A Judaeo-Arabic Biblical Glossary as a Source for Arabic Historical Dialectology

MS T-S Ar.5.58 is a translation glossary from the Cairo Geniza that contains a list of Judaeo-Arabic glosses for Hebrew words from the biblical book of Samuel. These Arabic words are fully vocalised with the Tiberian Hebrew pointing system, providing more precise phonetic information about the scribe’s native Arabic dialect than could be expressed with standard Arabic vowel signs. This pointing reveals linguistic features known from modern varieties of vernacular Arabic, including a conditional tendency to raise /a/ to /e/ and a reflex of ǧīm as /g/. The manuscript can be dated between the tenth and twelfth centuries, making it an important source for the history of spoken medieval Arabic and Middle Arabic writing.


Introduction
A perennial problem of Arabic historical dialectology is the relative paucity of manuscripts that clearly record non-Classical forms. This problem is compounded by the fact that extant texts which do contain colloquial features are either unvocalised or vocalised with just a few Arabic vowel signs. Only a fraction of these texts are fully vocalised, but even with those we are limited by the Arabic writing system itself: the three Arabic vowel signs (fatḥa, kasra, ḍamma) are insufficient to record all of the vowel qualities in dialectal Arabic. A vocalisation system with more than three signs could, in theory, record additional allophones more precisely, but no such system was common in the medieval Arabic written tradition. 1 Likewise, the Arabic script has no way to explicitly indicate stress patterns, nor can it easily mark dialectal reflexes of Classical consonants.
However, the Tiberian Hebrew writing system has signs for seven discrete vowel qualities, a sign for marking unstressed syllables, and a dot that distinguishes between stop and fricative consonants. A few early medieval Judaeo-Arabic texts make use of these signs between the tenth and twelfth centuries. It is thus most likely that T-S Ar.5.58 was written between 900 and 1200. This range places it among the earliest known Judaeo-Arabic texts with complete vocalisation.
The Writing System of T-S Ar.5.58 T-S Ar.5.58 records Arabic glosses with 'classical' Judaeo-Arabic orthography. This orthography is the most common type of Judaeo-Arabic writing, and mimics Classical Arabic by transcribing each Arabic character with a single Hebrew character. It was used between the tenth and fifteenth centuries, 6 and contrasts the rarer 'phonetic' orthography, which recorded the phonetic realisation of Judaeo-Arabic rather than imitating Classical Arabic. 7 The phonetic system also reflects a greater tendency to use plene spellings for short vowels in Arabic. 8 Conversely, the classical orthography conceals much of its internal vowel phonology, resulting in a relatively standardised Judaeo-Arabic writing system that could be read by Jews in communities that spoke different varieties of Arabic. 9 The scribe of T-S Ar.5.58, however, wanted a more precise record for their glossary of Samuel, so they transcribed all of the Arabic vowels using Tiberian Masoretic vocalisation signs.
During the early medieval period, groups of Hebrew scribes and scholars known as 'Masoretes' 10 created vowel signs to vocalise the text of the Hebrew Bible. Their primary goal was preserving Hebrew recitation traditions in the midst of an Arabicising linguistic landscape, and they developed three different vocalisation systems in service of that goal. These included the Palestinian and Babylonian systems, which saw use respectively in Palestine and Iraq, as well as the Tiberian system, named after the Masoretes of Tiberias on the Sea of Galilee. 11 The Tiberian Masoretic tradition proved the most authoritative of the Hebrew systems, and the Tiberian vowel signs supplanted almost all other Hebrew vocalisation systems in the Middle East and Europe. 12 The majority of vocalised Judaeo-Arabic manuscripts contain Tiberian pointing. 13 This system has nine graphemes that originally represented seven vowel qualities. 14 These signs included the ḥolem /o/ (ֹ ‫אֹ‬ ֹ ֹ ), qameṣ /ɔ/ (ֹ ֹ ‫א‬ ֹ ֹ ), pataḥ /a/ (ֹ ֹ ‫א‬ ֹ ֹ ), segol /ɛ/ (ֹ ֹ ‫א‬ ֹ ֹ ), ṣere /e/ (ֹ ֹ ‫א‬ ֹ ֹ ), ḥireq (ֹ ֹ ‫א‬ ֹ ֹ ), and qibbuṣ /u/ (ֹ ֹ ‫א‬ ֹ ֹ ). Another sign, šureq, also represented /u/ when written with a mater lectionis letter vav ‫.)אּו(‬ Lastly, the šewa sign (ֹ ֹ ‫א‬ ֹ ֹ ) represented silence at a syllable break, or /a/, equivalent to pataḥ. 15 This latter šewa is known as 'vocalic' šewa, and it occurs where the Masoretes pronounced an epenthetic vowel in place of a historic lexical vowel, predominantly in unstressed, open syllables. 16 Qameṣ (/ɔ/) and segol (/ɛ/) do not appear in the Arabic of T-S Ar.5.58, but the other seven signs do. They all seem to retain their original Tiberian functions, which allowed the scribe to record allophonic features like imāla and to use šewa as a marker of Arabic stress patterns.

Vocalisation in Middle Arabic
Consistently vocalised Middle Arabic texts are about as rare as hens' teeth, so much of the evidence for non-Classical medieval vowel phonology comes from somewhat roundabout sources. Joshua Blau describes four in particular: a late ninth-or early tenth-century Greek transcription of Arabic, 17 several 'phonetic' Judaeo-Arabic transcriptions with plene short vowels, 18 a thirteenth-century Coptic transcription of Egyptian Arabic, 19 and a twelfth-or thirteenth-century 'classical' Judaeo-Arabic letter with full Hebrew vowel signs. 20 He deems this last text 'comparatively late' for his analysis of early Middle Arabic features, 21 but he includes it nonetheless, as it is uncommon for a classical Judaeo-Arabic text to contain more than a smattering of vowel points. Geoffrey Khan has likewise shown that a number of vocalised Judaeo-Arabic manuscripts reflect features of medieval colloquial Arabic, 22 but like Blau's late letter, most of his sources cannot be easily dated before the twelfth century. He does refer to three parchment Geniza manuscripts which may be earlier, including a copy of the siddur ('prayer book') of Saʿadiya Gaon, 23 a translation of Ecclesiastes, 24 and a commentary on a example: T-S Ar.5.12, T-S Ar.5.17, T-S Ar.54.31, and T-S NS 301.25. See VIDRO, "Arabic Vocalisation in Judaeo-Arabic Grammars," 341-51. liturgical poem. 25 Along with this small parchment corpus, 26 we may now add the translation glossary of T-S Ar.5.58 as another source of vocalised Middle Arabic.
These four manuscripts follow roughly the same vocalisation practices, but they also show significant variation. For example, the scribe of the Ecclesiastes translation used ṣere (/e/) only sparingly, while the siddur scribe applied it with reckless abandon. These differences reveal that the scribes who pointed these manuscripts did not all follow the same rules for vocalising Judaeo-Arabic, and their work likely reflects slightly different varieties of spoken Arabic. As such, the pointing system of each manuscript must be evaluated on its own, and any patterns must be derived first on internal evidence before comparing with other texts. The following sections examine T-S Ar.5.58 through this lens.

Methodology
This section contains an edition of T-S Ar.5.58, with the hope of making its linguistic data available to Arabic scholars who may not read Judaeo-Arabic. It is split into four sections, each containing two columns of lexemes from the manuscript. These columns are arranged along with their line and verse numbers, a transcription of the Judaeo-Arabic in Latin characters, and an English translation of the Arabic form. The transcriptions are as specific as possible according to the vocalisation in the manuscript, and I have generally avoided giving additional details that the scribe could not have conveyed with the system of signs available to them. Some of the Arabic glosses are not literal renderings of the Hebrew largest extant sample of vocalised Judaeo-Arabic in Geniza collections, and Khan cites it more than any other manuscript in his analyses of vocalised Judaeo-Arabic. See KHAN, "The Function of the Shewa Sign," 105, 107-8; KHAN,208,210  words, but it is beyond the scope of this paper to analyse the lexical and theological implications of these differences. In some cases, I have reconstructed a vowel or part of a word where the text or vowel points were omitted or damaged. These reconstructions are indicated by [square brackets]. A few letters have (curved brackets), which indicate that they were most likely quiescent in speech. It must be noted that the vocalisation in the manuscript appears to reflect the scribe's aural perception of each word while reading aloud from a translation of Samuel. This context may have influenced their perception of vowel length and stress positions, but we cannot access this layer of information without the full translation of 1 Samuel that the glossary belongs to.
In addition to the folio's main columns, there are several notes in the margins of the recto. They are in a different hand from the primary text and relate to the lexical items. I have included these annotations as footnotes when it is possible to decipher them.

A Note on Šewa
In Hebrew recitation, the šewa sign (ֹֹ ‫א‬ ֹ ֹ ) does not inherently represent any one vowel quality.
Instead, it marks either silence (like sukūn) or an epenthetic short vowel, usually in an unstressed, open syllable. The quality of this 'vocalic' šewa can range between several different vowels (e.g. /a/, /e/, and /ə/) depending on its phonetic context and the particular reading tradition. In the majority of Tiberian Hebrew contexts, it was pronounced with a neutral open quality /a/. 27 Vocalic šewa in the Arabic of this text also predominantly denotes a short vowel in an unstressed, open syllable. Following the standard Tiberian usage, this sign likely represented epenthetic /a/ in most places, and it corresponds to positions where Classical Arabic has fatḥa. This epenthesis may correspond to broader Middle Arabic trends of reducing short vowels in unstressed, open syllables. 28 I have transcribed these instances of vocalic šewa (as well as the composite šewa sign, ḥateph pataḥ ֹ ֹ ‫א‬ ֹ ֹ ) as ạ, which should be interpreted as representing an open or open-mid short vowel. Some of these vowels could be greatly reduced in quantity, almost to zero, but the šewa sign does not specify their exact length. There are also three instances where vocalic šewa likely indicates /i/ due to a correspondence with the vowels of Classical Arabic particles. I have transcribed these with ị, which represents a short, front vowel in an open syllable, the precise quality of which can only be assumed from context.

17:13
and his second wạ-ṯēnīh he would go day after day and he supplied the troops wạ-ʾaḥḍir ʾal-ʿaskar 17:18 wheels of cheese 33 ʾaqriṣa al-gubun 30 The 3ms object and possessive suffixes are consistently written as qibbuṣ before haʾ (i.e. -uh). This form is probably an imitation of Arabic orthography, and was pronounced -u or -ū. 31 Conjunctive wāw is almost always transcribed as vav with šewa, imitating the Hebrew orthography.
32 Tāʾ marbūṭa is usually represented by haʾ ‫,)ה(‬ imitating Classical Arabic orthography. This haʾ was not pronounced as a consonant. 33 There is a note between columns 1 and 2 that corresponds to this gloss. It reads: ; 'the chunks of cheese, the measure of the […] is their largeness(?).' It is mostly unvocalised, but in contrast to the vowel points on al-gubun ('cheese') in the main text, this later hand writes it with two hireqs: al-jibin (or al-gibin). There are not enough marginal notes to say whether this second writer also meant to record a stop-plosive reflex of ǧīm.     There is probably a dageš in the gimel of this word, but a stain on the parchment obscures it. 42 The vav in this word is almost rubbed off. Only one dot remains below it, but the original sign was likely ṣere. 43 The expected Classical Arabic orthography has alif ‫;مخالة(‬ miḫlā), but the Judaeo-Arabic spelling is defective.

54
The gloss is unvocalised, and appears to be a direct borrowing of the Hebrew without changing it to an Arabic form (e.g. maḥuliyya).
55 This word appears to have epenthetic /i/, with the expected form being yakmạlū. 56 The scribe must have meant something like yaḫūf or yaḫāf ('he fears') here, but the remaining text looks more like yaḫfā ('he is hidden').

Observations and Analysis
Almost every vowel sign and diacritic dot conforms to its expected usage in Tiberian Hebrew, and it is clear the scribe was well-versed in the details of Tiberian pointing. This regularity allows for a confident reconstruction of the intended Middle Arabic vowel phonology in many of the glosses. Additionally, when viewed through a standard Tiberian lens, the use of dageš dots appears both regular and systematic. This consistency reveals the scribe's pronunciation of certain Arabic consonants as either stops (e.g. ‫)ج‬ or fricatives (e.g. ‫ذ‬ and ‫.)ث‬
59 Except in V2.3, where the construct form gūṯat ‫ֹת(‬ ‫גּות‬ ֹ ֹ ֹֹ; 'the corpse of') is spelled phonetically with tav. This spelling for the construct form of nouns ending in tāʾ marbūṭa is common in 'phonetic' Judaeo-Arabic orthography, in contrast to 'classical' Judaeo-Arabic; KHAN, "Judaeo-Arabic," 150. 60 Although see LEVIN, "The Imāla," XIX, as he finds that imāla of the vowel represented by -‫اة‬ is rare in modern Arabic dialects. /i/, or open with vocalic šewa. 61 It seems that in this scribe's Arabic dialect, such a syllable structure could induce a final /a/ to be raised and fronted, resulting in a high vowel somewhere between /e/ and /i/. This vowel raising correlates with the imāla of final /a/ known from both medieval and modern varieties of Arabic, 62 and is also recorded with plene spellings of yod in other medieval Judaeo-Arabic texts. 63 Similar raising occurs in words where Classical Arabic would have /ā/ in an open syllable. When such a vowel precedes a syllable with /i/, then the /ā/ is raised to /ē/: wạḥēmil ‫ֹל(‬ ‫ֹאמ‬ ‫ֹח‬ ‫ו‬ ֹ ֹ ֹ ֹ ֹ; 'and a bearer;' R2. Applying these rules to damaged areas of the manuscript, it is possible to extrapolate some missing vocalisation. The first vowel points on the word ‫ֹאת‬ ‫ֹּב‬ ‫אלאע‬ ֹ ֹ ֹ ֹ ֹ ֹ ֹ are too badly rubbed to read, but given that the Classical Arabic form of this word would have /ā/ in an open syllable preceding a syllable with /i/, the original vocalisation was probably with ṣere: ʾ[a]l[ē]ʿibāt ‫ֹאת(‬ ‫ֹּב‬ ‫ֹ]אע‬ ‫א[ל‬ ֹ ֹ ֹ ֹ ֹ ֹ ֹ ֹֹ ; 'the playing women;' V2.19). However, there are exceptions to this rule, and it seems some consonants prevented this vowel raising when they occurred before or after /ā/, including: ḫāʾ in tūḫātin ( ‫ת‬ ‫ו‬ ‫ן‬ ֹ ‫ֹאת‬ ‫כ‬ ֹ ֹ ֹ ֹ ; 'you will become a son-in-law;' V4.9) and qāf in ʿāqil ‫ל(‬ ֹ ‫ֹאק‬ ‫ע‬ ֹ ֹ ֹ ֹ ; 'surpassing;' V4.18). This phenomenon corresponds to the effect that pharyngeal and emphatic mustaʿliya letters have, preventing imāla of nearby a-vowels. 65 There is also one gloss that contains a ḥolem sign, indicating an Arabic word with the vowel /o/: wa-hạlōm ‫ֹלֹום(‬ ‫ֹה‬ ‫ו‬ ֹ ֹֹ ֹ ֹ; 'and onwards;' V4.1), analogous with the Classical Arabic halumma ( ‫م‬ ‫ل‬ ‫ه‬ ; 'onwards'). The ḥolem is written with a plene letter vav, even though no wāw appears in the Classical Arabic orthography, suggesting that this vowel was pronounced long (/ō/). There is also no indication of gemination on the final mem.
The signs segol (/ɛ/) or qameṣ (/ɔ/) do not appear in the Arabic columns, and indeed there is little reason for them to, as neither represents a cardinal Arabic vowel. By contrast, they both occur often in the Hebrew columns, closely matching the expected forms from 1 Samuel 17-19. These Hebrew forms do, however, differ in a few small details. 66 In particular, there are two instances where the scribe used a segol in a position where the standard Masoretic text has pataḥ. They write ha-nɛʿɛr ‫ֹר](‬ ‫ֹ[ע‬ ‫ֹנ‬ ‫ה‬ ֹֹ ֹ ֹ ֹ ֹ ; 'the boy;' V1.11) when the expected form is ha-naʿar ‫ֹר(‬ ‫ֹע‬ ‫הנ‬ ֹ ֹ ֹֹ ), and seḫɛl ‫ֹל(‬ ‫ֹכ‬ ‫ש‬ ֹ ֹ ֹ ; 'he behaved;' V3.18) instead of the expected sɔḫal ‫ֹל(‬ ‫ֹכ‬ ‫ש‬ ֹ ֹ ֹ ). These interchanges of segol for pataḥ suggest that the scribe per-61 Šewa in these cases may represent unstressed short /e/, but there is no way to know for sure from the signs in the manuscript. ceived /a/ and /ɛ/ as allophonic in Hebrew, a situation likely conditioned by the comparatively small vowel inventory of their native Arabic. As such, there was certainly no reason for them to use segol to transcribe allophones of /a/ in the Arabic glosses, since all of its phonetic functions could be covered by pataḥ. This usage contrasts the scribe's marking of /e/, which they perceived as distinct enough from /a/ to warrant the use of the ṣere sign in Arabic.
On the other hand, there are no clear Hebrew interchanges related to qameṣ (/ɔ/), and the scribe had no trouble differentiating it from pataḥ (/a/) in the Hebrew words. This phonological understanding implies that they distinguished between /ɔ/ and /a/ in their Hebrew pronunciation. If so, then they still maintained a key element of the Tiberian pronunciation tradition, in contrast to the more common Palestinian and later Sephardi Hebrew traditions, where pataḥ and qameṣ were both realised as /a/. 67 This detail suggests that the text was vocalised prior to the extinction of the Tiberian pronunciation system, no later than the eleventh century. 68 Silent šewa conforms to its Tiberian usage in the Arabic glosses, consistently marking the close of a syllable (like Arabic sukūn). Meanwhile, vocalic šewa occurs frequently and consistently to indicate an open or near-open vowel, predominantly in unstressed, open syllables. These syllables seem to be places where the scribe pronounced a short lexical vowel or a short epenthetic vowel that corresponds to a lexical vowel in Classical Arabic (usually fatḥa). Most of these vocalic šewas were probably realised as /a/, the same quality as Tiberian šewa and equivalent to pataḥ. However, it does seem that the scribe used šewa deliberately in contrast to pataḥ to highlight that a syllable was unstressed and open. Šewa also indicates /i/ several times, 69 corresponding to the vowel in the Classical Arabic particles biand li-: lị-yūʿayyir ‫ֹר(‬ ‫ֹיי‬ ‫ֹיּוע‬ ‫ל‬ ֹ ֹֹ ֹ ֹֹ ֹ ; 'to condemn;' R2.17); bị-laḥiyuh ‫ֹה(‬ ‫ֹי‬ ‫ֹח‬ ‫ֹל‬ ‫ּב‬ ֹ ֹ ֹ ֹ ֹ ; 'by its beard;' R4.6); lị-yuġannī ‫ֹי(‬ ‫ֹנ‬ ‫ֹג‬ ‫ֹי‬ ‫ל‬ ֹ ֹ ֹ ֹ ֹ ; 'to sing;' V2.17). The scribe also uses the combination sign of šewa and pataḥ (i.e. ḥateph pataḥ), but this usage is solely an orthographic convention to avoid marking a guttural letter with vocalic šewa. This practice matches the standard Tiberian rule for ḥateph vowel signs, and implies that hateph pataḥ is phonetically equivalent to šewa in the Arabic glosses. 70 Besides the expected usage in unstressed, open syllables, there are a few instances where it seems vocalic šewa must be in a closed or stressed syllable. This notation differs from the šewa in Tiberian Hebrew, where it cannot indicate a stressed vowel and only rarely appears in closed syllables. 71 For example, the first syllable of ʾạnā ʿayyarat (ֹ ‫ֹא‬ ‫ֹנ‬ ‫א‬ ֹ ֹ ֹ ֹ ‫ת‬ ֹ ‫ֹיר‬ ‫ֹי‬ ‫ע‬ ֹ ֹ ֹ ֹ ֹ ; 'I condemned;' R2.5) is marked with ḥateph pataḥ, unexpectedly suggesting that the first vowel is unstressed (i.e. the most likely reading is with stress on the second syllable (mantáqatuh), and yet the tav has šewa. The phrase wạ-nazạʿhum ( ֹ ֹ ‫ֹה‬ ‫ֹע‬ ‫ֹז‬ ‫ֹנ‬ ‫ו‬ ֹ ֹ ֹ ֹ ֹ ֹ ‫ם‬ ; 'and he removed them;' R4.11) is even more difficult to interpret, as the vowel after the zayin is marked by šewa, even though it seems that that syllable is both stressed and closed (nazáʿhum). These potentially irregular stress patterns do not match the expected stress patterns of modern Egyptian colloquial Arabic, 72 despite the likely Egyptian provenance of this manuscript.
There are two other places where a vocalic šewa appears in a closed syllable. Both mark the vowel before the kaph of a 2ms suffix: qiḥatạk ‫ּך(‬ ֹ ‫ֹת‬ ‫ֹח‬ ‫ק‬ ֹ ֹ ֹ ֹ ; 'your insolence;' R4.3) and yūsallimạk ‫ּך(‬ ֹ ‫ֹמ‬ ‫ֹל‬ ‫יּוס‬ ֹ ֹ ֹ ֹ ֹֹ; 'he will deliver you;' V2.2). 73 If the scribe pronounced this suffix as -ak -the same as modern Egyptian and Levantine dialects -then one would expect pataḥ (/a/) in these closed syllables. If these šewas do represent /a/, then the scribe may have had an orthographic policy specific to this suffix that permitted šewa in a closed syllable. Such a practice would approximate the graphic appearance of the equivalent Tiberian Hebrew suffix -ֹֹ ְ ֹ ֹ ‫ָך‬ (-aḫɔ). Without more evidence, it is difficult to determine the exact functions of šewa in every instance. The scribe may simply have conflated šewa and pataḥ due to their equivalent qualities, or they may have recorded šewa occasionally for what they heard as ambiguous vowels in closed or stressed syllables. At any rate, it is clear that they predominantly used vocalic šewa to record short vowels in unstressed, open syllables.

Dageš as a Marker of Arabic Stops and Fricatives
The twenty-two letters of the Hebrew alphabet are insufficient to transcribe the twenty-nine Arabic letters on a one-to-one basis, so in Judaeo-Arabic a few Hebrew characters each represent two different Arabic consonants.
Judaeo-Arabic scribes often addressed this consonantal ambiguity by placing a diacritic dot or stroke above a Hebrew character to indicate that it represented an Arabic letter that had no Hebrew equivalent. Historically, the first letters to receive this treatment were Hebrew ṣade ‫,)צ(‬ ṭet ‫,)ט(‬ and gimel ‫,)ג(‬ 74 which took dots to indicate Arabic ḍād (ֹ ֹ ‫צ‬ ֹ ֹ ), ẓāʾ (ֹ ֹ ‫ט‬ ֹ ֹ ), and ǧīm (ֹ ֹ ‫ג‬ ֹ ֹ or ֹֹ ‫ג‬ ֹ ֹ). 75 This system eventually expanded with diacritic dots on other Hebrew letters, but at first the new dots were a last resort, used only for Arabic phonemes (i.e. /ḍ/, /ẓ/, and /ǧ/) that did not exist in Hebrew phonetics. For other letters, instead of adding diacritics, early Judaeo-Arabic scribes preferred to use the Hebrew writing system to the fullest extent possible to indicate Arabic phonemes. This preference led to the application of the Hebrew dageš dot to Judaeo-Arabic to differentiate between stops and fricatives. 72 MITCHELL,An Introduction to Egyptian Colloquial Arabic, This notation also occurs in the Saʿadiya siddur from T-S Ar.8.3; see KHAN, "Vocalized Judaeo-Arabic," 210. By contrast, the 2ms ending is always written with pataḥ in the Ecclesiastes translation from T-S Ar.27.55, T-S Ar.53.12, and L-G Ar.I.150.
Based on the unambiguous readings in the manuscript, the distribution of dageš is as follows:  dageš marks gemination 10 times  bet occurs 11 times with dageš, and 6 times without; it always represents /b/ ( ‫ب‬ )  peʾ occurs 6 times, never with dageš; it always represents /f/ ‫)ف(‬  gimel occurs 8 times with dageš, every time representing /g/ ( Altogether, the scribe marks Arabic stops with dageš 58 times, while fricatives always occur without dageš (27 times total). There are only 13 instances where the scribe does not mark a stop with dageš. Six of these are bet, 77 which only ever represents a bilabial stop in Judaeo-Arabic, so pointing it with dageš at all is a redundant practice that the scribe retained from Hebrew. The other seven include one medial kaph, 78 one medial dalet, 79 and five tavs in final position. Two of these tavs are in feminine plural endings, 80 two are 76 There are three cases where there is a gimel in the text, but the manuscript is damaged and ambiguous as to whether they had dageš or not. second;' R2.8), without dageš in the tav (/ṯ/); and gūṯat ‫ת(‬ ֹ ‫גּות‬ ֹ ֹ ֹֹ; 'the corpse of;' V2.3), with dageš in the gimel (/g/) and final tav (/t/), but not in the first tav (/ṯ/). This consistent notation suggests that the scribe consciously distinguished these interdental consonants from their alveolar counterparts, definitely in writing, and probably in speech.

Conclusion
T-S Ar.5.58 is a parchment folio from the Cairo Geniza that contains part of a Judaeo-Arabic translation glossary for the Hebrew book of Samuel. It was most likely produced between 900 and 1200, and probably before the end of the eleventh century. The Arabic words in this glossary are fully vocalised with Tiberian Hebrew vowel signs, and this pointing system allowed the scribe to record vocalic allophones -most notably /e/ -from their native Arabic dialect. They also used the Tiberian šewa to indicate stress patterns in Arabic words, and their distribution of the dageš dot suggests that they realised Arabic ǧīm as a voiced velar stop /g/. These details would not be evident if the text were written in Arabic script, which makes this manuscript a unique source for the phonetic features of medieval vernacular speech.
More research is required to fully understand the linguistic features of this text, particularly with respect to its lexical inventory and verbal morphology. Further comparative study is also needed to determine the relationship between this manuscript, other vocalised Judaeo-Arabic manuscripts, and Middle Arabic texts more broadly. Such work is for the future, but it is hoped that the present edition makes these rare data points more accessible to all scholars of Arabic.