New Perspectives in Biblical and Rabbinic Hebrew

7 Most of the papers in this volume originated as presenta� ons at the conference Biblical Hebrew and Rabbinic Hebrew: New Perspecti ves in Philology and Linguisti cs, which was held at the University of Cambridge, 8–10th July, 2019. The aim of the conference was to build bridges between various strands of research in the fi eld of Hebrew language studies that rarely meet, namely philologists working on Biblical Hebrew, philologists working on Rabbinic Hebrew and theore� cal linguists.

337). The increasing use of -ōṯ in post-biblical Hebrew is statistically significant, as has most recently been shown by Tubul (2003;. Tubul shows that unmarked nouns, 1 mostly masculine singular, select -ōṯ in Mishanic Hebrew more than in earlier stages of Hebrew, while marked feminine singular nouns show only a minor increase of the plural -īm. The question of this gender-number morphological mismatch in all phases of Hebrew is a long-standing problem; attempts to explain it have mostly been synchronic and restricted to evidence from Hebrew. Although the existence of gender-number mismatch in other Semitic languages is peripherally acknowledged by Hebraists, 2 scholars have yet to examine this phenomenon comparatively. Given the failure to provide a coherent explanation for the mismatch internal to Hebrew, it is surprising that so little attention has been paid to this phenomenon in other languages. In what follows, I would like to position the Hebrew case within a wider Semitic context and ask whether this system is especially typical of Hebrew, and, if not, what that tells us about plural formation in Hebrew. My aim is not to provide a full 1 The term 'unmarked' here designates nouns with no overt gender morpheme in the singular. 2 See already comments on the situation in Palestinian Arabic in Cohen (1930, 282). reconstruction of the Semitic system; I hope to do that in a separate study. Rather, I wish to contextualise Hebrew and point to avenues which have yet to be explored. In §2.0 I review some inner-Hebrew explanations and show them to be insufficient. In §3.0 evidence from other Semitic languages is presented. In §4.0 I offer analysis, with some implications for Hebrew. §5.0 summarises my findings and offers a conclusion.

Hebrew Plural Morphemes
The main accounts of gender-number mismatch in Hebrew are lexical, morphological, or a combination of the two. That is, they treat plural formation as a lexical feature, unconditioned by nominal morphology, as a grammatical feature, or as arising from both lexical and grammatical factors. Schwarzwald (1991) suggests that a set of synchronic rules applies to a large number of substantive categories in the lexicon, and do not follow gender assignment (the Strong Lexicalist Hypothesis), while other categories, along with adjectives and participles, are marked in the grammar, and follow gender assignment (the Weak Lexicalist Hypothesis). According to this approach, different nouns are assigned plural formation in different layers of the grammar. Even if one accepts Schwarzwald's position, it provides no explanation for the seeming randomness of plural formation in Hebrew, which results in a synchronically incoherent system. This approach is also problematic synchronically, because it is unclear how nouns move from one category to the other, especially from the one governed by the weak lexicalist hypothesis to the one governed by the strong lexicalist hypothesis.
A set of phonological, morphological, and lexical criteria was already suggested by Cohen (1930), which form the basis for most subsequent discussions of the topic. Cohen suggested that -ōṯ is common with nouns with a rounded vowel, certain patterns (miqṭēl, maqṭēl), and quadriradicals (e.g., ‫ֹלת‬ ‫כְׁ‬ ְׁ ‫ש‬ ‫אַ‬ 'clusters'); -īm is common with certain patterns (qiṭṭūl, qa ̊ṭ ūl), segolates, and nouns denoting flora (e.g., ‫ים‬ ‫נִ‬ ‫אֵ‬ ְׁ ‫ת‬ 'figs') and fauna (e.g., ‫ים‬ ‫זִ‬ ‫עִ‬ 'goats'). Tubul (2003) proposes a similar set of criteria to account for the distribution in both Biblical and Mishnaic Hebrew. She discovered that gender is not a crucial factor in determining the plural morpheme, as 50 percent of unmarked nouns select -īm while the other 50 percent select -ōṯ, regardless of gender. For unmarked nouns selecting -ōṯ, she suggests two criteria: a. Phonological: nouns with a rounded vowel in the final two syllables of the stem tend to choose -ōṯ, which is also rounded, due to ad hoc vowel harmonisation. This tendency is claimed to be stronger in Mishnaic Hebrew. 3 b. Morphological: -ōṯ is associated with certain nominal patterns. 4 Marked feminine nous which select -īm typically belong to the semantic field of flora and fauna, although other categories are also possible. Tubul (2005) and others have noted that the plural formation of the most common nouns are not morphologically predictable, while those of the least common nouns are. Tubul (2005,267) suggests that speakers learnt common words as lexical items, whereas less commonly-used words were subject to regular morphological rules.
If there are rules governing the selection of the plural suffixes, nouns which are attested with both plural morphemes present a problem. Multiple unrelated explanations have been suggested in the literature to account for this phenomenon. Unfortunately, most of them are ad hoc and do not provide a motivation for the phenomena in toto. Tubul (2007) suggests that plural morphemes are grammatically determined. According to this hypothesis, different morphemes are used in different states; i.e., one suffix can be used for the absolute, while another is used for construct. These, however, are internally inconsistent (see Table 2); nouns may use different plural morphemes in different states, but the choice of morpheme for each state is unpredictable, so the assumption that plural morphemes are grammatically determined does not hold.
Another claim is that different morphemes correlate with semantic differentiation (Tubul 2003, 195); for example, qɛḇɛr with -īm is the plural of 'grave', while qɛḇɛr with -ōṯ is 'family burial grounds'. Thus, in example (1), the Israelites refer to simple graves, while in (2) the reference is to Josiah's family burial compound.
( Semantic differentiation, therefore, cannot explain variation in the case of most lexemes; as the examples above show, even with those lexemes whose variation is partially attributable to semantic differentiation, this tendency is inconsistent. Moreover, there is no coherent account for why a specific plural morpheme is selected to mark a given semantic nuance. But the rules governing the distribution of the plural suffixes may not be purely linguistic. Within Biblical Hebrew, Tubul (2007) argues that genre is also an important factor. At least for some nouns, plural morphemes are claimed to be distributed according to genre. Specifically, in poetry, nouns tend to select -ōṯ plurals, even when in other genres they do not. Here too there is no internal consistency; for example, in prose yaʿar 'forest' appears with -īm (Ezek. 34.25; 39.10), while in poetry it appears with -ōṯ (Ps. 29.9). But the exact opposite distribution is attested for dōr 'generation', which selects -ōṯ in prose (e.g., Num. 10.8), but -īm in poetry (e.g., Ps. 72.5). Genre is, therefore, unlikely to be a significant factor.
Finally, several scholars suggest that alternative plural morphemes are used when similar plural morphemes occur in the vicinity, e.g., in the same clause, as a result of textual attraction, or 'tonal agreement' (Sharvit 1990, 338;Tubul 2007). For example, in (5) below, the noun *ʾălumma ̊ 'sheaf' selects the plural morpheme -īm when in the vicinity of another form with the same morpheme (the masculine plural participle mǝʾallǝmīm 'were gathering'), but not otherwise: 5 where the evidence is squarely against such a hypothesis. In (6) below, the noun masmēr 'nail' is adjacent to a noun with the morpheme -ōṯ (ḏalṯōṯ 'doors'), and yet selects -īm. Meanwhile, in example (7) below, the adjacent noun selects -īm (šǝqa ̊l īm 'shekels') but masmēr 'nail' selects -ōṯ. In both cases, masmēr 'nail' belongs in the same syntactic segment as nouns with a different plural morpheme, and yet there is no perceivable effect on the selection of plural morpheme.

3.9)
Describing variation as a form of attraction has, therefore, very limited applicability. Indeed, in most cases, it has no effect at all.
Explanations thus far have appealed to multiple lines of reasoning which are internally inconsistent and which provide a coherent account of neither the mismatch in gender-number nor the alternation between plural suffixes in Hebrew. A diachronic explanation, to the best of my knowledge, has not been offered.
Both gender-number mismatch and the use of alternating morphemes are attested in other Semitic languages, and a review of the evidence there may shed some light on Hebrew.

Sound Plurals in Semitic
Many Semitic languages use broken plurals for some or most of their substantives. Such a system is found in the languages of the Arabian Peninsula (Arabic, Modern South Arabian, and Epigraphic South Arabian) and Ethiopia (Gəʿəz). A strong preference for sound plurals of the type that Hebrew shows is found primarily in Northwest and East Semitic. Other branches show a much more complicated combination of broken plurals and sound plurals, with a strong preference for the former. Therefore, the most relevant sub-branches for a comparison with Hebrew are Akkadian, Aramaic, and Ugaritic, 6 although evidence from Ethiopic and Arabic is also illustrative.
I have collected all nouns that show a morphologically unexpected plural, namely a reflex of Proto-Semitic (PS) *-ūn on feminine nouns and a reflex of PS *-āt on masculine nouns. 7 The comparison is not exhaustive, because for some nouns, plural forms are not attested. This is especially true for Ugaritic, but also on occasion for Akkadian, where some plural forms are consistently written with logograms, and so remain morphologically opaque. Another snag is that we cannot truly evaluate the ratio of predictable plural morphology with mismatched gender-number morphology, since we have no access to speakers, and sometimes only limited access to texts. The reliance on information from dictionaries is not without risk, since dictionaries list all forms, regardless of their actual status in the language. Nevertheless, for a broad comparison, there is no other resource to collect evidence.
All the languages examined for this study have a significant number of substantives the plural of which is unpredictable. With the exception of Aramaic, most languages show a strong preference for the use of *-āt at the expense of other plural forms, regardless of gender; namely, the number of substantives selecting an unpredictable *-āt is far larger than that of substantives selecting an unpredictable *-ūn. In Ugaritic the ratio of unpredictable *-āt to *-ūn is 2:1, in Biblical Hebrew it is slightly over 2:1, in Mishnaic Hebrew it is 3.5:1, and in Akkadian it is 30:1. The preference in Aramaic is reversed: the ratio of unpredictable *-ūn to unpredictable *-āt in Syriac is more than 5:1. Clearly, all languages show a strong preference for one of the plural morphemes and their distribution is therefore not random. In this respect, plurals was collected from Sokoloff (2002;2003), Cook (2015), Payne Smith (1879), which are all incorporated in the online CAL.
Hebrew patterns with Akkadian and Ugaritic, but not with Aramaic.
The set of nouns with unpredictable plurals is not easily compared, since not all languages share the lexicon and the shared set of nouns which select the same unpredictable plural morpheme is small. 8 Additionally, there is some variation even in the case of the most prototypical examples of unpredictable plural formation; for example, the plural of 'father', which selects *-āt in Hebrew ʾa ̊ḇ ōṯ and Aramaic ʾabāhātā, appears to select a masculine plural in Akkadian abbû, 9 despite the fact that Akkadian shows the most notable preference for *-āt. Table 3 below includes nouns that show the same unpredictable plural across Northwest Semitic (NSW) and Akkadian. This is a very short list, and includes two loan words from Akkadian, the first originally Sumerian (kissē 'chair' < kussû; mazza ̊l 'constellation' < mazzāzu).  however, Hebrew patterns more often with Akkadian than with its closest Northwest Semitic relatives. Table 5 below includes feminine singular nouns, either marked or unmarked for gender, which select the plural morpheme *-ūn. In Hebrew and even more so in Akkadian, this is a relatively small category, as neither language has many cases where feminine singular nouns select *-ūn. The data indicates that in the case of feminine nouns with *-ūn plurals, Hebrew patterns exactly with Aramaic, a bit less with Ugaritic, and very little with Akkadian.
Tables 4 and 5 show that Biblical Hebrew has slightly more variation than Aramaic, which shows greater propensity for *-ūn, and Akkadian, which leans toward *-āt; however, Hebrew patterns with Akkadian more than expected given its phylogenetic position.
The comparative evidence provides a number of useful leads. First, Hebrew is clearly not an exceptional system and several of the phenomena assumed thus far to be Hebrew-specific and/or Hebrew-internal should be examined from a much broader perspective. Second, Hebrew often patterns with Akkadian, not, as might be expected, with Aramaic or Ugaritic. Explanations given for the distribution of the plural morphemes in Hebrew clearly cannot be based solely on internal evidence ( §2.0 above).
The assumption in the scholarly literature is that, since plural morphemes are gendered, feminine nouns will choose *-āt and masculine nouns will choose *-ūn. This assumption is true for the distribution of the plural morphemes with adjectives and participles, where the plural morphemes behave predictably. If indeed plural marker distribution on adjectives is the same as that for substantives, as is widely assumed (Hasselbach 2007

Hebrew in Context
Given the comparative evidence, it emerges that some claims There are two ways of interpreting the evidence: (1) A change from -īm to -ōṯ can be interpreted as a sign of a diachronic shift in cases in which -ōṯ appears in texts known or thought to be late, as well as post-biblical Hebrew, while -īm is restricted to early texts or Transitional Biblical Hebrew (TBH).
(2) A localised use of -īm may be an indication of Aramaic interference in cases in which -īm appears in some late texts, while -ōṯ appears in texts of various periods, including Late Biblical Hebrew (LBH) and post-biblical Hebrew. Comparison to Aramaic is instructive, if the lexeme is attested there.
This tallies with the observation made by several scholars, that during its attested history Hebrew shifts to use -ōṯ more gener- It is, therefore, warranted to assume that for this noun the plural -ōṯ in Song of Songs is a sign of lateness, while the use of -īm likely positions a text no later than the Babylonian Exile.
appearance of -īm may be indicative of an Aramaism and, therefore, of exilic or post-exilic date. One must exercise caution, however, as not all nouns that select -īm are a result of foreign interference, and an early use of the plural morpheme -īm should be clearly distinguished from a hypothesised Aramaic calque -īm. I suggest that we can verify whether -īm is a result of language contact, if there is evidence that both pre-exilic and later texts  Given the distribution of the plural -ōt in pre-exilic, exilic, and post-exilic texts, the occurrence of -īm in Zechariah is exceptional, requiring explanation. I suggest that the choice of plural morpheme is a localised Aramaic calque, which failed to be adopted in later periods. 27 A second example comes from Song of Songs, another postexilic text. The language of Song has generally been dated to the Hellenistic period based on a number of late features. 28 Dobbs-Allsopp (2005) proposes that the plural of mɛḡɛḏ 'precious thing' 26 Sharvit (1990, 370) notes a few instances of pinnīm 'corners' in Mishnaic manuscripts, but they are rather likely related to pa ̊n īm 'face'. 27 Neither lexical item is attested in Persian period Aramaic texts. In Late Antique dialects pnītā takes -ātā (e.g., in Syriac), although the lexeme is likely derived from a different root (Aram. √pny versus Hebrew √pnn). The lexeme parsā is also attested in later texts, where its plural morpheme varies. JPA has parsātā, while Syriac uses parsānē as well. 28 The list of LBH features in Song includes orthography (the plene spelling of da ̊w īḏ 'David'), morphology (e.g., preference for the 1cs pronoun ʾănī), syntax (the decline of narrative wayyiqṭol), and lexicon (preference for ḥēḵ for mouth instead of the early pē). See Dobbs-Allsopp (2005) for a fairly thorough list and discussion. there, mǝḡa ̊ḏ īm (Song 4.13,16;7.14), is a late feature, on the basis of the discussion in Polzin (1976), who argued that preference for nominal pluralisation is a sign of lateness. The exact dating of Gen. 24, ex. (22), is difficult. Rofé (1990) has suggested that it is a Persian period text, based on its language. Rendsburg (2002) argues on the basis of the same features that it is not a late text, but rather a Northern text, which per Rendsburg includes multiple instances of Aramaisms, as a form of style switching. This analysis includes the lexeme mɛḡɛḏ, which Rendsburg argues is 'Aramaic-like' and Rofé suggests is late. 30 Hendel 29 See also Ezra 1.6 and 2 Chron. 32.23. 30 Akkadian *magattu~magadātu is a possible cognate. Von Soden (1966,16) reads magāṭātu and suggests that it is a loanword from Aramaic, whose root he suggests is √ghṭ (see Sokoloff 2002, 362). CAD assumes the root is √mgd. The lexeme is still listed as an Aramaic loan in and Joosten (2019, 81-82), however, note that despite the late features and signs of contact with Aramaic embedded in Gen. 24, the text also exhibits several CBH features which are generally absent from LBH. They, therefore, suggest that the text belongs to TBH.
What has not previously been noted in the context of this discussion is the form of the plural. Song of Songs pluralises mɛḡɛḏ with -īm, while all other biblical texts, whatever their date, use -ōṯ. Following the criteria I set above, the plural with -īm in a single text is exceptional, and since the language clearly did not stop using -ōṯ to pluralise this substantive, it may be a calque of the Aramaic plural. Indeed, all dialects of Aramaic where the lexeme is attested pluralise mɛḡɛḏ with a reflex of *-ūn: JPA maḡdānīn, Targum maḡdānīn, Syr. maḡdunē (or meḡdunē). The plural of mɛḡɛḏ in Song is, therefore, likely an Aramaic calque, which lends support to the LBH dating of the text. Some other possible cases are the plural of kinnōr 'lyre'. It generally selects -ōṯ in pre-exilic (2 Sam. 6.5) and post-exilic (Neh. 12.27) biblical texts. This is also true in Mishnaic Hebrew (e.g., m. Sukkot 5.4). The only exception is Ezek. 26.13, where it appears as kinnōrayiḵ 'your (fs) lyres'. In Aramaic, this noun also selects *-ūn, for example JPA kīnārīn, Syriac kēnārē. The plural of CAD although no Aramaic origin is noted. Abraham and Sokoloff (2011, 40) reject Sokoloff's etymology, but suggest no alternative. Given the context of the lexeme in Akkadian, that is, a dowry list, it is possible that it is a cognate of the Aramaic and Hebrew lexeme. Its attestation in Neo-Babylonian roughly fits the date of the first instance of the lexeme in Hebrew, if Gen. 24 is viewed as a TBH text. The evidence is, however, too scant for a clear conclusion.
ša ̊ḇ ūaʿ 'week' is typically pluralised with -ōṯ in both pre-exilic (Exod. 34.22) and post-exilic (2 Chron. 8.13) biblical texts. The same plural is found in Mishnaic Hebrew (e.g., Megilla 3.5). The only exception is Daniel, where all occurrences of the plural of ša ̊ḇ ūaʿ are ša ̊ḇ ūʿīm. The plural of this noun in Aramaic is similarly a reflex of *-ūn, e.g., JPA and CPA šābuʿīn, Syriac šabuʿē. Caution should be exercised, however, in evaluating such cases, and evidence from Aramaic and post-biblical Hebrew is crucial to establishing the direction of change.

Summary and Conclusions
This paper has dealt with the age-old problem of mismatched plurals in Hebrew. I have attempted to show that the study of such phenomena purely on the basis of Hebrew is misleading. In a comparative perspective, the distribution of plural morphemes in Hebrew is not exceptional and has parallels in other Semitic languages. The comparative evidence as well as documented diachronic shifts in Hebrew can be used for linguistic dating. The main methodological conclusion I hope readers draw from this paper is that the study of the structure of Hebrew must not be undertaken in isolation from evidence in other languages; in most cases comparative evidence is invaluable.