, however, indicates that the linguistic source of their orthography did not realize them with a simple medial ī. I suggest that this spelling emerged at a time when the glottal stop was still pronounced and the sequence اى was used to represent the noninitial glottal stop, in this case following an i-vowel. miʾah* = ماىه .3.2 The enigmatic spelling of miʾah has been the subject of much debate, but no consensus has been reached as to what the alif-yā sequence is meant to signify.34 The pronunciations māʾah and miʾāh, found in some vocalized Quranic manuscripts, most certainly reflect artificial, secondary vocalizations based on the synchronic interpretation of the word’s orthography.35 There is no etymological basis, from a comparative Semitic perspective or in the modern and ancient dialects of Arabic, for the vocalization of this word as anything other than miʾah and, following the loss of the glottal stop, as mīyah. In light of the discussion on above, the most natural explanation for this spelling is that it in fact reflects شــاىت and جاىــت miʾah, using the اى digraph for the glottal stop following the i-vowel. 3.3. Bi- and Li- before Words Beginning with a Glottal Stop Several words with an initial glottal stop following the prepositions bi- and li- are written with the digraph, the same environment as above.36 33. These orthographic oddities were first recognized and commented on by Marijn van Putten on Twitter (https://twitter.com/PhDniX/status/976754498151514112), although he attempted no explanation of them. 34. The most widely cited opinion is that of Diem, “Untersuchungen,” 102—namely, that the alif is preserved as a graphic archaism, and the yā following it reflects the contemporary pronunciation, miyah. This idea is followed by van Putten (“Hamza”), who terms it a mixed etymological spelling. 35. On these, see H. Sidky, “In Search of Lost Time: A Vocalized Muṣḥaf, Ibn ʕāmir, and the Evolution of the Syrian Reading Tradition,” forthcoming. 36. For previous treatments of this spelling, see van Putten, “Hamza,” 109–11, and references therein. https://twitter.com/PhDniX/status/976754498151514112 6 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) Table 2: The Digraph Following Prepositions ىاىىت ٮ ى اى ٮ biʾāyātin t y ʾ b ىاىىم م ى اى ٮ biʾayyāmin m y ʾ b ىاىى ى اى ٮ bi-ʾayyi y ʾ b ىاىىد د ى اى ٮ bi-ʾaydin d y ʾ b باىىكم كم ى اى ٮ bi-ʾayyikum km y ʾ b As van Putten has pointed out to me (personal communication), it is remarkable that in all such cases, a y occurs later in the word. Perhaps this acted as an orthographic conditioning environment for the preservation of this archaic spelling, at least marginally. Alternatively, it is possible that the following y blocked the loss of the glottal stop in this position in order to avoid the sequence yaya, a sort of dissimilation. If the glottal stop persisted longer in this environment than in others, this spelling would reflect a phonetic reality rather than being merely an instance of historical orthography. sī(ʾ) = ساى ;gī(ʾ) = جاي .3.4 A noninitial glottal stop following the long vowel ī is spelled using the digraph in the passives of *gāʾa ‘to come’ جــاى *gīʾa, and of *sāʾa ‘to be evil’ ســاى *sīʾa. śīʾ = شاى .3.5 The reading traditions of the QCT vocalize the word for ‘thing’ only as šayʾ, even when it is spelled ســاى. The modern dialects suggest that a by-form belonging to the fiʿl pattern, rather than faʿl, existed; the interchange of these two patterns is well attested.37 In the dialect of Beirut, for example, the diphthong /ay/ remains intact; Proto-Arabic *baytun produces bayt, while the word for ‘thing’ is realized as šī. This must go back to an earlier *śīʾ rather than *śayʾ. The Quranic spelling, in light of the previous identifications, suggests that its original dialect also exhibited a reflex of the fiʿl form, synchronically pronounced as śī and historically as *śīʾ; the orthographic form ســاى reflects the latter. 37. Faʿl/fiʿl by-forms are well attested in Arabic; for example, watrun, witrun ‘single’; kasrun, kisrun ‘bone with meat’; salmun, silmun ‘reconciliation.’ J. Fox, Semitic Noun Patterns (Winona Lake, IN: Eisenbrauns, 2003), 133. Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 7 3.6. The Spelling of the Glottal Stop with اى before an I Vowel Although the most common environment for the spelling of the glottal stop with the digraph is after the vowel i, there are a few examples of it spelled as such in a closely related environment, before i: ســاىل *suʾila; فاىــن /fa-ʾin/; and the case forms مالىــه /malaʾi-hu/ and ــاى ./nabaʾi/ نب In at least one clear case, the noninitial glottal stop is rendered with اى: as-sūʾā is spelled in Q 30:10, reflecting its complete emancipation from its original conditioning الســواى environment. However, one must note that the pronunciation of the alif maqṣūrah in the QCT was ē, and so the digraph may have been motivated by its proximity to this vowel, which was perhaps felt to be close to ī. The spelling of ruʾyā-ya with the digraph similarly points to the digraph’s use outside of the environment of /i/, although again the contiguous y may have played a role. Also remarkable about the spelling الســواى is the fact that the y of the digraph and the y of the alif maqṣūrah are treated as one, as in words terminating in y followed by a nisba ending, such as النبيــن for al-nabiyyīna (Q 2:61). The spellings of the verbs yayʾasu and tayʾasū as ياىــس and تاىســوا, respectively, could also reflect the use of the digraph outside the context of an i-vowel. But this is the case only if the vocalization of these verbs followed the Classical Arabic pattern in the dialect of the orthography. Prefix-conjugated verbs with the theme vowel /a/ (the vowel of the verb stem) have an i-vowel in the preformative prefix—the so-called Barths-Ginsberg law.38 If this law was operative in the dialect that gave rise to this spelling, these verbs would have been pronounced as yīʾas and tīʾasū, the exact environment in which we would expect to encounter this spelling. 4. The Background The evidence assembled above demonstrates that in the earliest stratum of Quranic orthography scribes had the option to employ the digraph اى to represent the noninitial glottal stop. The digraph was used most often after an i-vowel and marginally before one, and perhaps only once outside of that environment. The optional use of an orthographic device has several precedents in the Quran. As noted in the introduction, the representation of internal ā with alif was almost entirely optional outside of the CāC environment.39 But in our case, the digraph اى, while certainly an innovation from the Nabataean perspective, must be construed as an archaism with regard to the synchronic dialect of the Quran and its 38. That is, yafʿulu and yafʿilu but yifʿalu. on this law and its distribution in the Central Semitic languages, see J. Huehnergard, “Features of Central Semitic,” in Biblical and Oriental Essays in Memory of William L. Moran, ed. A. Gianto, 155–203 (Rome: Pontificio Istituto Biblico, 1995). This law is still operative in the Najdi dialects of Arabic today; see B. Ingham, Najdi Arabic: Central Arabian (Amsterdam: J. Benjamins, 1994). And it seems to have been operative in the northern Old Arabic dialects as well, as evidenced by the spelling ειραυ for yirʿaw in an Arabic inscription written in Greek letters from the Jordanian Ḥarrah; see A. Al-Jallad and A. al-Manaser, “New Epigraphica from Jordan I: A Pre-Islamic Arabic Inscription in Greek Letters and a Greek Inscription from North-Eastern Jordan,” Arabian Epigraphic Notes 1 (2015): 51–70. 39. A notable exception is the spelling of I-ʾ verbs, which always have the alif. I thank Marijn van Putten for pointing this out to me. 8 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) primary writing tradition, which had lost the glottal stop. This spelling must therefore stem from an orthographic tradition that, on the one hand, preserved the glottal stop in these environments but, on the other, had lost the means of using alif to represent it. How did this come to be? Digraphs are not a component of Semitic alphabetic writing in general, so this practice was clearly not inherited directly from any antecedent script. I suggest that the digraph emerged in a “script-contact” situation. In order to fully appreciate this phenomenon, however, we should give some attention to the diversity of the Arabic script in the sixth century CE. The corpus of sixth-century Arabic inscriptions and even earlier Nabataeo-Arabic inscriptions exhibits variation in both letterforms and orthographies. Focusing on just the sixth-century inscriptions, the following variation is attested: The orthography of the word for ‘I’: In the Ḥarrān inscription, the first-person pronoun is spelled ʾnʾ /ʾanā/, while in the Jebel Usays inscription it is spelled ʾnh. This latter spelling is found in an unpublished pre-Islamic Arabic-script inscription from the Tabūk area as well. The spelling ʾnh most certainly goes back to the Nabataean rendering of this pronoun in Aramaic, which survives even into the Nabataeo-Arabic period as evidenced by the Thaʿlabah inscription.40 The use of wawation: Nabataean orthography marked final triptotic personal names/ nouns with a w, the so-called wawation. In Nabataeo-Arabic, this was generally retained in personal names, but in the sixth-century (and later) inscriptions, wawation is used differently from text to text. In the early Arabic-script inscriptions from Najrān, wawation is deployed in the expected way, whereas in the Jebel Usays inscription this feature is missing altogether. In the Yazīd inscription, wawation is applied to the diptotic name Yazīd, suggesting an expansion of its usage.41 Letter shapes: Dots on dāls and lunate rēs: The Nabataeo-Arabic inscriptions occasionally added a supralinear dot to the dāl, a relic of a phase when the glyphs for dāl and rē were identical; the dāl was distinguished by a dot on top, in contrast to the Syriac tradition, where the dot for the dālet was added below. The Yazīd inscription exhibits these dots on the dāls, while other sixth-century Arabic inscriptions do not. Moreover, we find two types of rēs in the sixth-century inscriptions: a lunate form, as in the Jebel Usays inscription, and a linear rē, as found in the Ḥimà inscriptions and the sixth-century Arabic-script inscription of Dūmat al-Jandal.42 These examples of diversity indicate that there was no unified orthography for Arabic in the pre-Islamic period. Different traditions must have evolved locally where the Nabataeo- Arabic script was used. The chancelleries of different oases and tribal rulers could have 40. On this text, see U. Avner, L. Nehmé, and C. Robin, “A Rock Inscription Mentioning Thaʿlaba, an Arab King from Ghassān,” Arabian Archaeology and Epigraphy 24, no. 2 (2013): 237–59. 41. See Al-Jallad, “Yazīd-Inscriptie,” 197–98. 42. Ibid., 199–200. Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 9 developed their own particularities of writing, and these coexisted until the emergence of the Umayyad state and the top-down unification of the Arabic script.43 Script contact must be assumed for this period, as writing was used to send letters not only to one’s own kinsmen but also to people from neighboring groups, who may have had slightly different ways of writing Arabic. This situation forms the context for the scenario I develop below to account for the emergence of the digraph اى. In principle, orthographic developments emerge in an environment that allows for the reanalysis of a fixed, older spelling to produce a new one—a point of diffusion. If we look to the distribution of this digraph in the QCT, it is clear that it was most associated with the sequence iʾ, and indeed, the only word that is more often than not written with the digraph, even to this day, is ماىــه ‘one hundred.’ This stability suggests that the spelling of this word was fixed and widely adopted at a very early point in the history of the Arabic script. The number one hundred is our “patient zero,” so to speak. While the number one hundred is most likely to have been the model upon which the spelling of the glottal stop with اى was based, this does not explain why the digraph was used to represent the glottal stop in this word to begin with. The numeral is well attested in pre-Islamic Nabataean and Nabataeo-Arabic inscriptions, yet in all cases it is spelled mʾh/t, without a denticle, matching its etymology and pronunciation.44 It is at this point that we should turn our attention to the paleography of the final hē in Nabataeo-Arabic. The letter had multiple forms in the Nabataeo-Arabic hand—one form had the denticle of the h on top of its loop, while another form had it on the base line to its right. Figure 1: Shapes of the Hē in Nabataeo-Arabic45 43. On the scenario of the Arabic script evolving gradually from Nabataean at the courts of tribal chiefs in Northwest Arabia, see L. Nehmé, “Epigraphy on the Edges of the Roman Empire: A Study of the Nabataean Inscriptions and Related Material from the Darb Al-Bakrah, Saudi Arabia, 1st–5th Century AD” (Mémoire scientifique d’habilitation à diriger des recherches, École pratique des hautes études, 2013). 44. This evidence is assembled in L. Nehmé, “Aramaic or Arabic? The Nabataeo-Arabic Script and the Language of the Inscriptions Written in This Script,” in Arabic in Context: Celebrating 400 Years of Arabic at Leiden University, ed. A. Al-Jallad, 75–89 (Leiden: Brill, 2017), 88–90. 45. Nehmé, “Glimpse,” 49. 10 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) Figure 2: Example of a Denticled Hē in Nabataeo-Arabic (with the First Line's Final Word Reading ʿsylh)46 The latter form survived well into the period of the sixth-century Arabic script, as we find it in the spelling of the word ʾilāh at least twice. Figure 3: Umm al-Jimāl Inscription (with the First Line Reading ʿbd ʾl-ʾlh)47 Figure 4: Unpublished Graffito from Farīq al-Ṣaḥrā (#5); ʾlh = ʾilāh.48 46. Nehmé, “Glimpse,” UJadh 299. 47. L. Nehmé, “New Dated Inscriptions (Nabataean and Pre-Islamic Arabic) from a Site near Al-Jawf, Ancient Dūmah, Saudi Arabia,” Arabian Epigraphic Notes 3 (2017): 121–64. 48.  The original photograph can be found here: https://alsahra.org/2017/09/نقوش-عربية-بلكنة-نبطيــة/ (accessed June 17, 2021). https://alsahra.org/2017/09/نقوش-عربية-بلكنة-نبطية/ Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 11 To most modern eyes, the spelling of ʾilāh in the two Arabic inscriptions above looks more like ʾlyh, with a denticle before the h. Although it may be tempting to argue for ʾimālah in these cases—that is, a graphic representation of the ē allophone of *ā—we should note that the particular shape of the h, the loop without a tooth on top, prevents us from interpreting the denticle before the loop as a separate letter: it is part of the h. In a lecture I gave on this subject,49 I hypothesized that such a letterform must have persisted into the Islamic period, for reasons we shall see below. Shortly afterward, van Putten kindly shared with me a fascinating discovery he made while studying the pre–750 CE Quranic manuscript DAM 01-29.1 that confirms my hypothesis. The word ʾāḫirah is spelled with a final ىــه, where the hē lacks a tooth on top of the loop and is preceded by a denticle. This can only reflect the ancient shape of the hē discussed above. Figure 5: Al-ʾāḫirah Spelled with a Denticled Hē, االخرىه (DAM 01-29.1, fol. 3v, I. 2 [Q 3:45]) I subsequently noticed the denticled hē in an early Islamic inscription from the area of Medina.50 Although this text is undated, its paleography suggests that it was produced sometime after the second Islamic century, indicating that the digraph survived marginally, and was perhaps restricted to certain formulae. Figure 6: Malāʾikatu-hū (Source: @Mohammed93athar). 49. Delivered at the “Reading the Rasm” workshop held at Leiden University on December 3, 2018. 50. This text was posted online by Abū ʿAbd Allāh al-Mughadhdhawī on his famous Twitter account, Nawādir al-āthār wa-l-nuqūsh (@Mohammed93athar: https://twitter.com/mohammed93athar/status/ 1088434910254247936). https://twitter.com/mohammed93athar/status/ 1088434910254247936 https://twitter.com/mohammed93athar/status/ 1088434910254247936 12 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) 5. Scenario The word “one hundred” was certainly frequent in economic documents and in dating formulae. Since the spelling of this word in pre-Islamic times was always mʾh, what if the denticle of the Quranic ماىــه and that of later Arabic orthography is, at least in origin, no y at all? Could we not interpret this word as simply a continuation of the spelling mʾh, consisting of three letters? If so, then the final denticle, frozen in Arabic orthography, is not a denticle but part of the final shape of the h, similar to the one found in the Umm al-Jimāl inscription and in the Quranic MS DAM 01-29.1. In other words, the shape ىــه is simply one of the various forms of final h in the pre-Islamic Arabic scripts that continued, albeit marginally, into the Islamic period.51 For reasons that are lost to us now, this spelling of “one hundred” gained traction and was frozen graphically as a numerogram, as it is in the orthography of Arabic today. Now, if this gram spread to a writing tradition of Arabic that made use of another form of final h, scribes could have easily mistaken the denticle of the final hē for a separate letter, misparsing and graphically reinterpreting it as ه ى ا .م And since it was pronounced as miʾah, scribes could have inferred that the sequence اى was a way to spell the noninitial glottal stop. The distribution of this spelling suggests that it was always an optional strategy to represent this sound, and that it diffused by analogy to phonetically similar environments, but not exclusively to them. This theory explains the fact that the majority of the digraph’s attestations occur in the vicinity of an i/ī vowel, and usually in the exact phonetic environment of iʾ.52 Why would scribes devise a new way of representing the glottal stop if the Nabataean and Nabataeo-Arabic script already had a method to do so with alif? The answer, I think, lies in understanding the diffusion of Arabic writing traditions. The main orthographic stratum in the Quran clearly reflects a dialect that lost the glottal stop. What if the Arabic script spread from this starting point to a group that retained the stop? This, in fact, has already happened once in the Islamic period, namely, in writing Classical Arabic with Ḥijāzī orthography. Scribes may have abstracted from the spelling of ماىــه a plene method of rendering the glottal stop, a sound their language had but that was not represented in Ḥijāzī orthography. The emergence of a new orthographic practice based on the reinterpretation of a fixed spelling according to synchronic pronunciation finds an exact parallel in the development of another optional spelling strategy, the representation of internal ā with alif. Diem explains the emergence of the use of the alif to spell word-internal ā as follows.53 Words with an etymological glottal stop preceding the short /a/ would have been pronounced as ā, leading to the synchronic interpretation that internal alif signaled the vowel ā. The fact that this spelling in QCT orthography was most consistently applied to CāC nouns seems to imply 51. I thank Mila Neishtadt for pointing out the possible use of the denticled hē for the spelling of the word “one hundred” and both her and Marijn van Putten for the rich discussion with me on my Facebook post of March 30, 2018. 52. The fact that we have no reproductions of this spelling in the exact phonetic environment, namely glottal stop + h, has probably to do with the rarity of this sequence and the limited corpus available to us. 53. Diem, “Glimpses,” 258–59. Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 13 that the point of diffusion was spellings of CaʾC nouns, from which the practice expanded to represent internal ā in all environments. Original pronunciation: raʾs || orthographic representation راس: internal alif = consonantal ʾ Loss of glottal stop: rās || orthographic representation راس: internal alif = ā Figure 7: The Development of the Digraph 54 The digraph اى appears to be an embryonic attempt to indicate the internal glottal stop in a writing tradition that kept this sound but used Ḥijāzī orthography, which lacked a graphic means of representing it. It is impossible to reconstruct, at the current moment, why it was employed marginally in the writing of the Quranic archetype and later documents. Despite the fact that most later reading traditions had the need to represent the word- internal glottal stop, it is clear that the main tradition of writing Arabic, following the reforms of ʿAbd al-Malik, did not employ the digraph. The occasions on which it was used in the Quranic archetype and other early manuscripts became orthographic relics, similar to the spelling of ماىــه today, the purpose of which was no longer understood. Perhaps the introduction of the method of marking the glottal stop with supralinear diacritics finally extinguished any lingering use of the digraph in the writing of Arabic. 6. The Archaic Final Hē The emergence of the digraph was based on the graphic reinterpretation of the archaic final hē in the word The same archaic form seems to have been frozen graphically in .ماىــه the spelling of the word Torah in Q 3:3 (and passim) as 55.تورىــه The spelling of this word has been the subject of much speculation, with suggested explanations ranging from spontaneous ʾimālah (the raising of ā to ē) to the contamination of the word tōrāh and 54. DAM 01-29.1 is an early Quran manuscript that, as G.-R. Puin observed, spelled the glottal stop on occasion with the ʾ, continuing the older Nabataean practice; see Puin, “Vowel Letters and Ortho-Epic Writing in the Qurʾān,” in New Perspectives on the Qurʾān: The Qurʾān in Its Historical Context 2, ed. G. S. Reynolds, 147–90 (New York: Routledge, 2011), 170. This is the same document that used the archaic final hē in the word al-ʾāḫirah, cited above, further indicating that the alif spelling is ancient rather than a later and coincidental development. 55. I thank Yüsef Gursey for suggesting the possible application of the allograph of final hē to this word in a Facebook post of March 30, 2018. 14 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) ʾorayṯā;56 both of these are wholly unconvincing, and the latter especially strains credulity. A simpler solution presents itself in light of the interpretation of the word ــه the spelling :ماى of tōrāh consists of four letters and not five, the final denticle and loop being simply the archaic form of the hē. This produces the expected spelling of /tōrāh/, twrh. While this spelling made its way into the Quranic archetype, another rather common spelling that did not lends itself to a similar interpretation. A number of times, the word for “god,” ʾilāh, is spelled in early manuscripts as 57.الىــه This has usually been interpreted as a plene spelling of ʾimālah, reflecting the pronunciation of *ā as ē.58 Although this explanation is in theory plausible,59 it is striking that such a spelling occurs with any frequency only in this word. This suggests to me that we most likely have another example of the archaic hē graphically frozen in the high-frequency word ʾilāh, mirroring the pre-Islamic examples identified above. In contrast to الىه, there are two examples provided by Puin of the denticle’s representing a long ā: in the word riǧāl, spelled رجىــل in Q 72:6 in DAM 01-28.1 and DAM 01-29.1,60 and in the word ʿibād, spelled عبىــد in Q 40:31 in DAM 01-29.1. The latter occurs in a rhyme position, where the rhyme is formed with the syllable āC#. Given that both of these spellings occur very infrequently, and never in the pre-Islamic period, there is no reason to assume that they reflect an ancient practice. They could have emerged in the scribal milieu of Quranic copying, perhaps being the innovation of a small group. In any case, I do not believe they are related to the denticled hē. Their interpretation lies in the phenomenon of orthographic reanalysis in the wake of language contact. The original language of the QCT had a distinct reflex of the triphthong *ayV: it was pronounced as ē and rhymed separately from the alif mamdūdah.61 However, by the time Qurans were being copied, the text was read in dialects that realized the reflex of *aya as ā. The orthographic mismatch allowed for the emergence of a new orthographic convention. But let us first illustrate the scenario: QCT original: banē-hā = ىىىها some late seventh/eighth century readings: banā-hā = ىىىها 56. For a complete discussion of the opinions on this subject, see Diem, “Untersuchungen,” 248–50. 57. Puin, “Vowel Letters,” 168. 58. Ibid.; but F. Deroche, in La transmission écrite du Coran dans les débuts de l’islam: Le codex Parisino- Petropolitanus (Leiden: Brill, 2008), views it simply as a mater lectionis for ā. 59.  Puin also cites the spelling of Q 4:3 طــاب in the Cairo Edition as طيــب in the Samarkand codex, but this may be explained by a different reflex—the collapse of the medial triphthongs; see van Putten, “Triphthongs,” 49–50, 69. It is reported that in the ancient Ḥijāzī dialect, this very verb was realized as ṭēba. Nevertheless, the scenario developed for the spelling of ā with the denticle could account for the pronunciation ṭāba even here. 60. The latter attestation was recently identified by van Putten. 61. On the outcome of the triphthongs in the QCT, see van Putten, “Triphthongs.” ىه ر و ت h r w t Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 15 Just as one could extrapolate from the spelling ماىــه that the alif-denticle combination represented ʾ, it is possible to deduce from such QCT spellings that word-internal ā could be spelled with a denticle—but only in a dialect that had no ʾimālah; in other words, a dialect that did not realize the reflex of the triphthong as ē. If this innovation indeed emerged in such a scribal context, it never truly took off, and the practice was quickly marginalized by the internal alif spelling. 7. Concluding Remarks I hope to have shown that an archaic strand of Quranic orthography made use of the digraph اى to represent the glottal stop in a noninitial position. This peculiar method of spelling emerged in a situation of script contact, where the graphically frozen spelling of the word “one hundred,” ماىــه, with the archaic denticled hē, ىــه, was reinterpreted as consisting of four letters, ه ى ا ى with ,م ,representing the glottal stop. From this point ا the use of the digraph spread to the representation of this sound in similar environments. Moreover, relics of the archaic final hē explain the enigmatic spelling of the word tōrāh and the supposedly ʾimālah’ed spelling of the word ʾilāh in several early manuscripts. This study brings into relief an important issue in the study of the language of the Quran: its layered orthography. Although the text is generally consistent in its spellings, the significant variation on its margins harkens back to a period before Arabic orthography was standardized.62 For this reason, the desire to devise rules to account for every spelling in the Quran is perhaps misguided; only a historically informed approach, with due consideration of the diversity of Arabic scripts and spelling strategies in the pre-Islamic period, can fully explain its many orthographic enigmas. 62. For a discussion on the possible timing of the standardization of Arabic orthography as we know it, see C. Robin, “La réforme de l’écriture arabe à l’époque du califat médinois,” Mélanges de l’Université Saint-Joseph 59 (2006): 319–64. 16 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) Index of Early Arabic Inscriptions Cited Ḥarrān inscription: Monumental bilingual Arabic-Greek inscription from Ḥarrān, Syria, dated to 568 CE; for the latest edition, see Macdonald’s commentary in Z. T. Fiema et al., “Provincia Arabia: Nabataea, the Emergence of Arabic as a Written Language, and Graeco- Arabica,” in Arabs and Empires before Islam, ed. G. Fisher, 373–433 (Oxford: Oxford University Press, 2015). Jebel Usays inscription: A small rock graffito at Jebel Usays in southern Syria, dated to 528 CE; for the latest edition, see Macdonald’s commentary in Fiema et al., “Provincia Arabia.” Ḥimà (Najrān) pre-Islamic Arabic inscriptions: A collection of pre-Islamic Arabic-script inscriptions from the region of Najrān, the earliest dating to 470 CE. On these texts, see C. J. Robin, A. I. al-Ghabbān, and S. F. al-Saʿīd, “Inscriptions antiques de la région de Najrān (Arabie séoudite méridionale): Nouveaux jalons pour l’histoire de l’écriture, de la langue et du calendrier arabe,” Comptes rendus des séances de l’Académie des inscriptions et belles- lettres 3 (2014): 1033–1128. Dūmah sixth-century Arabic inscription: A Christian Arabic-script graffito from the region of Dūmat al-Jandal, northern Saudi Arabia; see L. Nehmé, “New Dated Inscriptions (Nabataean and Pre-Islamic Arabic) from a Site near Al-Jawf, Ancient Dūmah, Saudi Arabia,” Arabian Epigraphic Notes 3 (2017): 121–64. Umm al-Jimal inscription: A pre-Islamic Arabic-script inscription, undated. For the latest discussions, see R. Hoyland, “Mount Nebo, Jabal Ramm, and the Status of Christian Palestinian Aramaic and Old Arabic in Late Roman Palestine and Arabia,” in The Development of Arabic as a Written Language, ed. M. C. A. Macdonald, 29–46 (Oxford: Archaeopress, 2010), 40, and Nehmé, “New Dated Inscriptions.” Yazīd inscription: An early Christian Arabic inscription from the region of Qaṣr Burquʿ in Northeastern Jordan; see Y. Shdaifat, A. Al-Jallad, Z. al-Salameen, and R. Harahsheh, “An Early Christian Arabic Graffito Mentioning ‘Yazīd the King,’” Arabian Archaeology and Epigraphy 28, no. 2 (2017): 1–10. Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 17 Quranic manuscripts sigla BnF Arabe 331, Paris (+ Ms. Leiden Or. 14.545b +Ms. Leiden Or. 14.545c), 58 folios; first Islamic century. Samarkand Kodex (Faksimiledruck): Berlin, Staatsbibliothek, 353 folios; dated to about 700–850. British Library, Or. 2165, 121 folios; second half of the first Islamic century. Großer Korankodex: Kairo, al-Maktaba al-Markaziyya li-l-Makhṭūṭāt al-Islāmiyya, 1087 folios; not before 700. Saray Medina 1a: Istanbul, Topkapı Sarayı Müzesi M1, 308 folios; late first/early second century. Tübingen, Universitätsbibliothek, Ma VI 165, 77 folios; second half of the seventh century CE. 18 • AhmAd Al-JAllAd Al-ʿUṣūr al-Wusṭā 29 (2021) Bibliography Al-Jallad, A. “‘Moge God Yazīd de Koning Indachtig Zijn’: Nadere Beschouwingen over de Yazīd-Inscriptie en de Ontwikkeling van de Arabische Schriften.” In Mohammad en de Late Oudheid, edited by Josephine van den Bent, Floris van den Eijnde, and Johan Weststeijn, 198–208. Amsterdam: Verloren, 2018. ——— and A. al-Manaser. “New Epigraphica from Jordan I: A Pre-Islamic Arabic Inscription in Greek Letters and a Greek Inscription from North-Eastern Jordan.” Arabian Epigraphic Notes 1 (2015): 51–70. Avner, U., L. Nehmé, and C. J. Robin. “A Rock Inscription Mentioning Thaʿlaba, an Arab King from Ghassān.” Arabian Archaeology and Epigraphy 24, no. 2 (2013): 237–59. Cantineau, J. Le Nabatéen. Paris, 1930/1932. Reprint. Osnabrück: Zeller, 1978. Déroche, F. La transmission écrite du Coran dans les débuts de l’islam: Le codex Parisino- Petropolitanus. Leiden: Brill, 2008. Diem, W. “Some Glimpses at the Rise and Early Development of Arabic Orthography.” Orientalia, n.s., 45 (1976): 251–61. ———. “Untersuchungen zur frühen Geschichte der arabischen Orthographie: I. Die Schreibung der Vokale.” Orientalia, n.s., 48 (1979): 207–57. Fiema, Z. T., A. Al-Jallad, M. C. A. Macdonald, and L. Nehmé. “Provincia Arabia: Nabataea, the Emergence of Arabic as a Written Language, and Graeco-Arabica.” In Arabs and Empires before Islam, edited by G. Fisher, 373–433. Oxford: Oxford University Press, 2015. Fox, J. Semitic Noun Patterns. Winona Lake, IN: Eisenbrauns, 2003. Healey, J. The Nabataean Tomb Inscriptions of Mada’in Saleh. Journal of Semitic Studies Supplement 1. Oxford: Oxford University Press, 1993. Hoyland, R. “Mount Nebo, Jabal Ramm, and the Status of Christian Palestinian Aramaic and Old Arabic in Late Roman Palestine and Arabia.” In The Development of Arabic as a Written Language, edited by M. C. A. Macdonald, 29–46. Oxford: Archaeopress, 2010. Huehnergard, J. “Features of Central Semitic.” In Biblical and Oriental Essays in Memory of William L. Moran, edited by A. Gianto, 155–203. Rome: Pontificio Istituto Biblico, 1995. Ingham, B. Najdi Arabic: Central Arabian. Amsterdam: J. Benjamins, 1994. Nehmé, L. “Aramaic or Arabic? The Nabataeo-Arabic Script and the Language of the Inscriptions Written in This Script.” In Arabic in Context: Celebrating 400 Years of Arabic at Leiden University, edited by A. Al-Jallad, 75–89. Leiden: Brill, 2017. ———. “Epigraphy on the Edges of the Roman Empire: A Study of the Nabataean Inscriptions and Related Material from the Darb Al-Bakrah, Saudi Arabia, 1st–5th Century AD.” Al-ʿUṣūr al-Wusṭā 29 (2021) The Digraph اى in the Quranic Consonantal Text • 19 Mémoire scientifique d’habilitation à diriger des recherches, École pratique des hautes études, 2013. ———. “A Glimpse of the Development of the Nabataean Script into Arabic Based on Old and New Epigraphic Material.” In The Development of Arabic as a Written Language, edited by M. C. A. Macdonald, 47–88. Oxford: Archaeopress, 2010. ———. “New Dated Inscriptions (Nabataean and Pre-Islamic Arabic) from a Site near Al-Jawf, Ancient Dūmah, Saudi Arabia.” Arabian Epigraphic Notes 3 (2017): 121–64. Puin, G.-R. 2011. “Vowel Letters and Ortho-Epic Writing in the Qurʾān.” In New Perspectives on the Qurʾān: The Qurʾān in Its Historical Context 2, edited by G. S. Reynolds. New York: Routledge, 2011. Robin, C. “La réforme de l’écriture arabe à l’époque du califat médinois.” Mélanges de l’Université Saint-Joseph 59 (2006): 319–64. ———, A. I. al-Ghabbān, and S. F. al-Saʿīd. “Inscriptions antiques de la région de Najrān (Arabie Séoudite méridionale): Nouveaux jalons pour l’histoire de l’écriture, de la langue et du calendrier arabe.” Comptes rendus des séances de l’Académie des inscriptions et belles- lettres 3 (2014): 1033–1128. Shdaifat, Y., A. Al-Jallad, Z. al-Salameen, and R. Harahsheh. “An Early Christian Arabic Graffito Mentioning ‘Yazīd the King.’” Arabian Archaeology and Epigraphy 28, no. 2 (2017): 1–10. Sidky, H. “In Search of Lost Time: A Vocalized Muṣḥaf, Ibn ʕāmir, and the Evolution of the Syrian Reading Tradition.” Forthcoming. van Putten, M. “The Development of the Triphthongs in Quranic and Classical Arabic.” Arabian Epigraphic Notes 3 (2017): 47–74. ———. “The Feminine Ending -at as a Diptote in the Qurʾānic Consonantal Text and Its Implications for Proto-Arabic and Proto-Semitic.” Arabica 64, nos. 5–6 (2017): 695–705. ———. “‘The Grace of God’ as Evidence for a Written Uthmanic Archetype: The Importance of Shared Orthographic Idiosyncrasies.” Bulletin of the School of Oriental and African Studies 82, no. 2 (2019): 271–88. ———. “Hamza in the Quranic Consonantal Text.” Orientalia 86, no. 3 (2018): 93–120. ———. Quranic Arabic: From its Hijazi Origins to its Classical Reading Traditions (Leiden: Brill, forthcoming).