ʾinǧīl-in mubīn: A mixed archaic, Quranic, and Middle Arabic translation of the gospels and its implications for the nature of Middle Arabic

Abstract This paper undertakes a close linguistic study of a unique translation of the gospels into Arabic as attested in three manuscript witnesses. The translation is unique insofar as it imitates the Quran, especially in lexicon and rhyme. Linguistically it mixes numerous features specific to the Quran with features from both the Classical Arabic (ClAr) tradition, including poetic archaisms not typical of standard ClAr, as well as from Christian Middle Arabic. I argue that the regnant framework for Middle Arabic – that it exists on a spectrum from dialects to standard Classical Arabic – is insufficient for understanding this text. Instead, we need to conceptualize the high register for at least some communities as encompassing distinctively Christian features, which originated as living features and had achieved prestige, along with ones from Classical Arabic and Quranic recitation traditions, and even Old Hijazi.


Introduction
Scholarly understanding of the nature of the linguistic variation characteristic of premodern Middle Arabic texts has evolved in recent decades, from it being a distinct linguistic phase corresponding to a period of history (as in, e.g., Blau 1977), to a sociolinguistic phenomenon which has existed for as long as Arabic has existed, and which continues to the present (Lentin 2008;Khan 2011;Den Heijer 2012).Despite general acceptance of this re-conceptualization of Middle Arabic, many questions remain about both the linguistic origin and development of specific features attested in Middle Arabic, as well as whether, and to what degree, these features constituted a normative register at which a particular author was aiming.One such debate is the ongoing question of whether non-Classical Arabic features attested in Middle Arabic texts originated as hypercorrections, reflecting, originally at least, a lack of grammatical training or competency (as, e.g., Blau 1966-67;1970;1999;see also Hary 2007), or were deliberate but unsystematically used colloquialisms (as in, e.g., Lentin 2008: 219), or rather reflect the systematic use of innovative hybrid compromises between Classical Arabic and dialectal grammars (Bellem and Smith 2014).Despite differences in approaching and answering such questions, however, scholars of Middle Arabic by and large share the fundamental assumption, namely that Middle Arabic as a phenomenon "is a continuum, or a mix, between the H variety (usually identified as Standard or Classical Arabic) and the L variety (colloquial Arabic, also dubbed Neo-Arabic" (Den Heijer 2012: 6, emphasis mine).In other words, even once distinct Middle Arabic registers formed, thereby becoming prestigious vehicles of literary production in their own right (as, among others, Classical Judaeo-Arabic), they still originatedand remained sociolinguistically situatedin the middle of a spectrum between standard Classical Arabic at the high end, and colloquial dialects at the low one.
In this paper I seek first to challenge this shared assumption and, second, to add to the typology of Middle Arabic features, in the context of a close linguistic study of a unique medieval gospel translation in Arabic preserved in three manuscripts: Vatican Arabic 17 (1009 CE), Vatican Arabic 18 (993 CE), and Leiden Or. 561 (fifteenth c.CE).The translation is unique insofar as it imitates the style and lexicon of the Quran, even when these differ from what became normative Classical Arabic.Linguistically, forms specific to the Quran are mixed with features from not only the ClAr tradition, including, e.g., poetic archaisms not associated with standard ClAr, but also from Christian Middle Arabic, and even possibly colloquialisms.The three manuscripts attest very little variation in the consonantal frame of the translation; however, Vatican Arabic 17 and 18 (VAr.17 and VAr.18) are nearly fully vocalized, whereas Leiden Or. 561 (henceforth Or. 561) is much less, though not rarely, vocalized.And while the vocalization and orthoepic markings in the two earlier manuscripts, VAr. 17 and 18, are very similar to Classical Arabic, there are nevertheless several differences which have likely parallels in the language of the Quran.The phonology attested in the vocalization layer of the later manuscript, Or. 561, is much further from Classical Arabic, being virtually identical to that attested in other vocalized Christian gospel manuscripts from the medieval period.Or. 561 thus represents a deliberate Christianization of the recitation of the gospel.
There are several important implications of the data from the translation, and the manuscript witnesses to it, for the scholarly study of Middle Arabic.First, what we think of as typical, normative Classical Arabic (ClAr) features were not the only ones with which authors were interacting and mixing during the medieval period.In fact, the datahere and elsewhereincreasingly suggest the early linguistic diversity, reported by the early grammarians and attested in, e.g., vocalized Quranic manuscripts, was not forgotten completely but remained in use, even prestigious, among some groups of Christians for longer even than among Muslims.Second, I argue that several of the salient Christian phonological features attested in the vocalization layer of Leiden Or. 561 (such as tanwīn) which are both non-ClAr and also absent from any modern dialect, nevertheless originated as living features.At some point these features died out in the living speech of many, even eventually most, but not before becoming prestigious variants in their own right.As such, they were not ever pseudo-corrections, not always living colloquialisms (insofar as they were likely not living features of all scribes who employed them), nor ClAr/colloquial grammar hybrids; rather, they belong to a fourth category made up of once-living features which subsequently became prestigious variants for composition and recitation of (at least) the gospels.Such features were mixed with ones from ClAr or Quranic recitation traditions, as well as still othersarchaic and innovativein a high (H) register that was much more dynamic than conventionally conceptualized.I conclude that what is now normative ClAr represented, for at least some Christian communities, but one of several clusters of prestigious forms that might be used and mixed.
In what follows I begin with a discussion of several salient linguistic features attested in the consonantal skeleton of the text.These features are common to all three manuscripts.Subsequently, I discuss the vocalizations attested in VAr. 17 and 18, followed by those of Or. 561.I conclude by elaborating on the implications of these manuscripts for the study of Middle Arabic.

Description of the MSS
The three manuscript witnesses to this Quranic style gospel translation are: Vatican Arabic 17 (VAr.17), Vatican Arabic 18 (VAr. 18),and Leiden Codex Or. 561 (Or. 561).The earliest of the three manuscripts is VAr.18, which was produced according to the colophon in 993 CE.It contains only the Gospel of Luke, although it originally contained a complete translation of the four gospels.It consists of 93 folia of paper.VAr. 17 was produced just a few years later, in 1009 CE, and is a virtually complete copy of the four gospels.It contains 299 folia of paper.The colophons of these two codices indicate that they were in use both in Egypt and Constantinople, perhaps in contexts in which Christians and Muslims shared close social connections and in which Christians might want a gospel text that was comprehensible, and linguistically respectable, to Muslims (Kashouh 2012: 128).Indeed, VAr. 17 was copied by a Muslim named Ḥamdān ʿAlī.Finally, Or. 561 is the latest copy of this version, datable to the fifteenth century CE.It consists of 151 folia of paper.Kashouh has suggested that this translation represents a translation from the Peshitta, independent of other translations attested in the Christian gospel manuscripts (Kashouh 2012: 128-30).
Each of the three manuscripts is available online.1 Of the three manuscripts, Or. 561, is still clear and well-scanned.VAr. 17 is legible in most places, although the paper has darkened so as to make legibility challenging in some contexts.VAr.18 is the least legible of the three manuscripts, with darkening of the paper presenting a challenge, as well as an apparent smudging of the ink used.
While there is relatively little difference between the consonantal texts of the three manuscripts, there are some orthographic differences, especially between VAr. 17 and 18 on the one hand, and Or.561 on the other.More significantly, the vocalizations differ between the two groups.In VAr. 17 and 18, the vocalizations are essentially identical to what one would find in a normative ClAr manuscript.Vocalization is nearly ubiquitous, with a few exceptions (on which see below).In Or. 561, on the other hand, vocalization is much sparser, though not rare.Significantly, the linguistic patterns attested in the vocalization layer of Or. 561 differ from VAr. 17 and Var. 18, but with features in common with other Christian manuscripts.

Linguistic notes on the consonantal skeleton
In the description of the consonantal text of the translation I draw from Leiden Or. 561 unless otherwise noted.This decision is due not only to the consistent clarity and legibility of the text in that manuscript compared with the other two, but also because it is the most complete.As noted previously, there are very few differences, and those differences are typically lexical rather than grammatical (Kashouh 2012: 129; see also a collation of his test passages in appendix one, §12.6).The data, description, and examples in this and following sections are drawn from a systematic analysis of the entirety of the Gospel of Matthew, and 20 per cent each of the Gospels of Mark, Luke, and John.

Orthography and phonology
Each of the manuscripts is written in naskh script, and the orthography generally conforms to the norms of Classical Arabic with some notable exceptions, which I will discuss briefly here.Note that throughout this paper, the Arabic text is replicated from the manuscript exactly wherever possible.If this is not possible due to the limits of type font, I note the difference in parentheses.
Archaic spellings occur throughout the translation, often standing alongside later, more standard, spellings of the same words.One common orthographic archaism is the absence of word-internal alif to indicate a long vowel, especially in the words salām, e.g."ask (mpl) and you (mpl) will be given" (8b).This hybrid form, between Quranic ‫ﺳ‬ ‫ﻞ‬ and ClAr ‫ﺍ‬ ‫ﺳ‬ ‫ﺄ‬ ‫ﻝ‬ , suggests the presence of the medial hamzah of the root, /isʾal/, rather than the Hijazi /sal/, and it is indeed written with a hamzah above the sīn in VAr. 17 (180r).Outside of these contexts, as well as other words which even in ClAr orthography are never spelled with alif, such as the demonstratives hāḏā / hāḏihī / ḏālika, long ā is spelled with alif consistently.This strongly suggests an intentional archaizing of the spelling in certain places, closerthough not always identicalto the Quranic spelling where it differs from ClAr.
Another category of words spelled variably, with archaic and standard spellings, are ones which in the Quran are spelled with a final <-wh> / -‫ﻭ‬ ‫ﻩ‬ when absolute.Perhaps the most saliently Quranic features of the translation is the frequent spelling of assimilation with verbal V and VI, which occur alongside unassimilated ones, just as in the Quran (Fischer 2002: §47), listed in Table 1.
The example from folio 7a, ‫ﺍ‬ ‫ﺗ‬ ‫ﺮ‬ ‫ﻭ‬ ‫ﻱ‬ <ʾtrwy> attests a prothetic alif but is etymologically unclear.The context, hypocrites wanting to be seen praying (Matthew 6: 5 ff.), suggests it is a form V from the root *rʾy, in which case it perhaps reflects a variant from a variety in which hamzah had been lost and the middle root consonant shifted to a w, although this is the only example of such a variant in the portions of the manuscripts studied for this paper.Another assimilation pattern that occasionally occurs is the spelling of the nūn of the preposition min, "from", to following nouns, typically when the noun is prefixed with the definite article, listed in Table 2.
Unlike the assimilation in verbal forms V and VI, this type of assimilation is not found in the Quran, nor is it standard ClAr.These spellings could therefore reflect a living feature.This is not the only option, however, and given the otherwise standard nature of the orthography, we should consider other alternatives.In fact, a parallel in non-standard ClAr practice is attested.Wright lists the following kind of variation involving the preposition min among so-called poetic licenses: , "Fate has left no wealth in our possession", where Wright II: §242,.If the latter phenomenona poetic or H register variant used in the text for stylistic purposesis correct, then it provides another example of the non-normative ClAr features which are sprinkled throughout.
One example the parsing of which is not totally clear is: <wʾmṣyʾm>, "And this is not possible except with faithful prayer and fasting".
The prefixed mīm is familiar from the other examples of the assimilated form of min.Different, however, is the prothetic alif, which is absent in the other examples of assimilated min, as well as the absence of the definite article.The previous noun is preceded by the preposition bi, bi-ṣalāt "with prayer", and it seems likely that the preposition would have been the same, i.e. bi-ṣalāt … wa-bi-ṣiyām.In this case, the mīm underwent an unexpected nasalization of sorts, from the bilabial stop to nasal, perhaps due to an anticipatory assimilation to the final mīm, although this is not at all certain.The prothetic alif is most likely interpreted as representing a prothetic I, thus /im-ṣiyām/.
Despite the presence of clearly archaic, and in some cases specifically Quranic spellings, there are differences between the spelling of some categories of words in the Quran and the present translation, which in this case follows the ClAr orthography.The most prominent of these cases is the spelling of etymological III-Y roots with pronominal suffixes.In ClAr orthography, III-Y verbs are spelled with alif maqṣūrah when word-final; however, before pronominal suffixes it is spelled alif mamdūdah: ‫ﺑ‬ ‫ﻨ‬ ‫ﻰ‬ "he built", but ‫ﺑ‬ ‫ﻨ‬ ‫ﺎ‬ ‫ﻩ‬ ُ <bnʾh> "He built it".In the Quran, however, both of these are spelled with the yāʾ: ‫ﺑ‬ ‫ﻨ‬ ‫ﻰ‬ and ‫ﺑ‬ ‫ﻨ‬ ‫ﯩ‬ ‫ﻪ‬ <bnyh>.This reflects the fact that, phonetically, in the language behind the Quranic consonantal text III-Y verbs were realized with final /ē/ or /ay/, not ClAr /ā/ (Van Putten 2017: §6- §7, pp.55-9).Already by the eighth and ninth centuries CE, the ClAr spelling of these forms is attested ubiquitously in the corpora of the papyri and inscriptions (Van Putten 2022: §5.8, p. 118).Thus, in terms of the combination of orthographic features, this translation is Quranic and archaizing but only to the point of imitating practices of the post-eighth century CE period.Perhaps the most salient difference between the patterns of consonantal orthography in the translation, and Quranic and ClAr orthography, concerns the use of alif, especially to indicate the presence of the glottal stop in ways that classical orthography does not.Indeed, alif is used in numerous places to spell a glottal stop where, in ClAr orthography, either a glide or the hamzah would be used, e.g. ( 11b) ‫ﻭ‬ ‫ﺑ‬ ‫ﺮ‬ ‫ﺃ‬ ‫ﺕ‬ <brʾt> "she was healed" (= ClAr ‫ﺑ‬ ُ ‫ﺮ‬ ‫ﺋ‬ َ ‫ﺖ‬ ).In most cases of non-Quranic and ClAr spellings, word-internal combinations of vowels and the glottal stop lead to sequences of two consecutive alifs, which, along with sequences of two glides, were disallowed in Quranic spellings.In a few cases, the sequence spells etymological combinations of short vowels and glottal stop, as in the examples listed in Table 3.Most such sequences spell combinations of glottal stop and a long ā, shown in Table 4.
These spellings become even more frequent in Middle Arabic texts from the Ottoman period, for example (Lentin 1997: 111-2).One final orthographic usage of the alif that, though rare, occurs in the translation and is contrary to Quranic and ClAr orthography is the occasional use of alif presumably to mark overlong ā in the perfect 3ms of the verb qāl, e.g.(9b) ‫ﻗ‬ ‫ﺎ‬ ‫ﺍ‬ ‫ﻝ‬ "he said" <qʾʾl>.While the pattern of non-ClAr usage of alif to mark combinations of a vowels and the glottal stop is medieval and later, it is not qualitatively unique.Indeed, as Puin (2011: 170, 173) notes, such spellings of ʾ with alif are attested in an early Quranic manuscript, Ṣanʿāʾ DAM 01-29.1.Despite Puin's arguments for the historical anteriority of this practice, Van Putten (2018: 114-6) argues convincingly that this practice was innovative.Thus from very early on, and increasingly over time, the absence of a single letter with a single shape that represents glottal stop, especially given the dual role of alif to mark both glottal stop and long ā, gave rise independently to several scribal practices which use alif to mark glottal stop, even when it contradicts ClAr norms.
Phonetic spellings occasionally occur and are worth mentioning here.First, in a handful of cases, a prothetic alif is written before verbs that in other varieties of Arabic are form I: was healed"), or active, referring to Jesus (i.e."When he had healed").In either case, since form IV is virtually unattested with this verb, it is likely that either underlying /iṭhar/ or possibly passive /Vṭhir/ is indicated.2Lastly, in one place, etymological *ẓ is spelled with a ḍād instead of ẓāʾ, the sole example of an otherwise well-known and attested phenomenon already in the earliest Islamic era documents (Hopkins 1984: 40 .These phonetic spellings often (although not always) occur in contexts of direct speech, which might reflect their intentional use to create contrast between the refined narrative and character speech.If so, it is a helpful reminder that such colloquialisms need not reflect inability on the part of the composer of a text, but can be artfully and skilfully mixed with even quite refined texts, such as this.

Morphology
The present section presents phenomena that are saliently distinct from standard ClAr orthography, which include archaisms of various kinds, varied spellings of the same underlying form, and innovative spellings based on analogy.I address each of these categories in turn.
A number of non-ClAr archaic morphological forms are used, often alongside the standard ClAr forms.This is most commonly seen in the various forms of the demonstrative and relative pronouns used.According to the early grammarians, there were morphological differences between the demonstrative pronouns in eastern (Najdi) Arabic dialects and those of the Hijaz.The Hijazi forms are characterized by a final -h on the feminine singular, which is lacking from the Najdi form: Hijazi hāḏihī vs. Najdi hāḏī.The distal set is distinguished by the presence of an element l(i) in the Hijaz that is absent in the Najdi forms: Hijazi ḏālika (msg) / tilka (fsg) vs Najdi ḏāka (msg) / tīka (fsg) (Sībawayh "And that (man) was wise".
Additionally, addressee agreementin which a suffixed -kum is added to various demonstrative and deictic forms when addressing a group of peopleis well-attested in the manuscript, always with the Hijazi forms with infixed l(i): "That is my blood".This variant is well-attested in the modern dialects, from the Levant to Morocco (Vicente 2006: 569-72).However, it is also an archaic feature (Wright I: 268).With cases such as these, where a form is both archaic and dialectal, it can be difficult to determine the proper explanation of its occurrence.Yet perhaps we do not have to make such a choice.It seems to me quite likely that both aspects can play a role.The composer of this translation clearly chose intentionally from among a number of variants, but the fact that this form was both archaic and dialectal, but rare in writing, could have made it all the more appealing.The more frequent form is ḏālik(a), well known from the Quran and ClAr, e.g. ( 12b) The relative pronouns attested in the version are Hijazi, which are also those attested in standard ClAr, in each of singular, dual, and masculine plural contexts: "Woe to the pregnant women, and those who are breastfeeding in those days".The selection of demonstrative and relative pronominal forms is clearly intentionally archaic, in this case incorporating forms not common in standard ClAr.But whereas elsewhere the Quranic forms were less common, here the Quranic forms became standard for ClAr, and the use of archaic ones resulted in deviation from Quranic morphology.Still, the productive use of addressee agreement in demonstrative and other deictic forms clearly draws on the Quran against later ClAr norms.
There are several other cases in which archaisms are used alongside forms which are more common and more typical of ClAr.For example, in addition to the use of ladā, "at, by the side of", the archaic form ladun is attested, in combination with the preposition min, as well as with pronominal suffixes:  (Harrell 1962: 63).It is thus possible that, as with the case of hāḏāk above, we are dealing with forms chosen both for their local familiarity and the fact that they are accepted archaic forms which nevertheless go against the standard ClAr forms.
Also worth mentioning among the general stock of archaic forms used in the translation is the vocative suffix -āh, on nouns followed by the vocative particle yā: "And whoever says to his lord, 'you ignoramus!' will be subject to fire".
Another archaism connected with the language of the Quran over against what became standard ClAr is the shape of the jussive and imperative forms of geminate (II=III) verbs.In the Hijaz, un-metathesized forms were used, whereas elsewhere metathesized forms were the norm (Rabin 1951: 161  "the sages and ones with understanding".
Finally, the locative adverb set differs between Old Hijazi hunālika "there", with a deictic li-, and Old Najdi hunāka without it.In standard ClAr, the Najdi form hunāka is more common, whereas in the Quran, the Hijazi form hunālika is more common.In this translation, the Hijazi/Quranic form is more common: "that which is there".
As noted above, this form also inflects for addressee agreement: "There is someone who is greater than your temple".
In most cases, other than the archaisms, as well as Quranic forms mentioned above, the morphology follows ClAr norms.This is notable especially when it comes to, e.g., forms of the imperative of I-ʾ verbs, which vary depending on whether they occur by themselves or with a prefixed fa-or wa- (Fischer 2002: 128) Relatedly, in cases where III-W verbs occur alongside III-Y by-forms in ClAr, the present translation attests similar variation, sometimes using the III-W form, while elsewhere III-Y ones.For example, the root ḥnw "to be compassionate" occurs in some corpora as ḥny.In this translation, the III-W root form is used, e.g.(25b) ‫ﺣ‬ ‫ﻨ‬ ‫ﻮ‬ ‫ﺕ‬ ‫ﻋ‬ ‫ﻠ‬ ‫ﻴ‬ ‫ﻚ‬ "I have treated you compassionately".On the other hand, the root ksw "to clothe" likewise attests a by-form ksy, which is the form used in this translation, e.g.‫ﻓ‬ ‫ﻜ‬ ‫ﺴ‬ ‫ﻴ‬ ‫ﺘ‬ ‫ﻤ‬ ‫ﻮ‬ ‫ﻥ‬ "And you clothed me".Forms that are consistently III-W in Quranic and ClAr remain III-W in this translation.In that way, it is unlike the modern dialects, which attest a near-complete merger of III-Y/W verbs > III-Y.
In addition to archaic and standard ClAr forms, a few innovative morphological forms, relative to Quranic and ClAr, occur in the translation.Here again, many of the innovative forms involve III-W/Y roots.Perhaps most notable are the spellings of form I participles from III-W/Y roots, which are spelled with two yāʾs, indicating an analogical restoration of the root consonant, as in pre-modern corpora and some modern dialects (Blau 1966-67: §101, 199-200) This feature seems like a good candidate for colloquialism, insofar as it likely reflects the way this sequence (i.e./iyī/) was realized in the speech of many authors.The fact that it is consistent in the text suggests that its usage was intentional.

Morpho-syntax
More than any other features, those of nominal case inflection and verbal mood inflection, and specifically the differences between ClAr and all non-ClAr corpora, have framed and shaped the discussion of Arabic linguistic history in general, and the distinction between the pre-and Islamic periods specifically (Blau 1966-67;1977;1999;2002;2006;Hopkins 1984).3Due to this framework, and the pervasive belief that Middle Arabic texts are, by definition, characterized by non-ClAr nominal and verbal inflection, Middle Arabic texts are rarely subjected to systematic analysis in which a percentage, or the whole, text is tagged and the different inflections quantified and plotted.I have done just such a quantification for these categories.As we will see, the different layers of vocalized texts are each significant in the case of this translation.The manifestations of nominal case represented in the consonantal skeleton are almost completely standard, whereas those represented by the vocalization markings vary.

Nominal case inflection
The present section discusses those features associated with nominal case inflection that are represented orthographically by consonants, which include: the accusative when written on orthographically eligible words with an alif; the five nouns in construct (on which a glide corresponding to triptotic case is written); duals and sound masculine plural forms.
The topic of nominal case inflection written only with vocalization markings will be discussed in each manuscript's respective vocalization section below.What are we to make of these non-ClAr usages of tanwīn alif?Are they pseudo-correction (hyper/hypocorrections), as Blau has argued (1966-67: 317-45)?It is, of course, possible.However, these contexts are precisely the same as those in which non-ClAr tanwīn alif occurs elsewhere in the corpus of vocalized Christian gospels from St Catherine's monastery in the Sinai Peninsula.Stokes (2023b) has argued that this pattern is connected with non-standard usages of tanwīn alif in unvocalized manuscripts, especially Sinai Arabic 72, 74, and 75.In the Christian gospels he studies, subjects of verbs, as well as, e.g., the use of the plural form of the noun with the tanwīn alif following numbers in the thousands, are each common.In a subset of manuscripts, the genitive is also occasionally written with the tanwīn alif as well.Stokes argues that this distribution is due to several factors.Phonetically a development from phonemic distinction before tanwīn (*un/*in/ *an) merged to a single, non-phonemic vowel (variously realized, thus written -Vn).At the same time, early on there was only tanwīn alif which was used to write tanwīn.The variation in attested distribution of tanwīn alif was based on the degree to which scribes decided to adhere to orthographic and grammatical norms associated with the writing tradition they learned.Early on, they basically did so, with few exceptions (cf.Sinai Arabic 72, 74, and 54), although Sinai Arabic 75 is a notable exception.The development of vocalizationsnamely fatḥatān, kasratān, and ḍammatānenabled nuance that was impossible using only the consonantal spellings.The non-ClAr usages attested in this translation align with those from elsewhere in the Christian tradition and, as we will see, the vocalization patterns of tanwīn in Or. 561 is identical to those attested in vocalized Christian manuscripts as well.

Five nouns in construct.
When in construct, the so-called "five nouns" (Arabic al-ʾasmāʾ al-ḫamsah) -ʾab "father", ʾaḫ "brother", ḥam "father-in-law", fam "mouth", and ḏū "possessor"are written with a consonant corresponding to its triptotic case.Each of the five nouns occurs in construct in this translation.In total, there are 118 occurrences of one of these nouns in construct and, of these, 115 inflect as in ClAr (98%).In each of the three non-ClAr inflections, the nominative form occurs in a syntactically accusative role: "And he sent away everyone except the young girl's father and her mother, and those who were wanting him".
Once again, the vast majority of cases inflect as in ClAr.The three exceptions follow the pattern found by Kootstra (2022) in a detailed study of the papyri of the first three Islamic centuries, where the nominative form was by far the most commonly used form in nonstandard roles.

Duals.
A total of 99 occurrences of nominal forms inflected for the dual occur in the portion of the translation studied here.Of those, 96 are inflected as in ClAr (97%).In each of the three instances of non-ClAr inflection, the oblique occurs in a syntactically nominative context: "And she was twelve years old, and they were in awe (of her healing)".
Interestingly, the three examples of non-ClAr inflection each involve the word "two" (iṯnayn) or "twelve" (iṯnay ʿašar).While it is tempting to see colloquial influence in these examples, another interpretation is possible.Elsewhere in the Christian corpus, the phrase "the twelve" (al-iṯnā ʿašar) is frequently either nominative or oblique, regardless of case.This is true even in manuscripts in which the dual is otherwise regularly inflected.For example, in Sinai Arabic 76, the dual is inflected as in ClAr, with a few exceptions.In the first 20% of that manuscript, 66 of 71 duals are inflected as in ClAr; of the five which were not, two were iṯnay ʿašar "(the) twelve", both of which were oblique in nominative contexts, e.g.(Sinai Arabic 76, 37) "These are the twelve".It is thus possible, perhaps even preferable, to see these examples in the present gospel translation in a sense as intentional Christianismsthe use of a very familiar phrase in a way that contravenes typical inflection for a sociolinguistic purpose.
2.3.1.4.Sound masculine plurals.In the portions of the translation included in this study I identified 778 instances of sound masculine plural.Of those, 766 (98%) inflect as in ClAr.Of the 12 instances of non-ClAr inflection, nine are cases of a nominative form in a non-nominative context, and three cases of an oblique form in a nominative context.Of the nine nominatives in non-nominative contexts, eight involve the word ḥawāriyy, "disciple": Here again, given that the overwhelming majority of cases are inflected correctly, it is difficult to take these examples as unintentional mistakes.While they might be, the clustering of these non-ClAr inflections among certain lexemes, a pattern in common with the dual inflection as well, suggests it is more likely that these Christianized forms were intentionally mixed in with the otherwise ClAr inflection.
2.3.1.5.Case inflection in consonantal textsummary.The various manifestations of morphosyntactic case inflection in the consonantal frameworkshared across the manuscriptsoccur as in the Quranic and ClAr systems.The few instances in which it does not align are so relatively few in number, and cluster in contexts and lexemes that are familiar from other corpora in such a way that it suggests intentional use of non-ClAr inflection, perhaps in order to index a particularly Christian Arabic, or otherwise more colloquial sound.

Modal inflection
This section discussions the features of verbal mood inflection that are detectable in the consonantal frame of the translation, which include distinctions between indicative/nonindicative in dual and plural forms, as well as jussive/non-jussive distinctions in forms from II-W/Y, III-W/Y, and II=III (i.e.C 2 =C 3 ) roots.The topic of modal inflection represented only by vocalization markings will be discussed in each manuscript's respective vocalization section below.The data collected for verbal mood is drawn from the first 20% of manuscript (Leiden Or. 561), which includes the entirety of the Gospel of Matthew.
2.3.2.1.Dual and pluralindicative vs. non-indicative.Due to the fact that in the dual and plural, there is no morphological distinction between subjective and jussive, thus the only meaningful categories are indicative, marked morphologically with a final V-ni (dual) / V-na (masculine plural) and a non-indicative ending with final V-ø.In total, 68 dual or plural verbs in indicative syntactic contexts occur in the first 20% of the Gospel of Matthew, of which 67 are inflected as in Quran and ClAr (99%  However, Fischer (2002: §453, p. 231) mentions that in "earlier stages of Arabic" indicatives were also used, depending on the context and desired verbal aspect.Further, in this example the verb ‫ﺗ‬ ‫ﺪ‬ ‫ﺭ‬ ‫ﻭ‬ ‫ﻥ‬ occurs in rhyme position in a section in which the rhyme is CūR/CīR.It is thus probable that the indicative was used in order to maintain the rhyme.Indeed, the translation itself is paraphrasing here, since there is no verb with this meaning in the Syriac or Greek versions of Matthew.
In terms of non-indicative contexts, 92 occur in this portion of Matthew, with all 92 (100%) inflecting as in Quranic and ClAr.
2.3.2.2.II/III-W/Y and II=IInon-jussive vs. jussive.Similar to the data from indicative/nonindicative categories, II-W/Y, III-W/Y and II=III verbs, which distinguish morphologically between non-jussive and jussive forms in the consonantal text, follow Quranic and ClAr.In the selected portion of the manuscript 50 instances of verbs from one of these three categories occur in jussive contexts, with all 50 (100%) inflecting as in Quranic and ClAr.Of the 60 verbs from these three categories that occur in non-jussive contexts, all 60 (100%) are likewise inflected in line with Quranic and ClAr.

Modal inflection in consonantal textsummary.
As with nominal case marking, verbal moods which are explicitly marked in the consonantal portion of the translation inflect as in ClAr.

Syntax
The syntax of this translation is remarkable in a number of respects, to the degree that it deservesand requiresa full treatment.Further, it requires a nuanced and detailed study of the Quran, which it clearly imitates throughout, in order to fully understand.Due to the lack of such a study, as well as space limitations here, I offer only a few notes about some of the more salient recurring aspects of its syntax, especially when they stand apart from textbook ClAr.
Numerous "He said to her, 'It is not good for the bread of the children (of Israel) to be taken and given to dogs'".
The use of mā in this context is associated with the Quran, especially when unmediated by the prefix bi-, i.e. when it takes its predicate in the accusative, which is called mā al-ḥiǧāziyyah, "The Hijazi mā", which is primarily attested in the Quran (Fischer 2002: §367.1, p. 191).The pseudo-verb laysa is also well-attested as a nominal negator: "John came and did not drink or eat".
Of these two options, mā, either with bi-or without, is slightly more common than laysa.
Out of 34 nominal sentences from the first 20% of the Gospel of Matthew in which one of the two negators occurs, laysa occurs 14 times (41%) while mā occurs 20 (59%).Finally, the negator ʾin occurs rarely by itself to negate a nominal sentence: "there is only evil among them" (lit."there is not among them except evil").
While the negator ʾin by itself is relatively rare, it is much more common following the negator mā, the combination of which is used to negate both verbal and nominal sentences: folia), followed by a discussion of the most significant aspects of the rhyme for an understanding of the nature of the language of the translation.For clarity and convenience, I have adopted the signs from Van Putten (2019: 8-9): Low short vowel a The following section includes a list of the schemata, along with the relevant folia in which each occurs.It should be noted that, as in the Quran, the translation often switches back and forth between rhymes, thus any given folio might be listed as attesting more than one rhyme scheme.
By far the most common is VR, many of which are sound masculine plural -ūn/īn.Of the VǦ variants, the most common is Cūḏ (4b)/Cūd (28b; 40a).The most common VH variants are Cūh (5b-6a;39b) and 24a).Among the best evidence that final short vowels are not realized in rhyming position in this translation, as in the Quran, is the rhyme of 3ms pronominal suffixes with, e.g., adjective ending in -ūh, as in (6a) when the indefinite adjective ‫ﻣ‬ ‫ﻨ‬ ‫ﻘ‬ ‫ﻮ‬ ‫ﻩ‬ "coward, good-for-nothing", rhymes with ‫ﺗ‬ ‫ﺠ‬ ‫ﻔ‬ ‫ﻮ‬ ‫ﻩ‬ "it (the people) will treat him harshly".

vC rhymes
While relatively less common the vH pattern is not uncommon.These patterns are always combinations of the genitive i followed by a 3ms pronominal suffix (Ci-h).The fact that these are realized as genitives is highly likely given that the sentences are written so as to ensure that each final noun is syntactically genitive: vR: 7b-8a vH: 10b-11a; 13b; 14a-14b; 16b; 26b-27b; 36a.

AC rhymes
AR : 4b;23b;34a;AǦ: 5a;7b;11a;13a;21b;25a;26a;35a;40a;40b;24b;26a;34a;35a;36a.The rhyme pattern AC is among the most commonly used in the translation, again reflecting its popularity in the Quran.Of note are the instances of AǦ rhymes, which reveal and corroborate aspects of the underlying realization of the rhyme.First, many AǦ rhymes involve nouns ending in *-āʾv, which rhymes with other nouns ending in AC.Second, many examples, such as (5a) where all forms end in Cāq, involve words which are in different syntactic contexts and would therefore have different final vowels.This corroborates evidence from elsewhere that word-final vowels were not realized, at least in rhyme position, in the language of the consonantal text.Of the AH rhymes, the most common is Cās and Cāt.
All instances of aH are feminine singular nouns ending in -ah.The rhyme, however, tends to involve recurring nominal patterns as well.For example, (5a-5b) involves feminine singular active participles of form I verbs, i.e.CāCilah forms; (10b) involves mostly Caʿīlah forms; (12a-12b) are nouns ending in the feminine singular genitilic -iyyah; (20a-20b) consist of mostly mufāʿalah and fāʿilah forms; and (25a-25b) are mostly II-ʾ feminine singular participles of form I verbs, i.e.Cāyilah forms.As elsewhere, case is not consistent, which strongly suggests that these are realized Cah.Corroborating this realization is the fact that of the aǦ patterns, Cat occurs but only with strings of 3fs perfect verbs, not tāʾ marbūṭah.
In most cases, nouns or verbs from one of these categories do not rhyme with each other.This is as in the Quran, where, with two exceptions, words ending in *ayv (alif maqṣūrah) and *awv (alif mamdūdah) do not rhyme with the other.However, while the dispreference for rhyming nouns and verbs from the different categories is predominant in the translation, there are more exceptions to it than in the Quran.Interestingly, in the few places where III-W (alif mamdūdah) forms are regularly rhymed, they are not interrupted by III-Y (alif maqṣūrah) forms, but are occasionally interrupted by indefinite accusative ones, as in (18b).
However, these interruptions are few in number compared with others, such as III-W and indefinite accusatives.Importantly, final -ī / -iyy is occasionally rhymed with III-Y forms (as in, e.g., 19b where al-muḍiyy "the departure" (written with shaddah over the yāʾ) is rhymed with preceding ʾaġnā "richer" and following lil-qurā "to the villages"), but not with III-W forms or indefinite accusatives.
The most common of the CA rhyme patterns involves rhyming indefinite accusative nouns.Here again, the vast majority of the time the pattern is observed; in a few cases, however, it is interrupted, either by III-Y forms (as in, e.g., 9a-9b where masrā "place of departure" interrupts the pattern), or by e.g.,36b, where half of the words in the section are III-W, the other half are indefinite accusative).
Thus, while final ā forms appear to rhyme with each other, the integrity of each pattern is generally maintained.III-W and indefinite accusatives rhyme with each other more frequently than with III-Y forms, and the latter rhyme with final -ī / -iyy forms, but neither of the others does.The likeliest interpretation of this is that the phonetic realization of each is /ā/, but that the scribe who composed the translation was intimately familiar with the Quranic rhyme scheme, which included the orthographic distinction between alif maqṣūrah and alif mamdūdah forms.

aCC rhymes
In one stretch, final nouns and verbs consistently end in a geminate aRR rhyme pattern: The nouns are in varying syntactic contexts, which confirms again the lack of realization of final short vowels in rhyme.

vCC rhymes
The least common rhyme pattern is one in which nouns ending in masculine singular gentilic ending -iyy rhyme: vǦǦ: 18a 2.6.8.Summary of rhyme patterns The rhyme patterns attested in this gospel translation replicate most of those attested in the Quran, with VR, AC, and CA (indefinite accusatives) among the most commonly attested.Final short vowels were not realized in rhyme position.The CA rhymes further demonstrate the intimate familiarity of the composing scribe with the details of the Quran by largely maintaining the distinction between III-Y forms and others realized with final -ā, such as III-W and indefinite accusative ones, despite the fact that the occasional overlap between each suggests that, as in standard ClAr, the phonetic realization of all of these forms was /ā/.

Linguistic remarks on vocalization of VAr. 17 and 18
The earliest manuscript witness to this gospel translation, VAr.18 (993 CE), is vocalized often, though not fully.The next oldest manuscript, VAr. 17 (1009 CE), is almost fully vocalized.In both, the vocalizations are, with a few exceptions, identical to standard ClAr patterns.For example, case marking, where noted, fully conforms to ClAr.The assimilation of third person pronominal suffixes to preceding Ci, Cī, and Cay is also standard ClAr, with singular, dual and plural forms assimilating in these contexts.Other vocalizations are also consistent with ClAr, such as final *-āʾv ones, whereas, as we will see, this is not the case with Or. 561: (VAr.18, 13v) ‫ﻟ‬ ‫ﻘ‬ ‫ﺪ‬ ‫ﺟ‬ ‫ﺂ‬ ‫ﺀ‬ َ "He has come".
used with participles of verbal forms II-X; vocalizations of tanwīn (on which, see below); suffixed pronouns, especially the 3ms; and certain lexemes.Vocalization further occurs somewhat randomly, which nevertheless sheds some light on the phonetic and phonological layers of the underlying variety.I turn now to a brief discussion of those aspects of the vocalization that are both non-ClAr and which recur.At various places in Or. 561, a later hand makes various edits of the underlying text and vocalization.These edits are always clearly distinct from the original text because the original ink is light brown in colour, whereas the later editing hand is written in black ink.It is unclear how soon after copying the edits were made.The edits are of several kinds.First, where the original text has become unclear due to damage or smudging, the editor fills in what he thinks belongs, as in Figure 1. 5econd, in numerous places, the editing hand adds vocalizations, especially related to nominal case inflection.These case-related changes often involve marking out the non-ClAr ones and replacing them with the standard variants, as in Figure 2.
Third, certain words are crossed out and replaced with another word, as in Figure 3.
The grammatical edits and additions to the text are always in the direction of making the text closer to standard ClAr, and it is likely that the lexical replacements were geared towards replacing words felt either too unfamiliar or colloquial with words more commonly used and associated with ClAr as perceived by the editor.Other examples of this lexical replacement include: (4b) ‫ﺩ‬ ‫ﻭ‬ ‫ﺏ‬ "honey" was replaced with ‫ﻋ‬ ‫ﺴ‬ ‫ﻞ‬ (5a) ‫ﺣ‬ ‫ﻮ‬ ‫ﺏ‬ "fault" was replaced with ‫ﺫ‬ ‫ﻧ‬ ‫ﺐ‬ .
Interestingly, the later editor did not replace overtly Islamic words with others felt to be less Islamic.

Orthography and phonology
While vocalization of nominal and verbal forms is inconsistent and sparse compared with VAr. 17   6.
In a few of the above examples, such as ʿumādan and ġaḍub, it is possible that the pharyngeal (in the former) or the emphatic (in the latter) also influenced the vowel, thereby representing a backing effect on the vowel.This is attested elsewhere, as in the examples listed in Table 7.
In a final example attested in the portion of the manuscript included in this study, an example of ḍammah in a root in which i is expected, despite the absence of a bilabial or emphatic: "And we increased in insight and discernment".
The context in which vocalizations are most consistently used throughout the manuscript are those in which etymological glottal stop occurs.Of the various contexts, the three most relevant from the perspective of the vocalizations in Or. 561 are glottal stop in traditions, both linguistic and scribal.As argued in Stokes (2023b), the variety behind the vocalizations across the Christian manuscripts can be explained as above, namely as evidence for a variety in which short vowels before tanwīn had merged.It seems likely that at the point at which this merger took place, some contexts had ceased to be marked by case vowels + tanwīn, with accusative roles, especially adverbs, as well as subjects of certain verbs, lexemes of highly specific or individuated nouns, and nominal predicates were more salientand thus more consistently marked explicitlythan others.When the merger occurred, these contexts were still marked, but with a morpheme, composed of the single vowel and tanwīn (-Vn).When scribes attempted to write such a variety, they had to make use of the orthographic tools at their disposal, namely the three vocalization signs.Some made use of just one, others alternated fatḥatān and kasratān, while still others used all three.Crucially, most scribes attempted both to represent the underlying variety and adhere to the conventions of the writing tradition in which they had been trained.They accomplished this to varying degrees and in different ways.
This, however, is not the only pattern documented by the early grammarians, nor did they canonize it as the only acceptable one.Further, there are alternative patterns attested among both the canonical Quranic recitation traditions, as well as others found in vocalized Quranic manuscripts (Van Putten and Sidky forthcoming).
As shown in Table 8, in Or. 561 a very consistent non-ClAr pattern of harmonization is attested in which the 3ms suffix harmonizes to preceding Ci, Cī, and Cay (as in ClAr), but the dual (and presumably the plural) forms do not.
Significantly, the 3ms pronominal suffix is written -hi (length is not indicated) not only after the preposition bi-, but also nouns in genitive contexts.While case endings are not written explicitly, this alternation all but guarantees its presence.The dual pronominal suffixes do not assimilate after either Ci or Cī.The plural pronominal suffixes are not vocalized following Ci, Cī, or Cay in Or. 561; however, it seems a priori likely that they would pattern with the dual, which is attested in multiple places with a ḍammah, indicating a lack of harmonization.It is impossible to know whether nominative u or accusative a were realized.
Whether this harmonization pattern was ever a living feature, or rather was one of numerous variants which were aesthetically motivated, is uncertain.The complicating factor is the lack of attested vocalization of 3mp and 3fp forms.It is possible, for example, that the bilabial nasal /m/ in, e.g., the dual and 3mp suffixes, led to the rounded vowel /u/, whereas those without it (3ms, 3fp) assimilated.Regardless of whether or not it was once a living feature, it is consistently used throughout Or.561, and was clearly prestigious.Similar assimilation patterns are attested in the Quranic tradition (Van Putten and Sidky forthcoming) as well as in other vocalized Christian manuscripts (Stokes 2023a).4.2.7.Overview of the case system in Or. 561 The evidence for morpho-syntactic case inflection in Or. 561 suggests a system in which final short case vowels were realized when non-word final, at least before pronominal suffixes, as evidenced by the assimilation of the 3ms and 3du pronouns when suffixed to nouns in the genitive.Formerly indefinite triptotes are marked with tanwīn, which was probably realized with a single vowel, reconstructed here as /in/ or /ən/.Despite the loss of phonemic contrast between short vowels before tanwīn, the morpheme continued to mark a set of salient syntactic roles, including adverbs, subjects of certain verbs and clauses expressing existence or possession, as well as lexemes expressing groups or single members of a species.

General discussion
The three manuscript witnesses to the gospel translation under examination in this paper attest several linguistic layers, each of which is relevant for the study of Middle Arabic.The consonantal layer reflects a mixing of various forms, especially of archaisms, many of which were associated with the language of the Quran, while others were archaic poetic forms.While many of these archaisms are mentioned, and accepted, by the Arabic language grammarians, they are often not typical in standard ClAr.Also attested in the consonantal layer of the translation are a minority of forms which are common in Middle Arabic, with specific manifestationssuch as the non-ClAr use of terms for "the twelve (disciples)"that are salient in other Christian gospel translations as well.The result is a text that is highly Quranic, and generally sophisticated and archaic, with Christian-specific features sprinkled in here and there.
The vocalizations in the three manuscripts also communicate a great deal.While VAr. 18 and VAr. 17 both generally follow standard ClAr, VAr. 17 especially apparently mixes in still other features, possibly associated with a Hijazi and/or Quranic style, reflected especially in the use of fāyil as the form I active participle pattern, rather than the ClAr fāʾil one, as well as the apparent non-ClAr use of hamzat al-qaṭʿ for the definite article, imperatives, and derived stem perfects.
The vocalization attested in Or. 561 is distinctly non-ClAr in most cases, reflecting a deliberate Christianization of the phonology of the text.While we cannot be certain whether the scribe who copied Or. 561 relied on either VAr.18 or VAr.17, it is very possible that he did.In that case, the scribe had a copy not only of the consonantal text, but also vocalizations, which he then deliberately changed.The quality of the copying of the consonantal portion of the text, which is neater and more skilfully done in some respects than in either VAr.18 or VAr.17, rules out incompetence or carelessness.Further, the commonality of many of the features, such as the patterns of tanwīn and writing of final -āʾ, also suggests that rather than reflecting a very localized tradition, the scribe intentionally targeted a variety of Arabic associated to some degree with Christians, or at least commonly used by Christians to recite the gospels.
Returning to the topics of the linguistic nature of the features in Middle Arabic texts, as well as about the spectrum along which Middle Arabic texts are situated, the evidence in the translationdrawn from each of the three witnessesargues against many of the standard assumptions of the field.First, the assumption that if authors don't intentionally target a particular Middle Arabic sub-variety, they would target standard ClAr, is clearly not safe in every instance and must be demonstrated by close reading of the text, rather than assumed.In this translation, both the consonantal text and vocalization layers the use of features that are not standard ClAr.In some cases, the target was Quranic Arabic, even when it disagreed with standard ClAr.Elsewhere, archaic forms were used rather than more commonly prescribed ClAr ones.In still other cases, such as the fāyil participles, it seems as though the target is Old Hijazi, perhaps an early prestigious variety in the Umayyad Levant (Al-Jallad 2020).Finally, features likely associated with the vernacular of the composer or copyist were identified, as for example the occasional phonemic spellings of perfects, likely indicating syncope of the initial short vowel in an open syllable, as well as the shift of *ẓ > ḍ.Interestingly, the phonemic spellings almost always occur in direct speech, which possibly reflects a deliberate choice by the scribe to use a more vernacular register in those contexts.
Second, evidence from this gospel translation suggests that in at least some cases, Middle Arabic features which do not conform to ClAr norms, nor occur in modern dialects, are nevertheless most plausibly derived from once-living features, and can be included in discussions of the historical development of Arabic.I have argued here that the distribution of the tanwīn vocalizations attested in Or. 561 are virtually identical in many ways to those attested in other Christian gospels, which reflect a historical change, namely the merging of short vowels before nunation to a single vowel, likely realized /in/ or /ən/, which nevertheless retained for a while at least some salient syntactic functions.The harmonization of the 3ms pronominal suffix to preceding i, ī, and ay, but the lack of harmonization of the dual (and, presumably, plural) suffixes in the same contexts is another example.Both of these features are intentionally used, here and elsewhere (for tanwīn in Christian Arabic, see Stokes 2023b; for the pronominal harmonization, see Stokes 2021 for the same distribution in Judaeo-Arabic, and 2023a for a discussion of this and other patterns in Christian Arabic gospel translations).
Therefore, many of these Christian Middle Arabic features likely originated in living features, which became prestigious variants, and thus constituted a regular part of the H register.While it is questionable, even doubtful, that the feature of non-ClAr tanwīn was a living feature of the dialects of every scribe who employed it, it is also not clear when it ceased to be living, and in any event was clearly prestigious for centuries.We thus need to broaden the nature of the H register, at least for a subset of Middle Arabic texts, beyond textbook ClAr to include linguistic variation associated with the Quran and its recitation, as well as other variants which were prestigious in the centuries before a single set of prescriptive norms associated with ClAr became established.There were clearly multiple features and variants which held prestige, and were mixed to different effects, depending on the audience, genre, and likely other variables as well.The proposal here thus differs from previous ones in several ways, namely I argue regarding the phenomena attested in this gospel translation that: 1) non-standard features which also do not occur in modern dialects can reflect (at least originally) living features; 2) these features became prestigious; and 3) they were included in a nexus of prestigious variants which could, and among Christians often did, include ones not only from the ClAr tradition, but also Quranic recitation in its broadest sense, even preserving archaic features otherwise forgotten or proscribed over time among Muslims.8

Conclusion
This translation of the gospels into Arabic is so far unique in the Christian Arabic corpus insofar as it linguistically imitates the Quran in a multitude of ways, especially in the lexicon and rhyme scheme.The composing scribe mixes archaic forms, some of which are peculiar to the Quran or Old Hijazi, while others are attested primarily in poetry.Each of the three manuscript witnesses to the translation is vocalized to some degree.The earliest two manuscripts, VAr. 17 and 18, are vocalized in basically standard ClAr; however, VAr. 17 attests several phonological features that are not normal in ClAr, but may indeed be archaic.The third and latest manuscript, Or. 561, differs in many respects from ClAr on the one hand, and is nearly identical in those cases to features found in other Christian Middle Arabic translations of the gospels.There is therefore a movement away from a mix of ClAr, Quranic and other archaic features in the consonantal base of the translation, vocalized in basically ClAr early on, but Christianized in terms of phonology and morpho-syntax by the fifteenth century CE.It was argued that the mix of features demonstrates that, at least for some scribes, the high (H) register was much more varied and diverse than what has become normative ClAr, extending not only to archaic and Quranic forms, but also prestigious forms that were never a part of the ClAr tradition.Finally, several features, such as the non-ClAr distribution of tanwīn vocalizations in Or. 561 that likewise differs from dialectal tanwīn in modern dialects, were argued to have originated as living features which subsequently became prestigious variants in their own right.In addition to the unique mix of features, as well as the sophisticated imitation of Quranic rhyme, this translationand the manuscript witnesses to itrequire broadening of the spectrum along which Middle Arabic texts are plotted in the medieval period, as well as more nuanced typological categories for the non-ClAr features attested therein.
and 18, there are nevertheless indications of non-ClAr phonetic and phonological features based on what vocalizations are present.The first category of such features

Table 1 .
Assimilation of verb forms V and VI

Table 2 .
Assimilation of min to definite article in following noun

Table 4 .
Spellings of *ʾa ̄/ *a ̄ʾvII, 5, 77 f.; IV: 182, 411; al-Farrāʾ 2014: 11-12, 22, 94).The Hijazi forms are characteristic of the Quran, as well as in standard ClAr, whereas the Najdi forms are mostly limited to poetry.Proximal forms from both sets are found throughout the manuscript: Another possible archaism is the use of tifʿāl / tafʿāl nouns as maṣdars of form II verbs instead of the more common tafʿīl patterns (on which, see Wright II: 115-6): the retained wāw, would suggest /tawṯāq/ instead of /tiwṯāq/ and /tawfāh/ instead of /tiwfāh/.While these forms are far less common in Classical Arabic than tafʿīl, they are attested in various modern dialects, and indeed are the more common form in some North African dialects, such as Algerian (e.g.Dhina 1938:  327)and Moroccan Some archaisms are connected more explicitly with the language of the Quran.One is the use of shortened jussive forms of kān, spelled ‫ﻳ‬ ‫ﻚ‬ (ClAr yaku; Old Hijazi yak; Van Putten and Stokes 2018: 169): f.): Hijazi urdud "return" vs. non-Hijazi rudd(a/u/i).In this gospel translation, the Hijazi forms, which are the norm in the Quran, are regularly used: These un-metathesized forms are the norm for the manuscript.The determinative pronoun set, indicating "possessor(s) of X", are frequently used in the text:In the plural, while both ḏawū and ʾulū forms are attested, the latter are more common: . In the gospel translation, these forms occur with the same distribution: Also following the ClAr pattern are verbs such as ʾamara "he ordered", which when by itself is typically realized mur "order (msg)!", but when prefixed with a fa-or wa-is spelled with the alif:In a few cases, however, there is some variation in the spelling of what are apparently the same form.For example, the imperative of ʾatā, "he came", is spelled both . Examples from the translation include: aspects of the patterns of negation either pattern with the Quran against what later became normative ClAr, or attest interesting variations worth noting here.The best-known feature which patterns with the Quran, and Old Hijazi, against what became normative in ClAr is the use of mā to negate nominal sentences, either taking the predicate in the accusative case or introduced by the prefix bi, both of which are common: