The Marking of Poetry: A Rare Vocalization System from an Early Qurʾān Manuscript in Chicago, Paris, and Doha

* I am indebted to the registration staff at the Oriental Institute of the University of Chicago for permitting me to photograph and reproduce the OIM fragments discussed below. I am equally indebted to the multimedia team at the Museum of Islamic Art in Doha for providing images of the MIA fragments, and for their permission to reproduce them here. These institutions retain copyright ownership of their respective images. I would also like to thank François Déroche for sharing his expertise on this particular Qurʾān manuscript, and Magdalen Connolly for her keen observations and advice on an earlier draft of this paper. This work was supported by the Bill & Melinda Gates Foundation [OPP1144]. [Ed.: Note also Éléonore Cellard’s article on a related topic in this issue of JNES entitled “The Ṣanʿāʾ palimpsest: Materializing the Codices.”]


Introduction
In 1939, Nabia Abbott published her groundbreaking book on Arabic paleography, The Rise of the North Arabic Script. In its preface, she admitted that it was not her intention to produce such an extensive work-at least not at that time-and that her original goal was simply to catalog the Qurʾan manuscripts at the University of Chicago's Oriental Institute. She explained: Since these manuscripts cover a wide period of time and present a variety of scripts, it soon became apparent that this undertaking could not be satisfactorily accomplished without the aid of special scientific equipment: a knowledge of both the historical development of the North Arabic script and the progress of Ḳ urʾanic writing, especially in the early centuries of Islam. Investigation, however, soon revealed the fact that such knowledge is not available in any complete and up-to-date form. 1 The field of Arabic paleography has advanced considerably in the last eighty years, 2 but it remains indebted to Abbott's work in creating some of that first "special scientific equipment." However, there is one type of equipment that Abbott did not have in 1939: digital photography. As a result, the quality of the manuscript images at the end of Rise is frustratingly poor by modern standards, and many of their details have gone unnoticed.
This paper provides updated digital images of four fragments from the Oriental Institute Museum (OIM) that appeared in Abbott's Rise of the North Arabic Script, and calls attention to features of their paleography and vocalization which are not apparent from her original black-and-white plates. 3 In doing so, it demonstrates that these four fragments all belong to the same * I am indebted to the registration staff at the Oriental Institute of the University of Chicago for permitting me to photograph and reproduce the OIM fragments discussed below. I am equally indebted to the multimedia team at the Museum of Islamic Art in Doha for providing images of the MIA fragments, and for their permission to reproduce them here. These institutions retain copyright ownership of their respective images. I would also like to thank François Deŕoche for sharing his expertise on this particular Qurʾan manuscript, and Magdalen Connolly for her keen observations and advice on an earlier draft of this paper. copy of the Qurʾan, which is also the same Qurʾan as nineteen leaves in Paris' Bibliotheque Nationale de France (BNF) and two leaves in Doha's Museum of Islamic Art (MIA). All twenty-five folios of this Qurʾan contain a unique vocalization system that represents /a/ and /u/ with miniature red forms of alif and wāw. This system of "letter-form" 4 vocalization signs is unattested in other Qurʾan manuscripts and corresponds with a type of vocalization which medieval sources describe as shakl al-shiʿr: "the marking of poetry." The subsequent discussion analyzes the features of this shakl al-shiʿr system and contextualizes it within the history of both this Qurʾan manuscript and the history of Arabic writing as a whole.

The OIM-BNF-MIA Qurʾan
The commonalities in size, script, and vocalization of the OIM, 5 BNF, 6 and MIA 7 folios suggest that they all belonged to the same codex. They contain the following sūras, divided across ten groups of leaves (see Table 1). Although there are substantial gaps, the OIM fragments bookend the BNF and MIA material, including one case (section I) in which an MIA folio immediately follows a BNF folio. Each folio was once part of a bifolium, and the MIA folios have stubs. The following sections describe the common features of all the folios.

The OIM Folios
The four fragments which concern us here are Abbott's Nos. 10-13, known as OIM A6963, A6962, A6961, and A6993, respectively. They are from parchment folios, but their edges are badly torn, which makes it difficult to determine their original size. Based on the script style, Abbott dated them to the late eighth or ninth century, identifying the former pair (OIM A6963 and OIM A6962; Nos. 10 and 11) as successive folios from one copy of the Qurʾan, and the latter pair (OIM A6961 and OIM A6993; Nos. 12 and 13) as successive folios from another. 8 She measured OIM A6963 at 25.8 × 23 cm, but estimated that the leaf was originally about 35 × 26 cm with eighteen lines of text. The next leaf, OIM A6962, has at least part of all eighteen lines. OIM A6961 and OIM A6993 are much more badly torn than the former pair, but remnants of at least seventeen lines are visible on OIM A6961. Abbott pointed out that the length of the lines is the same across all four fragments, and for the latter pair, she went so far as to say that "comparison of the manuscript with the printed text would allow here too eighteen lines to the page, which like-wise must have measured originally about 35 × 26 cm." 9 Yet despite acknowledging this potential for the four leaves to have had the same original size and writing area, she was reluctant to say that they were from the same codex, concluding: The script of [No. 12] and of No. 13,belonging to the same copy of the Ḳ urʾan, differs from that of Nos. 10-11 only in that it is a little larger and provides more space between the letters. Except for this and a little difference in the verse division marks, these four numbers might easily belong to the same copy of the Ḳ urʾan. 10 Both pairs of fragments fit neatly into an A.I script style as defined by François Deŕoche. 11 Their differences are slight, and can be explained by two different scribes working together on the same manuscript. 12 With only these small samples, Abbott likely did not have enough information to make a more precise statement on the relationship between these fragments, but comparison with the BNF and MIA leaves shows that all four do belong to the same codex. Moreover, they were indeed written by two different hands.

The BNF Folios
The nineteen parchment folios of BNF Arabe 330f 13 make up seven discontinuous sections of the Qurʾan. 4 Revell, "Diacritical Dots": 180-81. 5  Ibid., 66-67. 10 Ibid., 66. 11 Deŕoche places this style in the second half of the eighth century, essentially in agreement with Abbott's chronology, but he also acknowledges the difficulty of dating manuscripts with A-type script styles, saying, "very few examples survive . . . no external evidence has been discovered for dating them": Deŕoche, Abbasid Tradition (1992), 35. 12 For further discussion of scribal collaboration, see Deŕoche et al., Islamic Codicology (2015), 198-99. 13 BNF Arabe 330 is a modern bound book compiled from seven different manuscripts. Of interest here is the sixth manuscript, BNF Arabe 330f, comprising folios 31 to 49 from that book. Images of BNF Arabe 330 are publicly accessible from the BNF online Gallica archive: https://archivesetmanuscrits.bnf.fr/ark:/12148/cc386217 (accessed 4 October 2020). The images of BNF Arabe 330f used below are reproduced in accordance with the Bibliotheque Nationale de France's non-commercial fair use policy.
Like the OIM fragments, it has eighteen lines, and its script is the A.I type. 14 In Les manuscrits du Coran, Deŕoche mentions that these leaves contain letter-form vowel signs similar to those in the OIM material. He suggests to compare BNF Arabe 330f to Abbott's Nos. 10 and 11, but he does not argue that they are from the same manuscript, nor does he mention Abbott's Nos. 12 and 13 at all. 15 It is difficult to determine a firmer connection between these folios based on the grainy plates in Rise of the North Arabic Script alone. In any case, the BNF leaves are generally better preserved than Abbott's fragments, so Deŕoche accurately determines their original size. He measures the intact leaves at 37 × 28 cm, 16 not far off from Abbott's estimate of "about 35 × 26 cm."

The MIA Folios 17
In 2013, the Museum of Islamic Art in Doha catalogued two parchment Qurʾan folios now known as MIA.2013.27.1 and MIA.2013.27.2. These leaves have been partially digitized as part of the Qurʾan Gateway project, 18 which has made it apparent that they also belong with the OIM and BNF material. Both leaves are classified as B.II script style in the Qurʾan Gateway database, 19 but this designation is a mistake. Like the OIM and BNF leaves, they are A.I type, although MIA.2013.27.1 also has some h ̣ijāzı̄features, as will be shown below.

Scribal Hands
Three main hands worked on the extant portions of this manuscript (see again Table 1). The first (Hand 1) wrote sections A through D, as well as the first page of section E (BNF Arabe 330f F39r). The second (Hand 2) wrote the rest of sections E, F, and H through J. The third hand (Hand 3) wrote only a single extant folio (MIA.2013.27.1), which makes up all of section G.
In general, the first hand is consistent in the forms of its letters, while the second shows more frequent variations. Three features differentiate these two hands at a glance. First, when writing consecutive lāms, the first hand uses nearly parallel strokes that are usually close together (Figs. 1a-c):    Second, the first hand angles the arm of initial ʿayn slightly upwards (Figs. 3a-d) The second hand sometimes manages to match the initial ʿayn of the first, but usually folds the arm further down, closer to the baseline (Figs. 4a-d):  The third hand appears in only one folio, but the two largest lacunae are on either side of it, so it is possible that this hand wrote additional folios that are no longer extant. While similar to the A.I style of the first two hands, it also shows influences of an earlier h ̣ijāzıs tyle, with taller ascenders that have a stronger tendency to lean rightwards. The main difference is in alif, which vacillates between the A.I style with a nearly vertical shaft and medium-sized lower return, and a h ̣ijazı̄style with an oblique shaft that extends rightward past a shortened return ( Fig. 7): 20 Unfortunately, the small amount of material here is limiting. Without evidence from more folios, we cannot rule out the possibility that this folio is not a third hand, and instead is a particularly inconsistent section written by the second hand.
Besides the main scribal hands, at least one later hand restored faded sections of the rasm with black carbon ink. Most words have undergone some amount of this repair work. In some cases, the restorer only retraced parts of letters, and it is often possible to see the contrast where the original strokes end and those of the restoration begin. 21 In other cases, the restorer overwrote extensive passages, including some entire pages. 22 The carbon ink had a less permanent quality than the original ink, which presumably is a metalo-gallic compound. 23 As such, while the original ink did fade considerably, the later ink is often completely rubbed off, leaving only the outline of words behind ( Fig. 8): Further evidence of later modification comes in the latter sections of the manuscript (H, I, and J), where a later hand has inserted alif to amend defective spellings of medial /a/. The first instance of this phenomenon is in Qurʾan 33:35, where someone inserted alifs into seven words with medial /a/. 24 I suspect that the reader was frustrated by the numerous defective spellings in this verse, and became convinced that it was acceptable to amend the rasm from that point onwards. These inserted alifs do not match the script style of the rest of the text, but some are similar to the miniature alif letter-form vocalization signs found throughout the manuscript. The same insertion occurs in Qurʾan 33:49 and 33:50, with the words sarah ̣ an , 25 azwajaka, and azwajihimu, then again with al-samāwāt in Qurʾan 34:22 and 34:24. 26 It also appears in andād an in Qurʾan 34:33. On the same leaf, for a defective spelling of qāla in Qurʾan 34:32, someone erased the sublinear hook of the lām so that it resembles an alif, then inserted an additional lām to create the appearance of a plene spelling. 27  The variations in the styles of the inserted alifs suggest that this later hand was not trained in calligraphy. Furthermore, while the plene spelling of medial /a/ is common in medieval personal-use Qurʾans, it is less common in model codices. 29 As such, these alifs suggest that an unprofessional hand-probably an owner of the codex-modified it after professional scribes produced it. This owner may have been the same person who restored the rasm with carbon ink. 21 This phenomenon is clearest on BNF Arabe 330f F38r. 22 For example, see BNF Arabe 330f F31r, F36v, F39v, F40r, F41v, F43v, F46r, F47r, F49r;and MIA.2013.27.2v. 23

Ornamentation and Division
All twenty-five leaves regularly separate verses with three to five oblique strokes. There are also ornamental dividers for groups of ten verses throughout the text, some of which are superimposed on the oblique dividing strokes. 30 There are four types of dividers that vary across the ten groups of folios (except in section D, which has none). 31 First, dark brown or black concentric rings, connected by small strokes, appear in sections B, C, F, and G (Figs. 10a-g):  Finally, section A contains only one divider. It is more intricate than the others, probably because it marks the 280 th verse of al-Baqara and is the last ten-verse division in that chapter. It is another green rosette, this time with many more "petals," and it includes a central red dot with a circular red outline ( Fig. 13): The different styles of ten-verse dividers do not correlate to the sections of the main scribal hands, suggesting that that the original scribes themselves did not add them. The regular appearance of verse dividers superimposed on the earlier layer of verse-dividing slashes reinforces this conclusion. Instead, someone else ornamented this Qurʾan after the rasm was complete, and the variation in the styles of the dividers suggests that this ornamentor did not have strict guidelines for their work. At the same time, the groups of dividers crosscut the OIM, BNF, and MIA leaves, indicating that they were ornamented together, and providing further evidence that they all belong to the same Qurʾan.
Just two of the twenty-five leaves contain the beginning of a sūra: OIM A6962v (section A) and BNF Arabe 330f F38r (section D). Both start with a header, written in red ink by a later hand, giving the title and number of verses of that sūra. The original scribes did not intentionally leave space for these headers. This type of heading is a common feature in early Qurʾans, 32 and it is an additional similarity between the OIM and BNF folios (Figs. 14-15):  Below its header, OIM A6962 also has an ornamental band (Fig. 16), which Abbott described in this way: The motif here is a simple one, consisting of green X's with four red dots around their centers, alternating with green lozenges containing red and green dots and accompanied by four red semicircles apiece, one projecting from each side. Green scrolls connect these successive elements. This band's colors are consistent with the ornamentation of the other folios. In contrast to the ruled lines of the main text, the upper and lower edges of this band are not straight, and the ink has bled through the parchment. These details indicate that the original scribes did not add the band, and it is likely the work of someone who was not professionally trained in ornamenting codices.

Diacritics
All twenty-five leaves contain thin slashes as diacritic marks on consonants. There are also dots and thicker strokes, added later in darker black ink along with the restoration of the rasm. These heavier marks sometimes overlap the earlier slashes and even the red vowel dots. See, for example, Figure 17, from the BNF material:  These inconsistencies show that diacritic marks were added to the manuscript more than once, including by people who had different ways of distinguishing fāʾ from qāf. It is highly likely that multiple people owned this Qurʾan, and each one added new diacritics marks to suit their needs as orthographic standards changed over time. This alteration follows the general trajectory of diacritic conventions for qāf and fāʾ, as eighth-century writers usually used a supralinear dot for qāf and a sublinear dot for fāʾ, while the use of two dots for qāf and a dot above for fāʾ appears from the ninth century onwards. 36 While it is not clear exactly how much later the heavier marks were added, this manuscript straddles the transition period between these two diacritic conventions.
There are no diacritic signs for sukūn, was ̣l, tashdıd, or takhfıf, but a red dot may indicate hamza. 37 A red semicircle, placed either above (with /a/ and /u/) or below (with /i/) a letter, also represents hamza. 38 Based on a comparison of ink shades, these semicircles were added at the same time as the letter-form vocalization signs. They sometimes reinforce red dots that were already present :

Recitation Tradition
Abbott pointed out that OIM A6963 and OIM A6962 (section A) extend the vocalization of the pronominal suffixes -kum and -hum to -kumūand -humū. 39 This phenomenon occurs regularly in OIM A6993 (section I) and throughout the BNF and MIA folios (Figs. 23-25).   Additionally, all of the leaves consistently indicate a lack of vowel harmony on words that end with the 35 Deŕoche et al. (Islamic Codicology [2015], 220-24; esp. 221 n. 67) cite both BNF Arabe 330f and the OIM fragments as examples of this old diacritical practice for distinguishing faʾ and qaf, but they do not mention them in their discussion of the vowel points. 36 Gruendler, "Arabic Script," (2001), 140. Note that maghrebıw riters retained the older convention even after the double dot for qaf became standard elsewhere. 37 Abbott, Rise of the North Arabic Script (1939), 40, 67. See George, "Coloured Dots" (2015): 14-15. 38 Abbott suggested that this semicircle also represents shadda once, on yud ̣arra ("he is harmed") in masculine singular pronominal suffix, 40 for example: kutubihūwa-rusulihūinstead of kutubihı̄wa-rusulihı( Qurʾan 2:285; OIM A6962r, line 9), fıhūikhtilāf an instead of fıhi-khtilāf an (Qurʾan 4:82; BNF Arabe 330f F31r, lines 16-17), and ʿalayhu instead of ʿalayhi (Qurʾan 18:81, MIA.2013.27.1r, line 6). 41 The manuscript contains a few other variations, 42 including an apparently regular shift of the /ay/ diphthong to /ı/, 43 or at least the regular appearance of a red dot below yāʾ in positions where an /ay/ diphthong would be expected. This marking may be a representation of imāla affecting the fath ̣a of these diphthongs, approximating a pronunciation closer to /e/. This feature appears to be systematic, but I have not examined every instance, and it is beyond the scope of this paper to analyze it more fully within the traditions of qiraʾāt. Suffice it to say that the consistent appearance of these variants across the OIM, BNF, and MIA material is further evidence that they all belong to the same copy of the Qurʾan.

Vocalization
All of the leaves are frequently, if not fully, vocalized, using a combination of red dots and miniature red letterform signs in the shape of alif and wāw. The red dots follow the standard arrangement, with a supralinear dot for /a/, a sublinear dot for /i/, and an intralinear dot for /u/. This last dot is sometimes superimposed on the rasm, especially for internal vowels. Two dots indicate tanwın, usually in the same position as their nonnunated counterparts. In general, the dots are a more uniform shade of red than the letter-form signs, and they have not faded as much. By contrast, the letterform signs have multiple lighter red or red-orange hues. These differences indicate the use of various thinner red ink solutions in the application of the letter-form signs As for the signs themselves, a miniature alif, usually placed above a letter, represents /a/. Similarly, miniature wāw represents /u/. This latter sign can appear above a letter, but it often occurs intralinearly on the left when there is enough space to avoid overlapping the rasm. It may also appear superimposed on a red dot, such that the dot fills the miniature wāw's open counter. Conspicuously, there is no evidence of a miniature yāʾ sign to mark /i/. The letter-form signs never contradict the red dots, and most appear to "reinforce" 45 dots that were already present. Only a few occur without any accompanying dots, and they never occur with tanwın, but otherwise they can indicate internal vowels, final vowels, and i'rābı̄case vowels.
It is difficult to determine precise numbers for the occurrences of each vowel dot and letter-form sign, as many have faded or rubbed off, but it is possible to make some estimations. The following ratios are broadly consistent across the entire manuscript. Of the vowels that are marked, 46 80-90% have a red dot alone. Miniature letter-form signs occur with 10-20% of marked vowels. 85-90% of miniature letter-form signs occur reinforcing a red dot. Only 1-3% of marked vowels have a letterform sign alone. The letter-form sign for /u/ appears roughly twice as often as the sign for /a/. It does not seem that there was a systematic motivation for the vocalizer who added these signs, which occur in almost every orthographic and grammatical context, although they are practically absent on long vowels. They are also clearly not the same as another medieval orthographic practice that involved the insertion of full red alifs to amend defective spellings of medial /a/. 47 Indeed, these letter-form signs do not change the rasm at all. Instead, they represent vowels in exactly the same way as the red dots, and they do not inherently indicate vowel length. The only discernible clue to this vocalizer's motivations is the occurrence of the miniature wāw sign 40 On this phenomenon, see Éleónore Cellard, "La vocalisation des manuscrits coraniques" (2015); Marijn Van Putten, "Arabe 334a" (2019). 41 This phenomenon also occurs with the dual pronominal suffix, which appears as f ıhumāin BNF Arabe 330f F49r, line 13. 42 Other differences include a representation of imala in jaʾa (he came/brought) with a red dot below (as if jiʾa) (BNF Arabe 330f F46v, lines 3 and 18); and a change of nazzala and anzala to nuzila and unzila in Qurʾan 4:136 (BNF Arabe 330f F35v, lines 5-6). 43  approximately twice as often as the alif sign. This tendency runs counter to the red dots in the text, which mark /a/ much more often than /u/ or /i/, largely due to the more frequent occurrence of /a/ in Arabic phonology. Most likely, the letter-form vocalizer recognized /a/ as a sort of "default" vowel and needed more orthoepic reminders for the comparatively infrequent /u/.
The examples presented in Figures 27 through 43 (following page) represent different contexts in which the signs appear. Here we see short medial vowels , long vowels (Fig. 31 48 ), and final vowels (both iʿrābı,. There is also significant variation in the shapes of the miniature alifs and wāws. The alif signs fluctuate in their height, curve, and the length of their right-hand hooks . Meanwhile, the wāw signs reflect different degrees of angular tails and open counters. The openness of the counter is often determined by whether or not the sign is superimposed on a red dot (Figs. 41-43).
The uniform ink shade of the red dots and their general tendency not to fade suggests that they were added to the text as part of a single scribal operation. It is impossible to say if the same scribe(s) who wrote the consonantal text also added the red dots, but whoever added them did so systematically and deliberately with a single solution of red ink. By contrast, the variation in the red shades of the letter-form signs and their higher tendency to fade suggests that a later user added them at several different times, using whatever red ink solutions they had on hand. Most likely, someone acquired this Qurʾan after specialists had already written the rasm and applied the initial red dots, and then that person added the letter-form signs in the course of personal use. This use must have involved multiple reading sessions-perhaps split across months, or even yearswhich further explains why there is so much variation in inks and shapes of the letter-form signs. This situation also explains why the signs are not used systematically for any particular vowel context: the user was not a trained scribe, and simply added a sign whenever they wanted a reminder. For this user, in comparison to the red dots, the letter-form signs must have been more intuitively linked to the phonemes that they indicated, and thus facilitated greater ease of reading.

The Marking of Poetry
The red-dot vocalization system was the standard for Qurʾan manuscripts from the early eighth until the late tenth or eleventh century. 49 However, it proved overly cumbersome for non-Qurʾanic writing, so supposedly the grammarian al-Khalıl ibn Ah ̣mad (d. 786/791) invented a new system of vocalization signs using miniature versions of alif, wāw, and yāʾ. According to Abbott, al-Khalıl's system was more convenient than the red dots, so it spread quickly across non-Qurʾanic genres. 50 Abbott also suggested that al-Khalıl's signs evolved into the modern Arabic vowel signs, 51 and argued that the four OIM fragments support this interpretation, writing: "The appearance of alif and w [sic] thus used confirms the theory that the modern fath ̣ah, d ̣ammah, and (by inference) kasrah have their origins in the letters alif, w, and y respectively." 52 She based this conclusion on the work of the eleventh-century tajwıd scholar AbūʿAmr al-Danı̄(d. 1053), 53 who recorded a report about al-Khalıl in his al-Muh ̣kam f ı̄Naqt ̣ al-Masāh ̣if (The Rules for Pointing the Codices): Abūal-Ḥ asan ibn Kaysan said: Muh ̣ammad ibn Yazıd said: The marking which is in books is from the work of al-Khalıl, and it is taken from the forms of the letters, so the d ̣amma is a small-form wāw, above the letter in order to not be confused with a written wāw. Then the kasra is yāʾ below the letter, and fath ̣a is a slanted alif above the letter. 54 48 This is the only instance I could find in which a miniature letter-form sign clearly marks a long vowel. 49 And even later, in some places. See Deŕoche, "Manuscripts of the Qurʾan" (2003); Deŕoche et al., Islamic Codicology (2015), 222-23. 50 Abbott, Studies in Arabic Literary Papyri, vol. III (1972), 7-9. See also, Revell, "Diacritical Dots" (1975)                 This description specifies that the system of letter-form vocalization signs is for books (kutub), in contrast to Qurʾans, which al-Danı̄usually calls codices (mas ̣āh ̣if ). The designation of yāʾ as the shape of kasra then seems to allude to an earlier stage of the Arabic vocalization system, prior to its final form, which now represents kasra with an oblique stroke. Slightly later, al-Danı̄refers to this letter-form system as shakl al-shiʿr, "the marking of poetry." He first relays a quotation from Abūal-Ḥ usayn ibn al-Munadı, the author of another book on pointing: He said: If you want to make the pointing rounded, there is no problem with that. If you want to make some of it rounded, and some of it with shakl al-shiʿr, then there is no harm in that, provided that you give the letters which are different their correct requirements. He said: Some scribes do not change the original rasm of the codex, but if they come upon a letter for which they know the naqt ̣ or the shakl is not correct, then they put whatever they prefer from the various readings, noting the different colors. 55 All of this occurs in the codices. 56 But then al-Danı̄adds further commentary, saying: [It is better] to refrain from the use of shakl al-shiʿr-which is the marking that is in books, that al-Khalıl invented-in mosque codices from the first, original, most correct versions, and others besides them; imitating, among the successors, those who began the pointing, and continuing in agreement with the predecessors. 57 Then later, after explaining how the red dots represent each vowel, he also writes: We only make the full vowels with rounded points according to a single form, an agreed shape-and we do not make the fath ̣a a reclined alif, nor the kasra a recurved yāʾ, nor the d ̣amma a small wāw-according to the practice of the previous people of Arabic. So, conceal the adoption [of signs] from these three letters, as evidence of that: following, among ourselves, the practice of those among the scholars of the past who began the pointing, in the presence of the companions. 58 Here al-Danı̄provides a clearer picture of what he believes were the original signs of shakl al-shiʿr: an inclined alif, a small wāw, and a yāʾ mardūda. In contrast to al-Munadı, he also discourages the use of these signs in publicly-visible copies of the Qurʾan. He emphasizes that abstention from shakl al-shiʿr demonstrates continuity with the "people of Arabic" and "scholars of the past" who first utilized vowel pointing. These scholars apparently lived among the companions of Muh ̣ammad, and al-Danı̄may be alluding to Abūal-Aswad al-Duʾalı̄(d. 689), an early grammarian whom both he and other medieval writers credit with the invention of the red-dot vocalization system. 59 The system of miniature alif and wāw signs in the OIM-BNF-MIA Qurʾan appears to be the shakl al-shiʿr that Ibn Yazıd, al-Munadı, and al-Danı̄describe. However, while the system in this Qurʾan is indeed based on the forms of letters, there is no yāʾ-shaped sign for kasra, and whenever space allows, the miniature wāw is placed on the left, rather than above. These details deviate from the medieval descriptions, but Abbott offers a potential explanation: these signs represent a transition period in the history of Arabic vocalization, at a time when some people were experimenting with new systems to replace the red dots in their Qurʾan codices. 60 Such a transition period would have been during roughly the ninth century, 61 prior to the introduction of the modern vowel signs to Qurʾans. There was likely some variation in the first applications of shakl al-shiʿr, and one variant may have more closely matched Abbott's inference and al-Danı's description of al-Khalıl's three vowel signs. 62 Either way, al-Danı's report probably reflects a homogenized vision of the shakl al-shiʿr, based on his eleventh-century understanding of this ninth-century transition period. By contrast, the vocalizer of the OIM-BNF-MIA Qurʾan used a specific stage or variant of the system, which apparently had no sign for kasra.
It must be reiterated that this codex is the only known Qurʾan manuscript in which shakl al-shiʿr vocalization appears, and this apparent dearth of sources complicates any attempt to extrapolate broader conclusions. 63 We therefore must ask the question: did any system of letterform vocalization actually exist outside of this codex, or is the story of al-Khalıl and the shakl al-shiʿr just a medieval etiological explanation for the modern vocalization signs? It is impossible to know from this manuscript alone. Nevertheless, even if this Qurʾan is wholly anomalous, then it is an anomaly that arose in the context of a transition period for Qurʾanic vocalization, just before the modern Arabic vowel signs began replacing the red dots.

Conclusion
As far as I am aware, OIM A6963, OIM A6962, OIM A6961, OIM A6993, BNF Arabe 330f, MIA.2013.27.1, andMIA.2013.27.2 are the only Qurʾan folios vocalized with the shakl al-shiʿr system of signs (Figs. 44-55, following two pages), and it is now clear that they all belong to the same codex. The paleography of this codex suggests that at least two professional scribes produced its consonantal text in the late eighth or ninth century, and they likely added red-dot vocalization at the same time. Then at least one person (though likely several people) altered the codex throughout its lifetime, modifying it to facilitate their personal use of the text. These modifications included: restoring damaged parts of the rasm, inserting alifs for medial /a/, decorating with ten-verse dividers and ornamental bands, adding red sūra headings, updating the diacritic system to align with newer standards, recoloring some of the red dots, and applying shakl al-shiʿr for a-and u-vowels. Most likely, these people were owners of the codex, but we have an incomplete picture of who could have accessed this manuscript, where they would have done so, and how that situation could have changed over time. Due to the disparate nature of the extant folios, it may ultimately be impossible to recover such socio-historical context. 64 The letter-form vowel signs in this Qurʾan are identifiable with shakl al-shiʿr, a medieval vocalization system which al-Danı̄attributes to al-Khalıl. The addition of shakl al-shiʿr to this Qurʾan may have technically contradicted a tradition of exclusively using the red dots for "Qurʾanic" vocalization, but that practice did not stop the later vocalizer. In fact, they were likely among the first people to vocalize a Qurʾan codex with new vowel signs, participating in the early stages of the transition from red dots to modern vocalization in Qurʾan manuscripts.
The lack of a distinct shakl al-shiʿr sign for /i/ in the OIM-BNF-MIA Qurʾan throws doubt on the idea that the modern kasra evolved from a letter-form yaʾ sign. Abbott and al-Danı̄both assumed that the original system had a miniature sublinear yaʾ, but such a sign is absent in the only clear extant example of shakl al-shiʿr. Perhaps the user who added this shakl al-shiʿr finished marking a-and u-vowels, but then ran out of time before they could reinforce the i-vowels; or perhaps they 61 In his Kitab al-Mas ̣ah ̣if, Ibn Abı̄Dawud (d. 929) transmits a report on vocalization from AbūḤ atim al-Sijistanı̄(d. 869), who describes the red-dot vocalization system but makes no mention of shakl al-shiʿr. If this report is authentic to the ninth century, then the shakl al-shiʿr may not yet have been invented during al-Sijistanı's lifetime, or he may not have been aware of it. Alternatively, he may have considered it a "non-Qurʾanic" system, and thus did not include it in a report about vocalizing the Qurʾan. See Abı̄Bakr Ibn AbıD awud, Kitab al-Mas ̣ah ̣if (2002) 1960), 4a, 11a, 20b. 63 Not only are there no other known Qurʾan manuscripts with the shakl al-shiʿr, it is not even clear that the system was ever used in non-Qurʾanic manuscripts. Abbott gives examples of several early ninth-century literary texts with what she calls "letter signs," "smallletter vowels," or "vowel symbols" (Abbott, Studies in Arabic Literary Papyri III [1972], 9, 11), but as far as I can tell, these manuscripts have the modern Arabic vowel signs. See also Abbott, Studies in Arabic Literary Papyri, vol. I (1957), document 1, and Abbott, Studies in Arabic Literary Papyri, vol. II (1967), documents 2, 6, 12, 13. George notes that "these signs are solely attested in secular documents for the third/ninth century onwards," but he refers to several manuscripts with the modern signs; George, "Coloured Dots (Part I)" (2015), 13-14 and n. 79. Geoffrey Khan describes sporadic modern signs in the Arabic papyri documents of the Khalili Collection, which likewise lack shakl al-shiʿr vocalization (Arabic Papyri           never learned the full version of the system in the first place. If so, then this two-sign shakl al-shiʿr might be idiosyncratic to them. In any case, the lack of evidence for shakl al-shiʿr with three unique signs does not necessarily mean that it did not exist-medieval scholars certainly believed that it did-but it does mean we cannot confidently say that the modern kasra sign evolved from the letter yāʾ. Instead, it may be that while modern fath ̣a is indeed an oblique stroke evolved from a slanted shakl al-shiʿr alif, and while modern d ̣amma is the shakl al-shiʿr waw fixed in a supralinear position, modern kasra is only an imitation of the fath ̣a sign, placed below a letter on analogy with the sublinear position of the red dot for /i/. Finally, the shakl al-shiʿr was not the only medieval Middle Eastern vocalization system that used miniature letters. Babylonian (i.e., Iraqi) Hebrew Bible manuscripts are attested from the tenth century with a system of vowel signs based on miniature Hebrew letters. This vowel system probably emerged in the latter half of the ninth century, first using miniature forms of the letters aleph ‫,)א(‬ vav ‫,)ו(‬ yod ‫,)י(‬ and ʿayin ‫)ע(‬ in supralinear positions. 65 Then, during the tenth century, these signs evolved into more economical shapes that could be written more quickly. 66 Similarly, West Syriac manuscripts appear from the late ninth or early tenth century that indicate vowels with miniature Greek letters, including alpha (Α), epsilon (Ε), eta (Η), omicron (Ο), and omicron/upsilon (ΟΥ). These signs supplanted the diacritic dot vowel system in the West Syriac world, and they remain in use today. 67 None of this is to say that Babylonian or West Syriac vocalizers necessarily modelled their signs after the shakl al-shiʿr or vice-versa, but none of these systems developed in a vacuum, and scribes of each language may have been aware of developments in the writing systems of the others.
The shakl al-shiʿr was just one of several methods which early Arabic writers used to record vowels in their language, and it ultimately gave way to more efficient notation methods with more abstract signs. The OIM-BNF-MIA Qurʾan is thus a witness to a relatively brief period in the history of Arabic writing, when standard practices for transcribing the language were still in flux. It provides only a small glimpse into the intellectual and social situations that may have led to its creation, and it raises as many questions about shakl al-shiʿr as it answers.
Most of these questions cannot be adequately addressed without evidence from additional manuscripts that contain this "marking of poetry." Such manuscripts might exist, with their letter-form vowel signs thus far escaping the notice of most scholars, but perhaps they do not. Either way, this Qurʾan is remarkable.