Remarks on syllable structure and metrical structure in Biblical Hebrew

In the Middle Ages Biblical Hebrew was transmitted in a variety of oral reading traditions, which became textualized in systems of vocalization signs. The two most important oral traditions were the Tiberian and the Babylonian, which were repre-sentedbydifferentvocalizationsignsystems.Thesetwooraltraditionshadtheirorigins in ancient Palestine. Although closely related, they exhibit several differences. These include differences in syllable and metrical structure. This paper examines how the syllable and metrical structure of the two traditions reflected by the medieval vocalization sign systems should be reconstructed. The Tiberian tradition exhibits an ‘onset typology’ of syllabification, where word-internal /CCC/ clusters are syllabified /C.CC/ and word-initial clusters are syllabified within the onset /CC-/. The Babylonian tradition exhibits a right-to-left computation of syllables resulting in a ‘coda typology,’ whereby the second consonant of a word-internal sequence /CCC/ is syllabified as a coda, viz. /CC.C/, and word-initial clusters are syllabified C.C, with the first consonant extra-syllabic.

In this paper I shall examine some features of the structure of syllables and metrical structure in the Tiberian and Babylonian reading traditions of Biblical Hebrew.
The Tiberian reading tradition is recorded by the Tiberian vocalization system, which is the system that appears in modern printed Bibles.This system of vocalization was developed by the Masoretes of Tiberias in the early Islamic period.It appears in a standardized form in medieval Bible manuscripts, which form the basis of modern printed editions.Although the Tiberian vocalization has come down to us as the standard form of written vocalization, the oral Tiberian reading tradition that it originally reflected was largely lost to knowledge in the later Middle Ages and Jewish communities read the Tiberian signs with other local traditions of pronunciation.The Tiberian reading tradition can now be reconstructed to a large degree on the basis of a variety of extant medieval sources.The reading tradition that is reconstructed from such sources does not correspond in many details to the descriptions of the pronunciation of Biblical Hebrew that appear in current reference grammars.In order to study the phonology of the Tiberian tradition of Hebrew, therefore, it is essential first to reconstruct through philological research the original form of the tradition.1 The Tiberian reading tradition of Biblical Hebrew has its roots in an ancient reading tradition that was transmitted orally in the first millennium C.E. before the creation of the written notation of vocalization signs.Another reading tradition of Biblical Hebrew was transmitted in Babylonia in the first millennium C.E., which, in the early Islamic period, came to be recorded in another type of sign notation created by Babylonian Masoretes, known as Babylonian vocalization.The Tiberian and Babylonian reading traditions are historically closely related and are likely to have derived from the same source, which can be identified as a proto-Masoretic type of reading existing in the Second Temple period.The Babylonian reading tradition has been preserved in numerous manuscripts with Babylonian vocalization.2Although the Tiberian and Babylonian traditions are closely related, there are, nevertheless, numerous small differences in phonology and morphology between the two.Some of these differences can be attributed to historical change that took place in the traditions over the course of their transmission in the first millennium.It is also possible that some differences reflect variations that existed already within the Proto-Masoretic reading in the Second Temple period.What is of interest to us in this 1 For a description of the current state of philological research see Khan (2020).Short overviews can be found in Khan (2013aKhan ( , 2013c)). 2 For a comprehensive description of this vocalization see Yeivin (1985).Short overviews are provided by Khan (2013d) and Heijmans (2016).
Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020) 7-30 paper is that the Babylonian tradition differed from the Tiberian in a number of aspects of syllable structure and metrical structure.

The Tiberian reading tradition
In order to understand the syllable structure in the Tiberian reading tradition it is important to know the distribution of vowel length.Long vowels in the reading tradition include those that are represented by full vowel signs when they are either (i) in a stressed syllable or (ii) in an open unstressed syllable.Elsewhere a vowel represented by a full vowel sign is pronounced short.Examples: In order to establish the synchronic phonological representation of the vowels of the Tiberian reading tradition one must distinguish between (i) vowels which are invariably long and include length in their underlying phonological representation and (ii) vowels whose length is determined by syllable structure and stress so are of unspecified length at a phonological level (Khan 2013e).
The long vowel phonemes include: long qameṣ /ɔ/, ḥolem /ō/, ṣere /ē/, long shureq /ū/, long ḥireq /ī/ (typically written with yod), e.g.To the second category of vowels we should add also /e/ and /o/ without specified length.These are represented by the ṣere and ḥolem vowel signs respectively in the stressed syllable of certain forms.Since stressed vowels are always long, on a phonetic level these are not distinguishable from ṣere and ḥolem representing phonemes with underlying length.This is necessary to account for apparent discrepancies in the historical development of vowels in several morphological forms, in which pataḥ (a vowel with no specified length feature) occurs in parallel with ṣere and ḥolem (Sarauw 1939, 56-64;Khan 1994).This applies, for example, to nouns with an originally doubled final khan Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020) 7-30 consonant.In forms deriving from the *qall pattern the vowel is pataḥ, e.g.‫ַ֫ר‬ ‫ב‬ /ˈrav/ 'much' , and in forms deriving from the *qill and *qull pattern, the vowel is ṣere and ḥolem respectively, e.g.[ˌqɔːðɔːˈʃiːm] 'holinesses ' (Exod. 29.37).
Another key starting point for analyzing the syllable structure of the Tiberian reading tradition of Biblical Hebrew is understanding the original function of the shewa sign.The quality of vocalic shewa in the Tiberian tradition was generally the same as that of the pataḥ vowel sign, i.e., the maximally low vowel [a].When occurring before a guttural consonant or the letter yod it was realized with a different quality through an assimilatory process.Before a guttural it was realized as a short vowel with the quality of the vowel on the guttural, e.g.[moʕoːnoː] 'his dwelling place' .Before yod it was realized as a short vowel with the quality of short ḥireq [i], e.g.‫ְבּ‬ ‫י‬ ‫וֹ‬ ‫ם‬ [biˈjoːom] 'on the day' .The default pronunciation of vocalic shewa with the quality of [a] was equivalent to that of the ḥaṭef pataḥ sign ( ֲ ).Both the vocalic shewa and the vowels expressed by ḥaṭef signs were short vowels that, in principle, had the same quantity as short vowels in closed unstressed syllables, which were represented in standard Tiberian vocalization by a simple vowel sign.So, the vocalic shewa in a word such as  and Linguistics 12 (2020) 7-30 It is an establised fact that every letter that has a 'light' (i.e.short) vowel requires a shewa unless this is precluded by a dagesh (in the following letter), as we exemplified at the beginning of our treatise, , or by a shewa that is adjacent to it, i.e. after it, as in ‫ַא‬ ‫ְב‬ ‫ָר‬ ‫ָה‬ ‫ם‬ ; the ʾalef has a short vowel, and were it not for the shewa that comes after it, we would have given it a shewa.4 The author of this treatise did not feel that there was a quantity difference between the vowel written with the ḥaṭef sign and the vowel represented by the full vowel sign.In his view it was the syllable structure which necessitated the notational distinction and not the quantity of the vowel segment.The shortness of the vowel in a closed syllable was indicated by the dagesh or shewa on the subsequent consonant.For the sake of economy of notation no additional sign was added to the vowel sign.
Further evidence for the quantitative equivalence of shewa and ḥaṭef vowels, on the one hand, and short vowels represented by full vowel signs in closed unstressed syllables, on the other, can be found in the use of the shewa and ḥaṭef signs in a variety of non-standard Tiberian manuscripts from the Cairo Genizah.These manuscripts sometimes represent a short /a/ vowel in a closed unstressed syllable with a ḥaṭef pataḥ or a shewa sign.5Even some of the standard Tiberian Masoretic codices contain a few cases of such a phenomenon, e.g. in Codex Leningradensis (L): 'on the magicians' (Exod.9:11), 'and we will kill him' (Jdg 16:2).6Furthermore in medieval Judaeo-Arabic texts with Tiberian vocalization shewa and ḥaṭef pataḥ are used to represent Arabic short /a/ in both open and closed syllables.7 In the Tiberian Masoretic literature a consonant with a vocalic shewa or a ḥaṭef vowel sign was not considered to stand independently, but was said to be bound to the following consonants.Thus the word ‫ִתּ‬ ‫ְס‬ ‫ְפּ‬ ‫֖ר‬ ‫וּ‬ 'you shall count ' (Lev. 23.16) was considered to have been composed of two prosodic units [tʰis-pʰaʀ̟ uː].The sources refer to these prosodic units by the Arabic term maqṭaʕ, which is used in the Arabic grammatical literature to refer to a syllable.The treatise Hidāyat al-Qāriʾ 'The Guide for the Reader' , written by the Karaite grammarian ʾAbū al-Faraj Hārūn in the first half of the eleventh century C.E., notes that syllables thus formed have the status of words, i.e., they can stand independently: khan Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020) 7 Another of it features is that it divides a word into (units) that have the status of words.This is because every letter at the end of a word is quiescent when it is deprived of an accompanying vowel and this letter that is deprived of a vowel is the stopping point (maḥaṭṭ) of the word and its place of division (maqṭaʕ), as in , in which the taw is the stopping point of the word, and ‫א‬ ‫וֹ‬ ‫ר‬ , in which the resh is the stopping point of the word, as so forth.A quiescent shewa in the middle of a word has the same status, for it is in a sense a stopping point on account of its quiescence, for example In various passages in Hidāyat al-Qāriʾ there are references to the fact that a vocalic shewa or ḥaṭef vowel is read more quickly than a following full vowel sign, e.g.'The shewa makes a letter mobile and causes it to be uttered quickly, so that one cannot tarry on that letter' ,8 'The shewa moves quickly forwards' .9By contrast a vowel is read more slowly, e.g.'A vowel has an indissoluble feature, namely slowness and steadiness' .10These descriptions can be interpreted as referring to the rhythmic structure of the prosodic unit consisting of a vocalic shewa followed by a vowel whereby this unit is a prosodic foot consisting of an iambic metrical pattern with a weak syllable followed by a strong syllable, which can be represented (. *).On a prosodic level, therefore, a word such as it will be noticed that the first syllable by itself has the status of a foot, i.e. [(tʰis).(pʰa.ˈʀ̟uː)], where feet are marked by rounded brackets.This is in conformity with the current state of research on the typology of the metrical phonology of the world's languages.The foot (pʰa.ˈʀ̟uː), as remarked, is iambic, i.e. it consists of two syllables, of which the second is the stronger.It is a binary foot consisting of a light syllable with one mora, viz.CV, and a heavy syllable consisting of two mora, viz.CVV.It is represented here (. *), where the star * represents the strong prominent syllable and the dot the weak syllable.In many languages with metrical phonology with binary feet, the feet may be binary either in the number of their syllables, as in the foot (CVCVV), which is known as a syllabic foot, or in the number of their morae, known as a moraic foot.This means that a heavy syllable with two morae, viz.CVV or CVC, normally represented in metrical phonology by (*), could function as a foot in the metrical scansion of a word alongside a syllabic foot.11The metrical parsing of [tʰis.pʰaʀ̟ uː] would, therefore, be [(*), (. *)].The CVC syllable with a vowel [tʰis], which constitutes an independent foot, would be metrically stronger than the first syllable of the foot (. *), which is represented by a shewa sign.This would conform to the medieval descriptions cited above, which state that a vowel has the feature of 'slowness and steadiness' whereas a shewa 'moves quickly forwards' .These differences in prominence can be represented by a metrical grid.12In the grid the relative prominences are marked by differences in heights of columns of index marks: (2) (tʰis) (pʰa.ˈʀ̟ uː) As can be seen, the syllable with the main stress is the most prominent.This stress occurs on the strong syllable of the (.*) foot.
The Masoretic notion of maqṭaʕ, therefore, can be equated with the notion of foot in the prosodic hierarchy rather than syllable.The foot is of relevance for some phonological processes in the Tiberian pronunciation tradition.This is a 11 For the typology of feet in iambic metrical systems see Hayes (1985Hayes ( , 1995)), Kager (1993Kager ( , 383, 2007, 200-201), 200-201).12 For hierarchical arrangement of linguistic rhythm on a grid see, for example, Liberman and Prince (1977), Hayes (1995, 26-31) and Halle and Vergnaud (1987).key justification for the reality of such metrical constituents (Nespor and Vogel 2012).The occurrence pattern of the allophones of Tiberian resh is a clear example of this.Tiberian resh had an advanced uvular allophone [ʀ̟ ], which can be considered its default realization,13 and a pharyngealized, apico-alveolar allophone [rˁ], which was conditioned by certain environments (Khan 1995(Khan , 2013b)).According to the medieval sources the apico-alveolar allophone occurred when resh is preceded by the consonants ‫ד‬ ‫ז‬ ‫צ‬ ‫ת‬ ‫ט‬ ‫ס‬ ‫ל‬ ‫ן‬ or followed by ‫ל‬ ‫ן‬ and when either resh or one of these consonants has shewa.This can be reformulated as the rule that apico-alveolar resh occurs when in immediate contact with a preceding alveolar, e.g., ‫ְבּ‬ ‫ִמ‬ ‫ְז‬ ‫ֶ֖ר‬ ‫ה‬ [bamizˈrˁɛː] 'with a pitch fork' (Jer.15.7), or in the same foot as a preceding alveolar, e.g., ‫ַדּ‬ ‫ְר‬ ‫֖כּ‬ ‫וֹ‬ [darˁˈkʰoː] 'his way' (Gen.24.21), or when the resh is in immediate contact with or in the same syllable as a following ‫ל‬ or ‫,ן‬ e.g., [ʕarˁleː-ˈleːev] 'uncircumcised in heart' (Jer.9.25), ‫ַר‬ ‫ְנּ‬ ‫֣נ‬ ‫וּ‬ [rˁannaˈnuː] 'rejoice (mpl)!' (Ps.33.1).What is of significance for the reconstruction of prosodic structure is that a resh preceded by an alveolar consonant with vocalic shewa was realized with the apico-alveolar allophone, e.g.,

‫ְצ‬ ‫ר‬ ‫וּ‬ ‫ָפ‬ ֔ ‫ה‬
[sˁɑrˁuːˈfɔː] 'refined (fs)' (2Sam.22.31), whereas in a form such as ‫ָצ‬ ֭ ‫ר‬ ‫וּ‬ ‫ף‬ 'refined' (Ps.12.7), where the alveolar has a vowel sign, the resh has its default realization as an advanced uvular [sˁɔːˈʀ̟ uːuf].This reflects the fact that one of the conditions for the occurrence of apical resh is, indeed, that it occurs in the same foot as a preceding alveolar, since in [sˁɑrˁuːˈfɔː] the alveolar ‫צ‬ is not in the same phonetic syllable but is in the same foot as the following resh-it is bound to the resh by the shewa, according to the description of the medieval sources-whereas in a form such as ‫ָצ‬ ֭ ‫ר‬ ‫וּ‬ ‫ף‬ [sˁɔːˈʀ̟ uːuf] 'refined (ms)' the ‫צ‬ is in a separate foot from that of the resh.
The distribution of the allophones of resh, therefore, reflects the phonological reality of the notion of foot (maqṭaʕ) in the Masoretic sources, viz., that a consonant with vocalic shewa belongs to the foot of the segments immediately following it.
The vocalic shewa can be regarded as having the synchronic status of an epenthetic vowel.In many cases vocalic shewas in the Tiberian reading tradition occur where historically there were originally lexical vowels.These vowels were of different qualities, e.g.,  and Linguistics 12 (2020) 7-30 the maximally open vowel [a], in some circumstances modified by assimilation to its phonetic environment.As remarked, before a guttural it was realized as a short vowel with the quality of the vowel on the guttural and before yod it was realized as a short vowel with the quality of short ḥireq [i].Shortness and nonrounded vowel quality and also the copying of the quality of an adjacent vowel are characteristic features of epenthetic vowels (Hall 2011(Hall , 1581)).In examples such as those cited for the reduction of lexical vowels to epenthetic vocalic shewa, the motivation for the vowel is no longer lexical but rather phonotactic, in that it breaks illicit clusters of consonants on the phonetic level.Lexical vowels can be reduced to zero in contexts where licit sequences of consonants are the result, e.g.‫ַמ‬

‫ְל‬ ‫ֵכ‬ ‫י‬
[malˈχeː] 'kings of' (< *malaḵē), [liʃmuːˈeːel] 'to Samuel' (< *la-Šamūʾēl).According to this analysis of vocalic shewa, it would have to be assumed that the original vowel is absent at some underlying level of the phonological derivation of words and an epenthetic vowel is introduced at the phonetic surface level.A /CC/ cluster at the onset of a syllable in word-initial position is broken by an epenthetic and this can be represented thus: (3) /mqō.mō/[ma.q ̟ oː.ˈmoː] 'his place' could be explained by the analysis of the shin as extrasyllabic and not part of the onset of the syllable.Moreover, if the cluster /ʃt/ were considered an onset, this would violate the normal principle of rising sonority of syllable onsets (Ewen and Hulst 2001, 136-141, 147-150;Hoberman 1989) One could argue that the conditioning of the allophones of the resh discussed above operated within the domain of the underlying syllable and ignored epenthetics introduced at the phonetic level, operating on a word such as [sˁɑrˁuːˈfɔː] as if the ṣade and the resh were in the same syllable: (5) /sˁrū.ˈfɔ/[sˁɑ.rˁuː.ˈfɔː]'refined (fs)'  and Linguistics 12 (2020) 7-30 Since, however, the allophones of resh are themselves a phonetic phenomenon, it is more satisfactory to regard the prosodic foot on the phonetic level as the relevant domain rather than a more abstract underlying level.
In word-internal position the sequence /CCC/ is always syllabified /C.CC/, i.e. the second consonant is syllabified as an onset, and the cluster of the second and third consonants at the onset of the second syllable are split by a vocalic shewa, e.g., (6) /yiχ.tvū/[yiχ.tʰa.ˈvuː]'they (m) write' A shewa under a geminated letter with dagesh within a word was likewise vocalic, e.g.
/ham.mlɔχī.m/[ham.ma.lɔː.χiːim]'the kings' (Gen.14.17).As remarked, although vocal shewa was a vowel on the phonetic level, it would be zero at a deeper phonological level, and so phonologically equivalent to quiescent shewa.It could have been this underlying phonological level that the marking of the shewa sign in the Tiberian vocalization was originally intended to represent and it could have been for this reason that the Masoretes used the same sign for both types of shewa (Khan 1987(Khan , 1991)).
When shewa occurred within a word after a long vowel, it was generally silent, e.g.‫שׁ‬ 'it (fs.) devoured' (2 Sam.18.8) (Khan 1987, 54-55), with the result that the long vowel appears to be in a closed syllable.Likewise some long vowels appear to occur in closed syllables at the end of a word, e.g.The most compelling evidence for this analysis is the appearance of this epenthetic as pataḥ before guttural consonants (the so-called furtive pataḥ), e.g.‫֫ר‬ ‫וּ‬ ‫ַח‬ [ˈʀ̟ uː.aḥ] 'spirit' , due to assimilation to the vocalic tract configuration of the guttural (for further details see Khan 1987).Metrically a phonetic sequence /CVː.VC/ with such an epenthetic can be supposed to constitute a trochaic foot consisting of a strong + weak syllable (* .).
The presence of the epenthetic in word-medial syllables in words such as [ʃoː.om.ˈʀ̟ iː.im] is demonstrated by the fact that the first syllable can take a secondary stress in the form of a conjunctive accent, e.g.
[kʰo.ˌʕoː.otˁ.ˈjɔː]'like one wrapped' (Cant.1.7).A secondary stress cannot clash with the main stress but must in principle be separated from it by another syllable.
The underlying syllable structure of words such as  and Linguistics 12 (2020) 7-30 would have been conditioned by a constraint against superheavy syllables consisting of codas greater than two morae.Following the analysis by Kiparsky (2003) of Arabic syllable structure, we may say that such unsyllabifiable consonants are licensed by moras adjoined to the higher node of the prosodic word rather than the syllable node.Kiparsky refers to these consonants as 'semisyllables': On the phonetic level the extrasyllabic consonants were syllabified by means of epenthetics.There was a constraint against word-final short CV syllables, since such syllables had to be combined in an iambic foot with a following bimoraic syllable.So the epenthetic came before the consonant, forming a trochaic foot on the phonetic level: [ˈqoː.ol](* .).This process of epenthesis in word-final position was extended by analogy to similar syllabic configurations in wordinternal position.
The discussion above concerning the epenthetic vocalic shewa has been concerned so far with cases in which it has developed from a historical lexical vowel.Another motivation for an epenthetic vowel was to introduce an ahistorical vowel between two consonants for orthoepic purposes.This applies in particular to the frequent insertion of an epenthetic after a guttural consonant (

‫א‬ ‫ה‬ ‫ע‬ ‫ח‬
) where there was no historical lexical vowel in a sequence where the guttural originally closed a syllable in word-medial position and was in contact with a following consonant.These epenthetics are regularly written

'cleansing'
The motivation for the introduction of the epenthetic between a guttural and a following consonant was orthoepic.Gutturals were weak consonants in the reading tradition and efforts were made to ensure that they were not slurred over.Acoustically the epenthesis made the gutturals more perceptible when separated from the following consonant14 and this facilitated their preservation in the reading.
Although there is a tendency for gutturals to be followed by ḥaṭef vowels where parallel forms have silent shewa, this is not a universal rule.It is sometimes possible to identify additional phonotactic and metrical factors that appear to have conditioned the occurrence of the ḥaṭef vowels in certain forms with gutturals in contrast to other forms that have silent shewa.As shown by de Caen (2003) and Alvestad and Edzard (2009) one factor that conditions the occurrence of ḥaṭef vowels on gutturals, at least on ḥet, is sonority of the following consonant.They have shown that a ḥaṭef vowel tends to occur when the following consonant is high in sonority, e.g. in ‫ַי‬ ‫ֲח‬ ‫֣ר‬ ‫וֹ‬ ‫שׁ‬ 'he ploughs' (Hos.10.11), 14 Cf. Hall (2011Hall ( , 1577Hall ( -1578)), who discusses this function of epenthesis in languages.where the consonant is a sonorant rhotic, but exhibits a greater tendency to be omitted when the following consonant is lower in sonority, e.g.

‫ֶי‬ ‫ְח‬ ‫ַ֥דּ‬ ‫ל‬
'he ceases' (1 Sam.9.5).This is motivated by the principle that the optimal contact between two adjacent syllables is where the onset of the second syllable is stronger than the offset (coda) of the preceding syllable (Vennemann 1988, 40), and so the contact is eliminated with a following weak sonorous onset by the intervening ḥaṭef vowel.
Variations, however, occur in inflections of the same verb, where the same consonants are involved, e.g.'they consider ' (Psa. 35.20).In such cases the ḥaṭef appears to have been motivated by a metrical factor, namely the disfavouring of a rhythmic clash.This is seen in a metrical grid representation of the two forms.In these grids feet are marked in the first row.It will be assume that feet after the main stress are extrametrical (marked with angular brackets ⟨ ⟩).Evidence for this extrametricality is presented below.( 12 The grid representation displays the varying degrees of the relative prominence of syllables.These include epenthetic syllables, syllables containing a vowel without the main accent, and syllables with the main accent (represented as levels 1, 2 and 3 respectively in the grids above).It is likely that the insertion of the syllable with the ḥaṭef vowel in the second form was favoured since it created grid euphony by repairing a potential rhythmic clash caused by two syllables of the same prominence before the stress,15 as shown in the following grid: 15 A clash is the occurrence of two adjacent metrically strong elements with the same prominence.A lapse is the occurrence two adjacent metrically weak elements; cf.Prince (1983), Selkirk (1984).In the standard Tiberian tradition a shewa is in principle silent on a guttural when the syllable of the guttural receives the main accent and it is followed by another syllable, e.g.[ʃa.maː.ʕa.ˈnuː.hɔː]'we heard it ' (Psa. 123.6).These phenomena can also be explained on metrical grounds if we posit, as remarked above, that syllables after the main accent are extrametrical and unfooted.Extrametrical syllables at the right periphery of words commonly occur in iambic metrical systems (Kager 2007, 204) Here the accent on the syllable after the guttural licenses the ḥaṭef in that it can be footed and bound metrically to this strong footed syllable.Some short vowels in open syllables are lexical vowels rather than epenthetic vowels.This applies mainly to a set of vowels represented by ḥaṭef qameṣ.In such cases the ḥaṭef qameṣ [ɔ] preserves the rounded quality of a historical lexical vowel of the morphological form and there has not been quality reduction and neutralization.They should be represented as the short phoneme /o/ in the phonological form of the word, e.g.