1.0. Introduction

1The development of Semitic vocalisation systems spans a massive gulf of time, beginning with the first use of matres lectionis letters and continuing to the standardisation of the modern Arabic and Hebrew vowel pointing systems. But the portions most commonly implied by the phrase ‘vocalisation system’—that is, the vowel signs themselves—were invented in the multicultural environment of the early medieval Middle East. Between the seventh and eleventh centuries, historically Aramaic-speaking Jews and Christians faced the challenge of preserving their biblical recitation traditions in the face of the growing dominance of the Arabic language. In the same period, Arab Muslims feared the corruption of the Qurʾanic recitation tradition as a result of contact with non-native Arabic speakers.

2Adherents to all three religions took steps to protect their languages. Syriac Christians first created a system of diacritic dots to record vowels in the Bible, and soon after, both the Jewish Masoretes and Arab grammarians implemented dot-based systems for marking vowels. Scholars have debated potential relationships between these dot systems for over a century,1 often without regard to the chronology of their sources (see below, §3.1).2 And indeed, the three vocalisation traditions are linked to such a degree that it is difficult to explain the history of one without putting it in context with the other two. The connections between them, however, are not necessarily graphic, and instead relate to phonological theories and terms that medieval grammarians developed to describe their vowel systems.

3This study thus aims to compare the phonological traditions of Syriac, Arabic, and Hebrew to demonstrate how they influenced each other over time. That is to say, it will look at the ways medieval linguists described their own languages, and compare the concepts that they used to discuss vowel phonology. In what follows, §2.0 will establish shared features in the Syriac and Hebrew vocalisation traditions prior to the spread of Arabic as the dominant language in the Middle East. §3.0 will examine the emergence of eighth-century Arabic phonetic terminology and its relationship with Syriac. Then §4.0 will explore some ways in which tenth- and eleventh-century Syriac and Hebrew grammarians blended Arabic phonological concepts into their own linguistic traditions.

2.0. The Hebrew-Syriac Connection

2.1. Early Syriac Relative Vowel Phonology

4Some of the earliest descriptions of a Semitic vocalisation system come from Jacob of Edessa (d. 708), a Syriac Christian bishop whose grammatical writings reflect a combination of Greek influence and native Syriac concepts. Three works in particular are crucial for understanding the history of Syriac phonology: his grammatical tract ‘On Persons and Tenses, ’ his ‘Letter on Orthography’ to George of Sarug (Phillips 1869), and his grammar, the Turrɔṣ Mamllɔ Nahrɔyɔ ‘The Correction of Mesopotamian Speech’ (Wright 1871), of which only six folios survive.

5Jacob addresses vowel phonology in the introduction of ‘On Persons and Tenses, ’ writing:

Then the tenses are three, past, present, and future, and sounds are thick and thin. Every saying, that is, [every] form, when it is thick or wide with sound, then it takes a point above. But when it is narrow or thin, then below. If it is intermediate, between narrow and thick, and there are two other [words] written the same as it, then it takes two points, one above and one below. This is called ‘restraining’. (Phillips 1869, )

6This passage shows that Jacob understood vowel phonology according to a relative classification system. Within this system, every word has a particular set of vowels that is comparatively different from the vowels of its homographs. These vowels are not absolutely defined, but rather for a given pair of homographs, Jacob would describe one as more ʿḇe ‘thick’ or pṯe ‘wide’, while the other would be more nqeḏ ‘pure’ or qaṭṭin ‘narrow’. Based on examples later in the text (Phillips, ), vowels most often associated with the ‘dot above’—i.e., relatively ‘thick’ vowels—were /ɔ/, /o/, and /a/. Meanwhile, those marked with a ‘dot below’—the relatively ‘thin’ vowels—were usually /u/, /i/, /e/, and /ɛ/. However, these attributions were not absolute. It seems that while Jacob interpreted vowel phonemes in terms of relative bulk or openness, he did not use any terms or graphemes to indicate particular vowels on a one-to-one basis. A vowel that was considered ‘wide’ in the context of one homograph could be called ‘narrow’ when compared to another.

7Jacob complicates this two-way relative system by the inclusion of meṣʿɔyɔ ‘intermediate’ vowels, which can only be identified in words that have at least two homographs. Such vowels are represented by ‘two points, one above and one below, ’ which Jacob refers to as mpaggdɔnɔ ‘restraining, bridling’. This term seems to describe only the physical two-dot grapheme, while the vowel phoneme itself is called meṣʿɔyɔ. This term almost always indicates the vowel /a/, but more importantly, it has no inherent descriptive qualities, and any meṣʿɔyɔ phoneme could be called pṯe or qaṭṭin in another context. It seems then that Jacob added the meṣʿɔyɔ term to his vowel phonology to align it with his understanding of consonants, which, in his grammar, he categorises as ʿabyɔṯɔ ‘thick’, meṣʿɔyɔṯɔ ‘intermediate’, and neqdɔṯɔ ‘thin, clear’ (Wright 1871, ). E.J. Revell (1972, 367) suggests that Jacob adapted these terms from Greek descriptors that meant, respectively, ‘rough’, ‘intermediate’, and ‘smooth’ with regard to voicing, modifying them to suit the Syriac language (see also Knudsen 2015, 77). As such, meṣʿɔyɔ was likely an addition to pre-existing Syriac vowel phonology—one based solely on relative degrees of bulk or openness—in order to fit Jacob’s wider Greek-inspired system.

8From this information, we can assume that Jacob of Edessa built on an older phonological tradition that used terms like ʿḇe, pṯe, qaṭṭin, and nqeḏ to describe vowels relative to each other, but not to name them. Since ʿḇe and nqeḏ were probably calques from Greek, examining pṯe and qaṭṭin may provide further insight into how early Syriac phonologists perceived vowel quality. These latter two terms appear to be descriptions of the lips while articulating vowels. For example, the mouth is relatively wide (pṯe) when one says /a/, whereas it is narrow (qaṭṭin) when saying /e/. Similarly, the lips open wider for /e/ and /o/ than they do for /i/ and /u/. Curiously, similar descriptions occur in the earliest work of the Hebrew Masoretes.

2.2. Early Masoretic Relative Vowel Phonology

9In an article on the etymology of Hebrew vowel names, Richard Steiner (2005, 379–80) argues that terms based on the roots ptḥ ‘opening’ and qmṣ ‘closing’ predate all other Hebrew vowel names, and that in their original form they distinguished minimal pairs of vowels according to lip movement. His main evidence for the relative antiquity of these two vowel terms is their appearance in the Masora magna and parva, as well as the fact that modern pataḥ and qamaṣ originated as the Aramaic active participles pɔtaḥ and qɔmeṣ (Steiner 2005, 374; 377–78; see also Khan 2000, 24). Meanwhile, the remaining names for Hebrew vowels are not in the Masora, and are contrived from later Hebraisms. Both of these features indicate that terms from ptḥ and qmṣ emerged in the eighth century, perhaps earlier, and Aron Dotan (1974) has identified rare usages of these roots to distinguish vowel pairs other than /a/ and /ɔ/ (see also Steiner 2005, 379). Both Steiner and Dotan thus conclude that the early Masoretes developed a relative system for describing vowels, as the latter writes:

It would appear that this use of the terms ץמק and חתפ occurred during a most ancient period, a time when these terms were not as yet serving to denote definite vowels. The vestiges of this use, both of the terms ערלמ , ליעלמ and the terms חתפ , ץמק indicate that in the period which preceded the invention of the vowel signs such a method of relative notation of vowels was current. It was therefore necessary to indicate the vowels which distinguish between homographs. (Dotan 1974, 32)

10This relative usage disappeared by the tenth century at the latest, when Hebrew vowels were reclassified according to backness and airflow, as will be shown below. Syriac underwent a similar development around the turn of the eighth century, with phonetic backness becoming associated with ‘height’.

2.3. The Pre-Arabic Relative Context

11The lack of absolute vowel notation prior to the eighth century gave rise to homograph lists in Syriac and Hebrew. In the Hebrew tradition, these lists divided homographic pairs according to stress, separating them with the Aramaic terms milleʿel ‘above’ and milleraʿ ‘below’. One of the first scholars to examine these concepts was Heinrich Graetz, who attempted to connect the Tiberian Masoretic tradition to Syriac on the basis of diacritic dots. He studied the homograph lists in Okhla we-Okhla and found that, in addition to their normal meanings related to stress, the terms milleʿel and milleraʿ were sometimes used to distinguish Hebrew homographic pairs that differed by one vowel (Dotan 2007, 622– 23). By analogy with the Syriac diacritic ‘dot above’ and ‘dot below’, Graetz identified this usage as part of a relative vocalisation system. Both Steiner and Dotan also see these terms as evidence of the earlier two-way, relative perception of vowels (Steiner 2005, 379; Dotan 1974). However, Graetz took an additional step, hypothesising that milleʿel and milleraʿ referred to diacritic dots that, just as in Syriac, were placed above or below a Hebrew word to indicate the relative quality of its vowels (Dotan 2007, 622–23). The problem with this idea is that a diacritic dot has been attested only once in the context of Hebrew milleʿel and milleraʿ lists, and in that manuscript the dot indicates stress, not vowel quality (Steiner 2005, 379; Dotan 2007, 623). Graetz’s theory also requires that the terms themselves were borrowed from Syriac, and that they persisted after the apparent ‘disappearance’ of the hypothesised Hebrew diacritic dots.

12Refuting Graetz, Dotan (2007, 623) insists that such terms ‘do not exist and never did exist in the supposed source language, Syriac, ’ but this may not be true. Returning to the afore-mentioned passage from ‘On Persons and Tenses, ’ Jacob of Edessa says:

Every saying, that is, [every] form, when it is thick or wide with sound, then it takes a point above. But when it is small or thin, then below.

13A word with thick vocalisation takes a dot men lʿal ‘above’, while its thinner homograph is men ltaḥt ‘below’. Jacob’s meaning here is clear, but these two prepositional phrases do not follow the typical Syriac practice of indicating above and below. Normally, one would expect the respective phrases lʿal men (h) or ltaḥt men (h) in this situation, and indeed that is what Jacob writes when he describes locations of diacritic dots in his ‘Letter on Orthography’ (Phillips (1869, lns. 13–14; lns. 2–3; for an example unrelated to diacritic dots, see ln. 16: the art of writing ‘is lʿal men all arts’). Jacob does not use men lʿal and men ltaḥt to discuss regular diacritic dots, but rather applies these phrases only to locate dots that are specifically related to vowels. That is, men lʿal and men ltaḥt are somehow unique phrases that have additional meaning related to vowel phonology. Furthermore, as is typical of Syriac, the second half of the above sentence does not repeat the word nuqzɔ ‘dot’, such that in a vacuum the line could be read, ‘Then what is small or thin is below.’ The phrase men ltaḥt thus appears to have an abstracted categorical usage, classifying the words it describes according to some conceptual ‘low’ quality. In the fourth chapter of ‘On Persons and Tenses, ’ i.e., ‘Sounds, ’ Jacob writes:

Above are, for example, shmayyɔnɔ, ʿɔḇdɔ, ʿḇɔḏɔ, ʿabbaḏɔ (?),3 malkɔ, and ṭɔḇɔ. Then below are shamminɔ, ʿaḇdɔ, and ṭeḇɔ. (Phillips 1869, ܙܝ )

14While his intention is undeniably to describe dot locations, Jacob does not use the word nuqzɔ with these instances of men lʿal and men ltaḥt. The prepositional phrases simply categorise the example words as ‘above’ and ‘below, ’ according to the two types of vowels. That is, the phrases serve as phonological terms, rather than descriptors of dot position. This development, which seems to have been on the verge of completion during Jacob’s life, may be the origin of the later Syriac phonological system that associated phonetic backness with height (Revell 1975, 181).

15At the end of the manuscript, the copyist inserts a brief passage that had been omitted from the introduction:

Then, again, as for the sounds which indicate ʾenɔ ʾemreṯ and ʾeʿbeḏ, and all the rest that are like them, and moreover, regarding ʾeddun, they have points below. Then those [sounds] which indicate ʾɔmar and ʾɔḵel, and the rest, they are above. (Phillips 1869, 32, fn. i)

16Phillips suspects that these instances of men lʿal and men ltaḥt should be reversed, in order to conform to the more common usage of diacritical dots that distinguish between first- and third-person verbs.4 However, the passage does not begin ‘as for the dots which indicate, ’ but rather ‘as for the sounds which indicate, ’ and, as such, the text should be interpreted in terms of the phonological system that Jacob has already explained. Through this lens, the syntactic placement of men lʿal and men ltaḥt makes sense: the first-person ʾemret (G perfect) and ʾeʿbeḏ (G imperfect) have ‘thinner’ vowels than their respective third-person homographs, ʾemrat (G perfect 3fs) and ʾaʿbeḏ (C perfect 3ms), so they ought to take a dot below. It seems that the copyist put dots above the first-person verbs according to the standard diacritic practice, as Phillips expected, even though, in this case, the dots that Jacob describes as men lʿal and men ltaḥt were meant to convey relative vowel quality. The following examples—the participles ʾɔmar and ʾɔḵel—are thus correctly classed as men lʿal, as the dot above distinguishes them from their respective homographs in the perfect, ʾemar and ʾeḵal. So again, in a case related specifically to vowel phonology, Jacob uses the uncommon constructions men lʿal and men ltaḥt in such a way that they appear to be phonological terms, conceptually divorced from the dots they once described.

17Recalling Dotan’s stance on the potential relationship between Syriac and the terms milleʿel and milleraʿ, he (2007, 623) asserted that such terms “do not exist and never did exist in the supposed source language, Syriac.” But Jacob of Edessa instructs that words with thick vowels take a dot men lʿal, while those with thin vowels take a dot men ltaḥt. Those particular phrases flirt with a theoretical usage, almost describing the phonology of words affected by dots, rather than the dots themselves. While still not explicit vocalisation terms, such descriptors mirror milleʿel and milleraʿ, at least on a conceptual level. It is possible that the Syriac phrases collapsed over time, with the nūn in men lʿal eliding to produce a geminated lamed in something like milleʿel. Similarly, men ltaḥt can be calqued as men lraʿ, which could collapse to milleraʿ.5 Simultaneous with this etymological shift, the Syriac terms became dissociated from the physical dots, becoming adjectives expressing the relative qualities of vowels. If this is the case, then the lack of attested evidence for the Hebrew dots hypothesised in Graetz’s theory is not irregular, but rather expected. That is, by the time the phrases men lʿal and men ltaḥt had a chance to become phonological terms in Syriac (c. 700–750), they had already lost their meaning related to dots. Consequently, the Masoretes could have adopted them without copying the Syriac diacritics. I know of no primary source that explicitly describes such a development, but Dotan is perhaps too quick to dismiss a Syriac connection.

18These similarities between the Syriac and Hebrew linguistic traditions suggest that the early Masoretes understood vowel phonology in much the same way as their Syriac Christian contemporaries. Both traditions qualified vowel phonemes on a hierarchy according to the relative openness of the mouth during articulation. For the Syrians, this meant that vowels could be pṯe ‘wide’ or qaṭṭin ‘narrow’ when compared to other vowels. Some early Masoretes also applied this principle, and described those same vowels as pɔtaḥ ‘opening’ or qɔmeṣ ‘closing’. Moreover, there is even evidence that both traditions used Aramaic terms, i.e., milleʿel ‘above’ and milleraʿ ‘below’, in some form to delineate between homographs with different vowels, suggesting that the terms may have entered into masoretic usage as Syriac loans. Over time these terms likely contributed to the association of height with phonetic backness in the Syriac and masoretic traditions. This concept eventually appeared in Saadya Gaon’s Kutub al-Lugha (Skoss 1952; Dotan 1997), which will be discussed below.

3.0. The Development of Arabic Vowel Terminology

3.1. The Chronology of Arabic Vowel Names and Their Relationship to Syriac

19The Arabic grammatical tradition emerged in this world of two-way relative descriptions, and early Arabic sources on vowel phonology reflect that context. They do not, however, indicate a wholesale borrowing of Syriac phonetic terms that became the Arabic vowel names (Versteegh 1993, 28–31; Talmon 2003, 289– 91).

20C.H.M. Versteegh has identified a Qurʾanic tafsīr by Muḥammad al-Sāʾib al-Kalbī (d. 763) as the earliest source for Arabic vowel names. In it al-Kalbī lists variant readings of the Qurʾan using unpointed Arabic, so he describes alternative vowels using words, rather than signs. In the sixty-eight variants that he records, al-Kalbī uses kasr, jarr, and khafḍ to describe i-vowels, fatḥ and naṣb for a-vowels, and ḍamm and rafʿ for u-vowels (Versteegh 1993, 125). Versteegh (1993, 126) notes that at this stage there was no consistent distinction between what are now considered vowel names (kasr, fatḥ, ḍamm) and declensional terms (jarr, khafḍ, naṣb, rafʿ), and concludes that “the later terms for the case endings were once part of a system to indicate vowels.” He takes these seven terms and compares them to the list of Syriac vowel names published by Adalbert Merx in 1889 (Versteegh 1993, 29–31), which Merx (1889, 50) collected based on what Gregory bar Hebraeus (d. 1286) wrote about what he claimed were the names of vowels used by Jacob of Edessa (d. 708). To say that this chain of transmission is tenuous would be generous.

21Versteegh suggests that five vowel names in Bar Hebraeus’ grammar—pṯɔḥɔ, zqɔp̄ɔ, rḇɔṣɔ, ḥḇɔṣɔ, ʿṣɔṣɔ—are the source of the Arabic terms fatḥ, naṣb, khafḍ, kasr, and ḍamm. While he is correct in pointing out parallels between the two sets of terms, incorporation of the Syriac sources from before the thirteenth century reveals a more complicated picture. The most obvious connection is the pair of pṯɔḥɔ and fatḥ, cognates that mean ‘opening’. Similarly, ʿṣɔṣɔ and ḍamm, while not cognates, both mean ‘contracting’, and ḥḇɔṣɔ and kasr can both (loosely) mean ‘pressure’ (Versteegh 1993, 30). The problem, then, is a chronological one. As we have already seen, Jacob of Edessa did not name any Syriac vowels, and only thought of them as relatively open or closed. There is no evidence that he had a word like ḥḇɔṣɔ or kasr to indicate a third type of vowel, and in fact when Jacob of Edessa uses the root ḥbṣ in his ‘Letter on Orthography’, it indicates an orthographic contraction rather than anything phonological (Phillips 1869, ln. 17). The earliest example of the use of the root ḥbṣ in relation to a vowel seems to come from Elias of Tirhan’s (d. 1049) grammar (Baethgen 1880, see below for the use of ḥbṣ for both /u/ and /i/), and it is not clear that either he or Elias of Ṣoba (d. 1049) used ʿṣɔṣɔ as a vowel term at all. As such, while the dual concepts of vowel ‘opening’ (and thus ptḥ) and ‘contracting’ could have entered Arabic from Syriac in the eighth century, the terms ḥḇɔṣɔ and ʿṣɔṣɔ are much later inventions, possibly calqued from kasr and ḍamm into Syriac. In any case, they cannot be the direct source of the Arabic vowel names. On the other hand, it would not be surprising if some of the earliest vowel descriptions in the Syriac, Arabic, and Hebrew traditions were all independently derived based on mouth movement. For example, pṯe ‘wide’ and qaṭṭin ‘narrow’ in Syriac, fatḥ ‘opening’ and ḍamm ‘contracting’ in Arabic, and ptḥ ‘opening’ and qmṣ ‘closing’ in Hebrew.

22Versteegh’s treatment of zqɔp̄ɔ and rḇɔṣɔ is more problematic. He attempts to explain their relationship to Arabic, writing:

The other phonetic concept that can be reconstructed from the terminology is that of the progressive lowering (of the tongue?) towards the front of the mouth. According to Revell (1975: 181), sounds at the back of the mouth are regarded by the Syriac grammarians as high, those at the front as low. Thus, the grammarians used the terms zqāphā ‘raising’ and rbāṣā ‘lowering’ for ā and ē, respectively.6 These vowels were indicated by a supralinear dot (ā) and a sublinear dot (ē), corresponding to their relative height. It is obvious that the position of the vowel dot in the Abu al-Aswad story7 is in accordance with this Syriac practice. It is equally obvious that the Arabic terms naṣb and khafḍ, as well as rafʿ, may be interpreted lexically in the same sense as the Syriac terms. (Versteegh 1993, 30)

23Versteegh accepts Revell’s idea that Syriac grammarians perceived sounds at the back of the mouth as ‘high’. This concept of phonetic ‘height’ is likely a natural development from the earlier Syriac context, which created terms from men lʿal and men ltaḥt. Versteegh and Revell, however, assume that the principle of ‘high’ and ‘low’ vowel sounds entered the Arabic tradition along with calques of zqɔp̄ɔ and rḇɔṣɔ; that is, naṣb and khafḍ. This conclusion is untenable on both chronological and linguistic grounds. The root zqp in the context of vowel phonology is not attested in any Syriac source before a commentary written by Ḥunayn ibn Isḥāq (d. 873), a century after naṣb appeared in al-Kalbī’s tafsīr (Hoffmann 1880, 10, ln. 13; 14, lns. 21–23). The earliest source I know of with rbṣ in a similar context is Elias of Ṣoba’s eleventh-century grammar, again, well after al-Kalbī (Gottheil 1887, 7–8). That said, while the Syriac terms zqɔp̄ɔ and rḇɔṣɔ cannot be the source of Arabic naṣb and khafḍ, respectively, Arabic grammarians did incorporate some height-based principles into their explanations of vocalisation.

3.2. Early Vowel Phonology in the Arabic Tradition

24After completing the list of the twenty-nine Arabic letters in his grammar (the Kitāb), Sībawayh (d. 793 or 796) says that there are actually thirty-five letters,8 some of which branch off of the others. Two of these additional letters are “the ʾalif which is tilted with great ʾimāla” and “the ʾalif of tafkhīm” (Harun 1982, IV: 432: and Here ʾimāla ‘inclination, bending down’ indicates the shift of an ʾalif towards /i/, such that the resulting sound is not /a/, but /ɛ/ or /æ/. Its opposite is tafkhīm ‘magnifying, thickening’, which indicates the shift of /a/ towards /ɔ/.9 This term may be related to the principle that Jacob of Edessa illustrated with his classification of /ɔ/ as a ʿḇe ‘thick’ vowel.10 But beyond this similarity, Rafael Talmon points out that Sībawayh uses another term specifically to indicate an ʾalif that does not undergo ʾimāla: naṣb (Talmon 1996, 291; 2003, 239). Apparently, at some very early stage, naṣb and ʾimāla were contrastive terms that distinguished the allophonic variants of ʾalif.

25The use of naṣb and ʾimāla to describe ʾalif probably began well before Sībawayh wrote the Kitāb, perhaps even before any Arabic vowels had absolute names. The main evidence for this conclusion comes from the first chapter of the Kitāb, where Sībawayh presents a systematic usage for the Arabic vowel names fatḥ, kasr, and ḍamm as distinct from the case names naṣb, jarr, and rafʿ. Prior to his time, all of these terms could indicate both vowels and cases, as seen in the work of al-Kālbī (Versteegh 1993, 125). Sībawayh was the first person to separate the two sets (Talmon 2003, 283),11 relegating fatḥ, kasr, and ḍamm to the status of phonological descriptors, whereas the so-called ʾiʿrābī ‘declensional’ terms were reserved for vowels with grammatical import. Sībawayh’s use of naṣb to indicate the quality of ʾalif is thus anomalous: according to his own instructions, it is a declensional term, and not a word for describing internal vowels. This inconsistency suggests that the duality of ʾimāla and naṣb was fixed in the Arabic tradition long before Sībawayh isolated naṣb as the name for the accusative case, and he is merely transmitting this early convention when he uses naṣb to describe an allophone of ʾalif (see Harun 1982, IV: 125–26, 143, for this contrastive use of ʾimāla and naṣb).

26Sībawayh includes one other variant of ʾalif in his discussion of naṣb and ʾimāla. He first states that there are seven letters which prevent ʾimāla when they precede ʾalif: ṣād, ḍād, ṭāʾ, Ẓāʾ, ghayn, qāf, and khāʾ, and then explains:

You abstain from ʾimāla for these letters because they are letters which are elevated towards the top of the palate, and if the ʾalif is pronounced from their point of articulation, it goes up towards the top of the palate. Thus, when [the ʾalif] is with these elevated letters, they overpower it, just as the kasra overpowers it in masājid12 and other variations [that have ʾimāla]. So when the letters are elevated, and the ʾalif goes upwards, and [the letters] draw near to it, then the articulation is in a single manner, which is less burdensome for them. (Harun 1982, IV: 129)

27This passage describes the production of a backed a-vowel that, like ʾimāla, only occurs in specific phonological contexts. In this case, that context is immediately after a velar or emphatic consonant, and the vowel itself requires shifting the articulation of /a/ back towards the soft palate, approximating /α/ or /ɔ/. Given that Sībawayh highlights the parallel between this vowel and ʾimāla, one might expect him to call it ʾalif al-tafkhīm, as he does in his description of the alphabet; but he does not. In fact, the term tafkhīm does not appear anywhere in this or any other of the Kitāb’s chapters on ʾimāla. Instead, this backed version of ʾalif is included along with just one of many irregular situations that affect the normal ʾimāla rules. If Sībawayh is indeed transmitting an earlier phonological tradition that contrasted naṣb and ʾimāla, then perhaps that tradition did not have terminology to distinguish /a/ from /ɔ/, and instead referred to both as naṣb— that is, ‘not ʾimāla.’ As such, naṣb and ʾimāla were effectively relative vowel terms, each indicating a particular allophone as either relatively fronted (ʾimāla—/ɛ/, /æ/) or relatively backed (naṣb—/a/, /α/, /ɔ/). This usage of naṣb (standing upright) and ʾimāla (bending down) thus conforms to the two-way relative descriptions of vowels in the early Syriac and Hebrew traditions, paralleling the association of ‘high’ with backness and ‘low’ with frontedness.

28The term naṣb must have become associated with the specific quality of an unaltered ʾalif—/a/—prior to al-Kalbī’s time. Then, by analogy with naṣb and according to the understanding of back vowels as ‘higher’, rafʿ ‘rising’ and khafḍ ‘lowering’ were linked to /u/ and /i/, respectively. Throughout this process, naṣb retained its now-secondary use as the opposite of ʾimāla, as evidenced by Sībawayh’s Kitāb, and, by extension, it retained some function as a way to denote /α/ in certain contexts. It seems then that naṣb is the likely source of Syriac zqp ‘standing upright’ as a descriptor of /ɔ/, first seen in Ḥunayn ibn Isḥāq’s commentary, mentioned above. Syriac grammarians had a concept of ‘openness’ in their vowel phonology as early as Jacob of Edessa, so when they began naming their vowels, pɔṯaḥ—later, pṯɔḥɔ—was the obvious term for /a/.13 Then when ninth-century Syrians needed a way to describe their secondary a-vowel, /ɔ/, they looked to their Arabic contemporaries, and calqued the second term which they used to distinguish a-vowels (i.e., naṣb). The results were zɔqep̄ and zqip̄ɔ, which became zqɔp̄ɔ ‘standing upright’ by the eleventh century.

29This process also fits Versteegh’s expected development of the vowel term rḇɔṣɔ, which, in direct contrast to zqɔp̄ɔ, he suggests can mean ‘lowering’. As such, one could conclude that when Syriac grammarians needed a term for their secondary i-vowel, /e/, they calqued the second Arabic term for i-vowels, khafḍ ‘lowering’. The Syriac root rbṣ, however, does not exactly mean ‘lowering’ or ‘depressing’ as a physical motion, but rather refers to ‘compression’, and the vowel name rḇɔṣɔ probably derives from the articulation of /e/ with relatively compressed lips in comparison to more-open vowels. Neither is it attested as a vowel descriptor in Syriac before grammars of the eleventh century, which complicates this reconstruction of the term’s origin. Furthermore, these later sources—particularly Elias of Tirhan’s grammar— may also have incorporated an Arabic tripartite division of vowels into the older Syriac relative vowel system, further distorting the picture.

3.3. Reinterpretation of Vowel Phonology in the Arabic Grammatical Tradition

30According to Versteegh and Revell’s argument, when Arabic grammarians adapted the Syriac vowel dots for Arabic, they also calqued their vowel terms, using a Syriac theory of ‘height’ that was linked to phonetic backness. As discussed above, there is no terminology in the early Syriac tradition that supports the idea that the Arabic case names are calques of Syriac terms, but the Arabic vowel names are certainly related to some phonological conception that relates backness to height. Arabic grammarians, however, reinterpreted this earliest vowel phonology, and instead explained non-consonantal phonemes based on physical motion, specifically associating them with the movement of airflow during articulation.

31In contrast to the idea of height-as-backness, Ilan Eldar proposes that medieval Arabic grammarians understood vowel phonology as effects on air. Taking into account how rafʿ ‘rising’ usually indicates a high position, whereas naṣb describes something which is set upright (Eldar 1983, 45), he argues that naṣb, rafʿ, and khafḍ ‘lowering’ were interpreted in terms of the direction of airflow during vowel articulation. He focuses on the relationship between Arabic case names and Hebrew vowel phonology (see below), but for now it is sufficient to explain his theory with respect to Arabic. In short, /a/ is called naṣb because when one articulates it, the flow of air proceeds straight ahead, unimpeded; it is thus ‘fixed in place’ or ‘standing upright’. By contrast, when articulating /u/, the airstream moves upwards; it is rafʿ. Then for /i/, the air tilts downwards, making it khafḍ.14 Eldar points out that Sībawayh (d. 796) emphasises the relationship between vowel sounds and air (Eldar 1983, 48). In his description of the alphabet in the Kitāb, Sībawayh writes:

Among [the letters] are the layyina [‘soft, flexible’], which are wāw and yāʾ, because their articulation is widened for the air of the sound, more than the widening of other [letters] besides them, as you say: wa ʾayy^un and al-wāw, and if you want, you can make the sound occur with lengthening.
[Also] among them is the hāwī [‘airy, breathy’], which is a letter whose articulation is widened for the air of the sound even more than the widening of the articulation of yāʾ and wāw—because you press your lips together for wāw, and you raise your tongue in front of the palate for yāʾ—and it is ʾalif.
These three are the subtlest of the letters due to their articulations’ widening, and the subtlest and widest of them is ʾalif, then yāʾ, then wāw. (Harun 1982, 435–36)

32Sībawayh distinguishes the three Arabic matres lectionis according to their effects on air during speech. Wāw and yāʾ are different from ʾalif specifically because their articulation requires some obstruction of airflow, either by the lips or the tongue, whereas ʾalif is a pure hāwī ‘airy, breathy’ letter. He arranges them in order of ‘wideness’, which seems to relate to the amount of airflow allowed by each letter, and corresponds to the relative openness of the vowels.

33The introduction of Kitāb al-ʿAyn also stresses the effect on air when discussing the matres lectionis. Convention attributes this text to al-Khalīl ibn Aḥmad al-Farāhīdī (d. 786 or 791), an early scholar of prosody and one of Sībawayh’s teachers. In reality, most of the text was compiled after his death, probably by another student, al-Layth ibn al-Muẓaffar (d. c. 803). Despite this, the book’s arrangement and parts of the introduction are probably original to al-Khalīl, and in any case the material in the introduction is quite old (Sellheim 2012a; 2012b). In its preliminary discussion on the letters of the alphabet, the text reads:

34Al-Layth said: Al-Khalīl said:

‘In Arabic there are twenty-nine letters. Among them are twenty-five sound letters which have spaces and steps, and four letters of the [oral] cavity, which are the soft wāw, yāʾ, and ʾalif, as well as the hamza. They are called jawf because they exit from the cavity, but do not occur at one of the steps of the tongue, or the steps of the throat, or the step of the palate. Instead, they are airy, in the air, for they do not have a space to attach to besides the cavity. He [al-Khalīl] frequently used to say: the soft ʾalif, the wāw, and the yāʾ are airy, that is, they are in the air.’ (Makhzūmī 1985, 57)

35The so-called ṣiḥāḥ ‘strong, firm’ letters contrast with the layyina ‘soft, flexible’ ʾalif, wāw, and yāʾ. The primary difference between them is that the former letters connect to specific points within the mouth, whereas the latter exist entirely as an effect in the air. Sībawayh cites al-Khalīl in his Kitāb more than any other source, but notably does not use al-Khalīl’s phonetic terminology in his chapters on phonology (Versteegh 1993, 16); and yet here Kitāb al-ʿAyn agree. These early Arabic grammarians understood vowels differently from consonantal phonemes, associating them not with any particular ‘back’ or ‘front’ locations in the mouth, but rather describing them based on airflow during articulation. The matres lectionis, then, are called layyina because they alone among the letters incline as streams of air.

36These two early sources thus support Eldar’s argument that Arabic vowel terminology was created based on airflow, or at least that it was interpreted that way by later scholars. Eldar cites a key passage from Ibn Sīnā’s (d. 1037) Risāla fī ʾAsbāb Ḥudūth al-Ḥurūf (Eldar 1983, 46–47; the English translation is my own):

As for the sounding ʾalif and its sister, fatḥa, I believe its articulation is with the loosing of air smoothly, without obstructions.
For the sounding wāw and its sister, ḍamma, I believe its articulation is with the loosing of air and a little contracting of the articulation point,16 while inclining smoothly upwards at it.
For the sounding yāʾ and its sister, kasra, I believe its articulation is from the loosing of air and a little contracting of the articulation point, while inclining smoothly downwards at it.

37It seems that Ibn Sīnā reached the same conclusion as Eldar, attributing a unique direction of airflow to each of the Arabic vowels, quite likely based on the names of case vowels (rafʿ, naṣb, khafḍ; ‘rising’, ‘standing upright’, ‘lowering’, respectively). This passage fully illustrates the tripartite division of Arabic vowels according to airflow, but Eldar does not discuss the full significance of Ibn Sīnā’s word choice. The root myl ‘inclining’ used here is the same as that of the term ʾimāla, which suggests that, at least for Ibn Sīnā, even the allophonic variants of ʾalif could be explained as tilting streams of airflow. This conception of vowel phonology must have been current, at least in some circles of Arabic grammarians, by the early eleventh century, and it also appears in Syriac and Hebrew grammatical texts at roughly the same time.

4.0. Two Examples of Syncretisation in Phonological Systems of the Tenth and Eleventh Centuries

4.1. Elias of Tirhan’s Syriac Grammar

38As the Arabic language and its grammatical tradition became dominant across the Middle East, Syriac and Hebrew grammarians adapted elements of the Arabic tripartite division of vowels to fit their older relative systems. Perhaps no author is more emblematic of this development than Elias of Tirhan (d. 1049), who wrote a Syriac grammar specifically for an Arabic-speaking audience in the first half of the eleventh century. In his chapter on vowel pointing, Elias groups the vowels by association with the matres lectionis; three for ʾalap̄: zqɔp̄ɔ /ɔ/, pṯɔḥɔ /a/, and rḇɔṣɔ or sheshlɔ /e/;17 two for waw: ḥḇɔṣɔ /u/ and massaqɔ or rwaḥtɔ /o/; and one simply called yoḏ /i/ (Baethgen 1880, lns. 15–18). Besides the terms which he presents in this chapter, Elias describes vowels a few other ways throughout the text, including: ḥḇɔṣɔ (Baethgen 1880, lns. 16–21), ḥḇiṣtɔ (Baethgen 1880, lns. 1–5) for /u/; and two versions of waw, which he calls meṯḥḇaṣɔ ‘contracted’ and meṯrwaḥɔ ‘widened’ (Baethgen 1880, lns. 19–21).

39At work here is the old Syriac tradition of ‘wide-and-narrow’ vowels: /u/ requires contraction of the mouth, and is thus meṯḥḇaṣɔ. Its ‘widened’ counterpart is then /o/, which is meṯrwaḥɔ. Ḥḇɔṣɔ ‘contracting’ and rwaḥtɔ ‘widening’ are likewise Elias’s names for /u/ and /o/. All of these terms describe mouth movement and depend on the principle of two-way contrastive vowels laid out by Jacob of Edessa. This idea explains how roots like ḥbṣ can refer to an u-vowel here, but other authors use it to mean an i-vowel:18 it has meaning only in comparison to other vowels.

40There are also indications of Arabic influence here. Most prominent is massaqɔ19 ‘raised up’, which stands out as a C-stem form in a group of terms otherwise derived from G-stem participles. This uniqueness suggests that it came into use separately from the other terms, probably as a calque of the Arabic marfūʿ ‘raised’, but it preserves the relative nature of other Syriac vowel terms. Elias applies it to the ‘higher’ (more-backed) of a pair of vowels—/o/ as opposed to /u/—following the Syriac association of height with backness. There is even evidence that the Arabic phonetic theory based on airflow affected Elias of Tirhan’s understanding of vowels. He was writing for an Arabic-speaking audience, so many of his explanations are meant to resonate with people familiar with Arabic. He explains that there are three zawʿe ‘movements’ in Syriac (Baethgen 1880, lns. 19–21), directly translating the Arabic word for ‘short’ vowels, ḥarakāt ‘movements’, which to him are vowels that are written without matres lectionis. As such, the Syriac zawʿe are pṯɔḥɔ (/a/), rḇɔṣɔ (/e/), and zqɔp̄ɔ (/ɔ/), and he considers them each to be pelgut ʾalap̄ ‘half-ʾalap̄’ (Baethgen 1880, ln. 21– ܕܟ , ln. 2). This grouping of terms parallels the Arabic triad of naṣb (/a/), khafḍ (/i/), and rafʿ (/u/), with one central vowel having unobstructed airflow (/a/), and the others being pronounced with relatively ‘upward’ (/ɔ/) and ‘downward’ (/e/) movement. Similarly, it corresponds to the Arabic allophones of ʾalif: naṣb (/a/), ʾimāla (/ɛ/ or /æ/), and tafkhīm (/ɔ/). Moreover, while explaining a case where one should read /o/ instead of /u/, Elias says lʿel ʾapeqn lbart qɔlɔ ‘we pronounce the sound upwards’ (Baethgen 1880, lns. 5–6). While he may be referring to the idea that /o/ is a ‘higher’ (more-back) vowel than /u/, his language mirrors that of Ibn Sīnā (d. 1037), potentially indicating a direction of airflow.

4.2. Saadya Gaon’s Hebrew Grammar

41Vowel phonology in the Hebrew tradition underwent a similar development in the post-Sībawayh era, with elements of the earlier relative system combining with an airflow theory by the eleventh century. At the centre of this process was Saadya Gaon’s (d. 942) ‘vowel scale’, which he recorded in the fifth chapter of his grammar, Kutub al-Lugha. In this chapter, titled Al-Qawl fī al-Nagham ‘Discourse on Vocal Melody’, he lists the Hebrew ʾiʿrāb ‘vowels’ from high to low: ḥolem /o/, qɔmeṣ /ɔ/, pataḥ (or p/fatḥa) /a/, segol /ɛ/, ṣere /e/, ḥiriq /i/, and shureq /u/ (Skoss 1952, 285).20 This scale is a fully-articulated version of the milleʿel and milleraʿ comparisons of earlier masoretic homograph lists. It is also precisely what would be expected if a Syriac phonologist undertook the same exercise, ranking the vowels from high to low (perhaps men lʿal to men ltaḥt?) according to backness. The one exception is /u/, which Saadya seems to remove from the scale in order to support a morphological principle for which he argues later on (see Skoss 1952, 316).

42Saadya confirms that his organisation of vowels is based on backness, saying:

As for the explanation of the third chapter, which is the knowledge of the places in the mouth, and their levels, we say: if one chose to interrupt their vocal melody at the first point where it could be cut off after its ascension from the throat; then ḥolem would emerge, with [the ḥolem]’s force proceeding ahead of [that point], not turning upwards or downwards. But if one wanted to take [the vocalic melody] past this point and then interrupt it, the force of qɔmeṣ would appear, and its movement would be specifically towards the top of the palate. (Skoss 1952, 292, lns. 7–13)

43He proceeds in this manner for the rest of the vowels, saying for each one that you tajāwaz ‘pass’ the mawḍiʿ ‘articulation point’ of the previous vowel. But beyond showing how Saadya arranges vowels according to backness, this passage reveals the degree to which he is familiar with the Arabic grammatical tradition. His explanation of /ɔ/ (i.e., qɔmeṣ) is the same as Sībawayh’s, and his progression through the mawāḍiʿ ‘articulation points’ and marātib ‘levels’ of the vowels mimics the language that both Sībawayh and al-Khalīl use in their classifications of consonants (Harun 1982, IV: 431–36; Makhzumi 1985, 52–57). Additionally, his explanation of the quwwa ‘force’ of each vowel is reminiscent of Arabic descriptions of airflow, focusing on the ḥaraka ‘movement’ ilā fawq ‘upwards’ or ilā ʾasfal ‘downwards’. At the same time, Saadya modifies this principle, stating explicitly that /o/ is the ghayr ḥāʾida ‘unwavering’ vowel, in contrast to Ibn Sīnā’s understanding that /a/ was the vowel that does not tilt up or down (i.e., naṣb).

44Many of Saadya’s vowel names seem to be novel, with only the Aramaic qɔmeṣ and pataḥ attested in the Hebrew tradition prior to this text. Segol ‘a cluster of grapes’ is likely derived from the name of the Hebrew accent sign with the same form, but the other four may be Saadya’s own tenth-century Hebraisms, all based on mouth movement.21 However, these innovations did not immediately catch on, and until at least the eleventh century, grammarians continued referring to /o/, /u/, /e/, and /i/ by either phonetic transcription or the number of dots in each sign (Khan 2000, 24; Steiner 2005, 377–78; Dotan 2007, 633). In fact, rather than accepting Saadya’s scale as fully authoritative, his successors modified it to better align it with Arabic phonology.

45Sometime in the eleventh century, an anonymous Hebrew grammarian took the Arabic concept of tripartite airflow and merged it with Saadya’s vowel scale in an abridged version of Al-Qawl fī al-Nagham that is partially extant (Eldar 1981, 105–18). Titled Kitāb Naḥw al-ʿIbrānī ‘The Book of Hebrew Inflection’, the abridgement maintains a scale arranged by phonetic backness, but also divides the vowels into three groups: al-rafʿ (/o/ and /u/), al-khafḍ (/e/ and /i/), and al-naṣb (/ɔ/, /a/, and /ɛ/). Unlike in Saadya’s version, the abridger does not use any of the ‘modern’ vowel names besides qɔmeṣ (/ɔ/) and pɔtaḥ (/a/), albeit in the Arabicised forms al-qamṣa and al-fatḥa. Instead, the author refers to /o/, /u/, /e/, and /i/ by spelling them phonetically, and also calls /i/ and /ɛ/ “the one dot” and “the three dots,” respectively. It places vowels on a scale by ranking their status in the three groups: /o/ is the greater rafʿ, /ɔ/ the greater naṣb, /a/ the middle naṣb, /ɛ/ the lesser naṣb, /e/ the lesser khafḍ, and /i/ the greater khafḍ.

46The author also follows the original text in removing /u/ from the scale, although the fragment breaks off before explaining the reason behind this choice. Presumably, /u/ was the ‘lesser rafʿ’, as that classification would correspond to the Arabic notion that /u/ emits an upward stream of air, while also following Saadya’s original scale and being phonetically ‘lower’ than /o/. As another example of the same principles: calling /e/ the ‘lesser khafḍ’ indicates that one should pronounce the vowel with a downward inclination of air, but not quite as inclined as the ‘greater khafḍ’ (/i/). Then the location—fifth from the top of the scale—designates the lesser khafḍ as the fifth-most-backed of the vowels. This syncretic Arabic-and-Saadyan scale thus classifies every vowel according to both its effect on airflow and relative amount of backing, combining principles from both the Arabic and Masoretic phonological traditions.

5.0. Conclusion

47The development of Syriac, Arabic, and Hebrew phonological thought as it relates to vocalisation had significant inter-linguistic overlap during the medieval period. Early Syriac and masoretic sources show that both traditions perceived vowel phonology according to a relative system. This system distinguished homographs by the comparative ‘openness’ of their vocalisation and, at least in the Syriac tradition, it used dots above or below a word to indicate its vowels. Then, over time, terms like milleʿel and milleraʿ developed out of the perceived connection between dot position and vowel quality, and phonetic backness came to be associated with ‘height’.

48The Arabic grammatical tradition emerged in this relative context, and although the early uses of naṣb ‘standing upright’ and ʾimāla ‘bending down’ reflect height-based principles similar to those of the Syrians and Masoretes, later Arabic grammarians interpreted their vowel names as designations of the direction of airflow when articulating vowels. Before the late eighth century, one of these terms—naṣb ‘standing up’—had an extended usage that helped distinguish allophones of ʾalif, including a back vowel between /a/ and /ɔ/. It is likely that the Syriac name for /ɔ/, zqɔp̄ɔ ‘standing up’, is a calque of this term. Other Syriac vowel names may also be Arabic calques, but it is difficult to tell due to the syncretisation of phonological systems that happened in the tenth and eleventh centuries.

49Elias of Tirhan’s eleventh-century Syriac grammar exhibits this syncretic phenomenon, as he incorporates some of the Arabic tripartite division of airflow with the old Syriac system of ‘wide-and-narrow’ vowels. Saadya Gaon’s tenth-century Hebrew grammar also demonstrates this phonological blending, as his vowel scale combines the masoretic hierarchy of vowels with the Arabic emphasis on airflow.

50This discussion is by no means an exhaustive account of all the connections between medieval Semitic vocalisation traditions, but rather it shows that it is possible to discern such links by comparing the phonological theories that authors used to describe their own languages. There is much more work to be done in order to connect the dots.