[Mova](https://md.mova.elamur.ai/free/en/) · [this page in HTML](https://mova.elamur.ai/free/en/docs/languages/hbo/03-phonology/)

[home](https://md.mova.elamur.ai/free/en/) / [docs](https://md.mova.elamur.ai/free/en/docs/) / [languages](https://md.mova.elamur.ai/free/en/docs/languages/) / [Ancient Hebrew](https://md.mova.elamur.ai/free/en/docs/languages/hbo/) / Phonology of Biblical Hebrew

# Phonology of Biblical Hebrew

**On this page**

1. [Contents](#contents)
2. [1. Three layers of "pronunciation"](#1-three-layers-of-pronunciation)
3. [2. The consonants](#2-the-consonants)
4. [3. Begadkefat: stops and fricatives](#3-begadkefat-stops-and-fricatives)
5. [4. Gutturals and resh](#4-gutturals-and-resh)
6. [5. The vowels](#5-the-vowels)
7. [6. Shewa and the hatef vowels](#6-shewa-and-the-hatef-vowels)
8. [7. Syllables](#7-syllables)
9. [8. Stress, pause and maqqef](#8-stress-pause-and-maqqef)
10. [9. Vowel change: reduction and lengthening](#9-vowel-change-reduction-and-lengthening)
11. [10. Assimilation and sandhi](#10-assimilation-and-sandhi)
12. [11. The weak letters א ה ו י](#11-the-weak-letters-א-ה-ו-י)
13. [12. Historical sound changes](#12-historical-sound-changes)
14. [13. Other reading traditions](#13-other-reading-traditions)
15. [14. In Mova](#14-in-mova)
16. [15. Sources](#15-sources)

This chapter describes the sounds of Biblical Hebrew as the Tiberian Masoretes wrote them down (the pointed text of our Torah trees), the rules that tie the written signs to sounds, and the older sound changes that explain why the forms look as they do. It separates three layers that grammars often mix: the **reconstructed Tiberian reading** (how the Masoretes of Tiberias actually pronounced the text in the 9th–10th centuries CE, Khan 2020), the **school pronunciation** of the grammars of the 19th century (GKC, Davidson), with its system of long and short vowels, and the **pre-Masoretic history** of the language (proto-Semitic mergers, the Canaanite shift). Every rule is given with Torah examples from our MACULA trees, and the last sections say what Mova's tools see of it and what they do not.

## Contents

1. [Three layers of "pronunciation"](#1-three-layers-of-pronunciation)
2. [The consonants](#2-the-consonants)
3. [Begadkefat: stops and fricatives](#3-begadkefat-stops-and-fricatives)
4. [Gutturals and resh](#4-gutturals-and-resh)
5. [The vowels](#5-the-vowels)
6. [Shewa and the hatef vowels](#6-shewa-and-the-hatef-vowels)
7. [Syllables](#7-syllables)
8. [Stress, pause and maqqef](#8-stress-pause-and-maqqef)
9. [Vowel change: reduction and lengthening](#9-vowel-change-reduction-and-lengthening)
10. [Assimilation and sandhi](#10-assimilation-and-sandhi)
11. [The weak letters א ה ו י](#11-the-weak-letters-%D7%90-%D7%94-%D7%95-%D7%99)
12. [Historical sound changes](#12-historical-sound-changes)
13. [Other reading traditions](#13-other-reading-traditions)
14. [In Mova](#14-in-mova)
15. [Sources](#15-sources)

Conventions. Transliteration follows the SBL academic style (see the README): ʾ b/ḇ g/ḡ d/ḏ h w z ḥ ṭ y k/ḵ l m n s ʿ p/p̄ ṣ q r ś š t/ṯ; vowels a ā e ē ê i î o ō ô u û, vocal shewa ə, hatefs ă ĕ ŏ. Phonetic transcriptions in square brackets [ ] are IPA and follow Khan 2020. Examples carry the verse and the sentence id of our trees (`wlc:BOOK:n`); the Hebrew is printed without the cantillation marks, the English is JPS 1917. Counts are from our Torah trees on 2026-10-17.

---

## 1. Three layers of "pronunciation"

The consonantal text of the Bible is much older than its vowel signs. The signs for vowels, gemination, spirantization and accents were created by the Masoretes of Tiberias between about the 7th and 10th centuries CE to fix an oral reading tradition that had been handed down with the text (Khan 2020, vol. 1, §I.0.4; GKC §7). Three layers must be kept apart:

| layer | what it is | where it comes from |
| --- | --- | --- |
| Tiberian reading (reconstructed) | the pronunciation the Masoretes intended, e.g. qamets = [ɔ], resh with two allophones, *bet rafe* = [v] | medieval treatises (Hidāyat al-Qāriʾ), Karaite Arabic transcriptions; Khan 2020 |
| school pronunciation | the European academic reading, with "long" and "short" vowels, qamets gadol [aː] vs qamets hatuf [o] | Kimḥi's medieval grammar → GKC 1910, Davidson 1916 |
| historical phonology | what the forms come from: proto-Semitic *ḫ, *ġ, *ś; case endings; *ā > ō | comparative Semitics; GKC §§1–2, §§18–29; Suchard 2019 |

Khan states the problem plainly: the textbooks, "even in its earliest stages in eleventh-century Spain, did not have direct access to the way the Tiberian Masoretes were pronouncing Biblical Hebrew" (Khan 2020, vol. 1, Preface, p. xiii). Most rules of this chapter work in both the Tiberian and the school description, because they are rules about the *signs*. Where the two differ (vowel length, the value of qamets, of resh, of sin), the difference is said.

---

## 2. The consonants

Hebrew writes 22 consonant letters (five with a final form: ך ם ן ף ץ). In the Tiberian reading they stand for 23 consonant phonemes plus the fricative allophones of the six *begadkefat* letters. The table gives the letter, its transliteration, the school value, and the Tiberian value reconstructed by Khan.

| letter | name | translit. | school value | Tiberian (Khan 2020, §I.5.1) | notes |
| --- | --- | --- | --- | --- | --- |
| א | ʾalef | ʾ | glottal stop / silent | [ʔ] | often quiescent: §11 |
| בּ / ב | bet | b / ḇ | b / v | [b] / [v] | begadkefat |
| גּ / ג | gimel | g / ḡ | g / g (gh) | [g] / [ʁ] | spirant = voiced uvular fricative |
| דּ / ד | dalet | d / ḏ | d / dh | [d] / [ð] |  |
| ה | he | h | h | [h] | silent as final mater: §11 |
| ו | vav | w | w / v | [v], [w] | [w] e.g. before furtive patah |
| ז | zayin | z | z | [z] |  |
| ח | ḥet | ḥ | ḥ (ch) | [ħ] | merged with *ḫ: §12 |
| ט | ṭet | ṭ | ṭ | [tˁ] | "emphatic", pharyngealized |
| י | yod | y | y | [j], [ɟ] when doubled |  |
| כּ / כ ך | kaf | k / ḵ | k / kh | [kʰ] / [χ] |  |
| ל | lamed | l | l | [l] |  |
| מ ם | mem | m | m | [m] |  |
| נ ן | nun | n | n | [n] | assimilates: §10 |
| ס | samekh | s | s | [s] |  |
| ע | ʿayin | ʿ | ʿ | [ʕ] | merged with *ġ: §12 |
| פּ / פ ף | pe | p / p̄ | p / f | [pʰ] / [f] |  |
| צ ץ | ṣade | ṣ | ṣ (ts) | [sˁ] |  |
| ק | qof | q | q | [q] |  |
| ר | resh | r | r | [ʀ̟] and [rˁ] (allophones) | behaves like a guttural: §4 |
| שׂ | sin | ś | s | [s] | same as samekh in the reading |
| שׁ | shin | š | sh | [ʃ] |  |
| תּ / ת | tav | t / ṯ | t / th | [tʰ] / [θ] |  |

Classed by place and manner (school phonemes; the emphatics ṭ ṣ q are pharyngealized or uvular):

|  | labial | dental/alveolar | palatal | velar | uvular | pharyngeal | glottal |
| --- | --- | --- | --- | --- | --- | --- | --- |
| voiceless stop | p | t, ṭ |  | k | q |  | ʾ |
| voiced stop | b | d |  | g |  |  |  |
| voiceless fricative | (p̄) | (ṯ), s, ś, ṣ, š |  | (ḵ) |  | ḥ | h |
| voiced fricative | (ḇ) | (ḏ), z |  | (ḡ) |  | ʿ |  |
| nasal | m | n |  |  |  |  |  |
| liquid |  | l, r |  |  |  |  |  |
| glide | w |  | y |  |  |  |  |

(The fricatives in brackets are the allophones of §3, not independent phonemes in the Biblical period.)

**The letter groups of Sefer Yetzirah.** The medieval tradition classified the letters by the organs that make them. Sefer Yetzirah (ch. 2) gives the five places of the mouth — throat אחה"ע, palate גיכ"ק, tongue דטלנ"ת, teeth זסשר"ץ, lips בומ"ף — and divides the letters into three "mothers" (אמ"ש), seven "doubles" (בג"ד כפר"ת) and twelve "simple" letters. The doubles are exactly the begadkefat letters plus resh: the book calls them doubles because each has a hard and a soft pronunciation, and its inclusion of resh agrees with the Tiberian evidence that resh, too, had two realizations (Khan 2020, vol. 1, §I.1.20; §I.0.13.5 on the commentaries on Sefer Yetzirah as a source). The crate `yetzirah` encodes these groups (see §14).

**Final forms.** כ מ נ פ צ take a special shape at the end of a word: ך ם ן ף ץ (GKC §5b). The final kaf normally carries a silent shewa (לְךָ has qamets instead: *ləḵā*). Final forms carry no difference in sound.

---

## 3. Begadkefat: stops and fricatives

Six letters — ב ג ד כ פ ת, remembered as *bəḡaḏ kəp̄aṯ* — have two pronunciations: a stop, marked with a dot inside the letter (*dagesh lene*), and a fricative, without it (GKC §6n, §13, §21; Khan 2020, §I.1.25, §I.3.1.10).

**The rule.** A begadkefat letter is a fricative when a vowel (full or vocal shewa) immediately precedes it, in the same word or across a close word boundary; otherwise it is a stop (GKC §21a–b). So:

- **stop** at the beginning of a word after a pause or after a word ending in a consonant, and after a silent shewa (a closed syllable);
- **fricative** after any vowel, including vocal shewa and the hatef vowels;
- **doubled** letters are always stops (the *dagesh forte* includes the stop dot), GKC §21c.

| environment | example | reading |
| --- | --- | --- |
| start of verse | בְּרֵאשִׁית בָּרָא (GEN.1.1) | *bərēʾšîṯ bārāʾ* — both stops: the first after silence, the second after a closed syllable (-*šîṯ*) |
| after a vowel inside the word | דְּבָרִים (EXO.20.1) | *dəḇārîm* — ב fricative after the vocal shewa |
| after a "medial" shewa | וַיְדַבֵּר (EXO.20.1) | *wayḏabbēr* — ד fricative, see below |
| after a vowel across words | יְהִי־כֵן | *yəhî-ḵēn* — כ fricative after *-î* joined by maqqef |
| after a closed syllable | מִדְבָּר | *miḏbār* — ב stop after the silent shewa of ד |
| doubled | הַשָּׁמַיִם, יְדַבֵּר | doubled consonant, always a stop |

The case of וַיְדַבֵּר "and he spoke" (EXO.20.1, `wlc:EXO:519`) shows the role of the shewa. The wayyiqtol prefix normally doubles the next consonant (וַיִּקְרָא), but a yod with shewa drops the doubling (GKC §20m). The shewa under י now follows a short vowel and would be "silent" by the school rule, yet the ד after it is spirantized — *wayḏabbēr*. The school grammar calls such a shewa "medial" (GKC §10d): it stands where a doubled consonant once was, and it behaves as vocal for the begadkefat rule.

> **GEN.1.1** · `wlc:GEN:0` בְּרֵאשִׁית בָּרָא אֱלֹהִים אֵת הַשָּׁמַיִם וְאֵת הָאָרֶץ *bərēʾšîṯ bārāʾ ʾĕlōhîm ʾēṯ haššāmayim wəʾēṯ hāʾāreṣ* — in-beginning created God OBJ the-heavens and-OBJ the-earth — "In the beginning God created the heaven and the earth." (JPS 1917)

Here ב of בְּרֵאשִׁית is a stop (start of the text); the ב of בָּרָא is a stop because the previous word ends in a consonant (*-ṯ*); the final ת of רֵאשִׁית and of אֵת is a fricative [θ] because a vowel precedes it. Khan's reconstruction of this verse is [baʀ̟eːˈʃiːiθ bɔːˈʀ̟ɔː ʔɛloːˈhiːim ˈʔeːeθ haʃʃɔːˈmaːjim veˈʔeːeθ hɔːˈʔɔːʀ̟ɛsˁ] (Khan 2020, vol. 1, §I.5.4.1, p. 615) — note the vocal shewa read as [a], the long vowels everywhere stressed, the epenthetic echo vowel in [ʔeːeθ], qamets as [ɔ], and the ṣade as [sˁ].

**Across word boundaries.** A begadkefat letter at the start of a word is spirantized when the previous word ends in a vowel and the two words are joined by a conjunctive accent or maqqef (GKC §21b). The rule is suspended after a disjunctive accent — so the accents (chapter on cantillation) decide the phonology of word junctures:

> **GEN.1.2** · `wlc:GEN:1` וְהָאָרֶץ הָיְתָה תֹהוּ וָבֹהוּ *wəhāʾāreṣ hāyəṯâ ṯōhû wāḇōhû* — and-the-earth was formless and-void — "Now the earth was unformed and void" (JPS 1917)

Here תֹהוּ (Strong 8414) begins with a fricative [θ] (the printed text has no dagesh: the previous word ends in the vowel *-â* and is joined by a conjunctive accent), while the lemma, quoted out of context, is pointed תֹּהוּ with a stop. Khan's transcription accordingly has [hɔːɔjˈθɔː ˈθoːhuː vɔːˈvoːhuː] (vol. 1, p. 615). The dictionary key of the word (`8414` in torah-en.tsv) uses the lemma with dagesh, the trees keep the form without it — one of several places where FORM and LEMMA of the same word differ only by a dot.

**Dagesh lene vs dagesh forte.** The same dot marks two things: the stop of a begadkefat letter (lene) and the doubling of a consonant (forte). A dot in a begadkefat letter after a vowel can only be forte (the letter is doubled); after a consonant or at the start of a word it can only be lene (GKC §12, §13). In הַשָּׁמַיִם the dot in ש is forte (the doubling that follows the article); in בְּרֵאשִׁית the dot in ב is lene.

---

## 4. Gutturals and resh

The four gutturals א ה ח ע and, in part, resh ר share three behaviours (GKC §22; Khan 2020, §I.2.5.4):

**(a) They are not doubled.** Where the morphology demands a doubled consonant (after the article, in the D-stem, in the wayyiqtol prefix, after מִן), a guttural or resh cannot take dagesh forte. The language then either lengthens the preceding vowel (**compensatory lengthening**) or keeps the short vowel as if the doubling were still there (**virtual doubling**, "dagesh forte implicitum") (GKC §22c).

| morphology | strong consonant | before א ע ר (usually lengthening) | before ה ח (usually virtual doubling) |
| --- | --- | --- | --- |
| article *ha* + doubling | הַשָּׁמַיִם *haššāmayim* | הָאָרֶץ *hāʾāreṣ*, הָעָם *hāʿām*, הָרָעָב *hārāʿāḇ* | הַחֹשֶׁךְ *haḥōšeḵ*, הַהוּא *hahûʾ* |
| Piel perfect | דִּבֶּר *dibber* | בֵּרַךְ *bēraḵ*, מֵאֵן *mēʾēn* | נִחַם *niḥam* (Piel), טִהַר *ṭihar* |
| wayyiqtol 1cs | וַיִּקַּח | וָאֹמַר *wāʾōmar* | — |

> **GEN.1.4** · `wlc:GEN:3` וַיַּרְא אֱלֹהִים אֶת־הָאוֹר כִּי־טוֹב וַיַּבְדֵּל אֱלֹהִים בֵּין הָאוֹר וּבֵין הַחֹשֶׁךְ *wayyarʾ ʾĕlōhîm ʾeṯ-hāʾôr kî-ṭôḇ wayyaḇdēl ʾĕlōhîm bên hāʾôr ûḇên haḥōšeḵ* — and-saw God OBJ-the-light that-good and-divided God between the-light and-between the-darkness — "And God saw the light, that it was good; and God divided the light from the darkness." (JPS 1917)

Both treatments are in one verse: הָאוֹר (lengthening before א) and הַחֹשֶׁךְ (virtual doubling before ח).

In our Torah trees the article token carries the form הַ 4 793 times (with accent variants more), הָ 1 827 times and הֶ 69 times; the vowel is chosen by the next letter. A further 1 679 articles have an empty FORM: they are absorbed into a preceding preposition (לָאוֹר, בָּאָרֶץ), see §10 and chapter 04.

**(b) They prefer a-vowels.** Before or after a guttural, a-class vowels replace i- or u-class ones (GKC §22d). Examples: the Qal yiqtol of a III-guttural verb has patah (יִשְׁמַע *yišmaʿ*, not *yišmōʿ*); the segolate נַעַר *naʿar* "lad" (not *neʿer*); יַעֲמֹד *yaʿămōḏ* "he stands" (prefix vowel a before a guttural).

**(c) Under a guttural, shewa becomes a hatef.** A guttural cannot easily carry a simple vocal shewa; it takes a "hurried" vowel: ֲ hatef patah, ֱ hatef segol, ֳ hatef qamets (GKC §10f, §22l). אֱלֹהִים *ʾĕlōhîm*, אֲנִי *ʾănî*, חֳלִי *ḥŏlî*, עֲמֹרָה *ʿămōrâ*. A silent shewa may stay under a guttural at the end of a closed syllable (יַחְדָּו *yaḥdāw*), but often a hatef replaces it there too, matching the preceding vowel (יַעֲמֹד, נֶעֱלַם) (GKC §22i; Khan 2020, §I.2.5.4).

**(d) Furtive patah.** A word-final ח, ע or consonantal ה (with mappiq) after a vowel other than a cannot be pronounced straight after that vowel; a short [a] glide is inserted before it, written as a patah under the guttural but read *before* it (GKC §22f): רוּחַ *rûaḥ*, נֹחַ *nōaḥ*, מִזְבֵּחַ *mizbēaḥ*, גָּבֹהַּ *gāḇōah*. Khan's reading of GEN.1.2 has [vaˈʀ̟uːwaħ] for וְרוּחַ, with a [w] glide between the long u and the furtive vowel (vol. 1, p. 615, n. 13).

> **GEN.6.10** · `wlc:GEN:147` וַיּוֹלֶד נֹחַ שְׁלֹשָׁה בָנִים אֶת־שֵׁם אֶת־חָם וְאֶת־יָפֶת *wayyôleḏ nōaḥ šəlōšâ ḇānîm ʾeṯ-šēm ʾeṯ-ḥām wəʾeṯ-yāp̄eṯ* — and-begot Noah three sons OBJ-Shem OBJ-Ham and-OBJ-Japheth — "And Noah begot three sons, Shem, Ham, and Japheth." (JPS 1917)

**Resh** shares (a): it is almost never doubled and takes compensatory lengthening (בֵּרַךְ, הָרָעָב). In the Tiberian reading it had two allophones, an uvular/advanced [ʀ̟] in most places and a pharyngealized alveolar [rˁ] next to dentals and before or after certain consonants (Khan 2020, §I.1.20). Suchard (2019, §2.1) lists a backing of *r* among the changes shared by Biblical Hebrew and Biblical Aramaic, and so places it inside the common reading tradition.

**ʾalef and he** are also "weak": they often lose their consonantal value and leave only the vowel (quiescence; §11).

---

## 5. The vowels

### 5.1 The signs

The Tiberian system has seven full vowel signs plus shewa and the three hatefs (GKC §8; Khan 2020, §I.2.1, §I.5.2):

| sign (on ב) | name | Tiberian value (Khan) | school value | typical length in the school grammars |
| --- | --- | --- | --- | --- |
| בַ | patah | [a] | a | short; long when stressed in pause |
| בָ | qamets | [ɔ] | ā (gadol) / o (hatuf) | long (gadol) or short (hatuf) |
| בֶ | segol | [ɛ] | e | short (and long in some forms) |
| בֵ | tsere | [eː] | ē | long |
| בִ | hireq | [i] | i / î | short (î with yod) |
| בֹ / בוֹ | holem | [oː] | ō / ô | long |
| בֻ / בוּ | qibbuts / shureq | [u] | u / û | short (qibbuts) / long (shureq) |

The Tiberian system therefore records **seven qualities** — [a ɔ ɛ e i o u] — and not length. Length was a matter of stress and syllable: in the Tiberian reading every stressed vowel and every vowel of an open unstressed syllable was pronounced long, and vowels in closed unstressed syllables were short (Khan 2020, §I.2.2.1). Qamets is one quality, [ɔ]; it is long in *qāṭal* [qɔːˈtˁaːl] and short in the closed unstressed syllable of כָּל־ [kʰɔl] or חָכְמָה [ħɔχˈmɔː] (Khan 2020, §I.2.1.2–4).

The school grammars instead derive two "qametses" — qamets gadol (ā, from proto-Semitic *a lengthened) and qamets hatuf (o, from *u in a closed unstressed syllable) — and teach rules to tell them apart (GKC §9u, §10i). Both descriptions predict the same spelling; the difference is only in what the sign is said to stand for. This manual transliterates in the school way (ā vs o), which is what MACULA's English glosses, BDB and GKC also assume.

### 5.2 Vowel classes

The school grammars group the vowels into three historical classes, because a vowel changes within its class when the stress moves (GKC §7a, §9, §27):

| class | short | tone-long / lengthened | reduced (shewa / hatef) | unchangeable (with mater) |
| --- | --- | --- | --- | --- |
| a | a (patah) | ā (qamets) | ə / ă | — (original *ā* > ô, see §12) |
| i | i (hireq), e (segol) | ē (tsere) | ə / ĕ | î (hireq-yod), ê (tsere-yod) |
| u | u (qibbuts), o (qamets hatuf) | ō (holem) | ə / ŏ | û (shureq), ô (holem-vav) |

**Changeable** vowels (a, ā, i, ē, u, ō without a mater) lengthen in an open syllable before the stress and reduce two syllables before it; **unchangeable** vowels — those written with a vowel letter (*plene*) and those that come from an old long vowel or a contraction — never reduce (GKC §25). That is why דָּבָר "word" reduces its first vowel in the plural (דְּבָרִים) but כּוֹכָב "star" keeps its ô (כּוֹכָבִים; GEN.1.16: וְאֵת הַכּוֹכָבִים).

### 5.3 Matres lectionis

Before vowel signs existed, the letters ה ו י (and א) were used to mark some vowels — mostly long and final ones (GKC §7b–g; see chapter 01). In the pointed text a mater combines with a vowel sign: וּ = û, וֹ = ô, ִי = î, ֵי = ê, ֶי = e/ê, final ָה = â. Spelling with or without the mater (*plene* vs *defective*) varies freely in the Torah: אֲבוֹת and אֲבֹת "fathers" both occur (in our trees the lemma אָב appears as אֲבֹת 48 times and as אֲבוֹת 7 times). Such pairs have one lemma in the trees but are different FORM strings; Mova's comparison of lemmas with other sources (`macula compare`) compares consonants without vowels, but plene/defective spelling still differs in consonants — see §14.

---

## 6. Shewa and the hatef vowels

The sign ְ (shewa) has two values (GKC §10; Khan 2020, §I.2.5):

- **silent shewa** (*šəwāʾ nāḥ*): marks the end of a closed syllable, no vowel — יִשְׁ־מֹר *yišmōr*;
- **vocal shewa** (*šəwāʾ nāʿ*): a very short vowel — בְּרֵאשִׁית *bərēʾšîṯ*.

The school rules for telling them apart (GKC §10a–d):

1. Shewa under the first consonant of a word is vocal: בְּרִית, שְׁמַע.
2. Of two shewas in a row in the middle of a word, the first is silent and the second vocal: יִקְטְלוּ *yiqṭəlû*.
3. Shewa under a doubled consonant is vocal: הַמְּאֹרֹת *hamməʾōrōṯ* (GEN.1.16).
4. Shewa after a long vowel (in an unstressed open syllable) is vocal: כֹּתְבִים *kōṯəḇîm*; after a short vowel it is silent: מַלְכִּי *malkî*.
5. At the end of a word shewa is silent (and normally not written, except in final kaf and after a two-consonant final cluster, אַתְּ, וַיֵּשְׁתְּ).

**How was it pronounced?** In the Tiberian reading the vocal shewa had a default quality [a], short; before a guttural it echoed the vowel after the guttural, and before yod it was [i] (Khan 2020, §I.5.3). Hence [baʀ̟eːˈʃiːiθ] for בְּרֵאשִׁית and [voˈħoːʃɛχ] for וְחֹשֶׁךְ (GEN.1.2; Khan 2020, p. 615). The common modern reading [ə] or [e] is not Tiberian.

**Hatef vowels** are written on gutturals instead of vocal shewa (§4) and sometimes on non-gutturals to mark that the shewa is vocal (e.g. under a consonant that follows a long vowel or that is a former double: הַלֲלוּ) (Khan 2020, §I.2.5.5; GKC §10h). Hatef qamets also appears as a lexical vowel (קֳדָשִׁים *qŏḏāšîm*, plural of קֹדֶשׁ, EXO.26.33).

---

## 7. Syllables

### 7.1 Shapes

A Biblical Hebrew syllable always begins with **one** consonant (GKC §26a). There are no vowel-initial syllables; the conjunction וּ "and" is the one apparent exception (GKC §26a, §104e). There are no initial clusters: a cluster at the start of a word is broken by a vocal shewa (בְּרִית *bərîṯ*). Syllables are:

| type | shape | example |
| --- | --- | --- |
| open | CV | בָּ־רָא *bā-rāʾ* |
| closed | CVC | מִדְ־בָּר *miḏ-bār* |
| doubly closed (word-final only) | CVCC | אַתְּ *ʾatt*, וַיֵּשְׁתְּ *wayyēšt*; segolates avoid it, see §12 |

The vocal shewa is not a syllable in the school analysis but a "half-syllable" attached to the following syllable (GKC §26m); Khan analyses it as an epenthetic vowel with its own short syllable in the Tiberian reading (Khan 2020, §I.2.5.2).

### 7.2 Quantity rules of the school grammar

GKC §26e–o states the "laws of the syllable" that the school grammars rely on to tell long from short and qamets from qamets hatuf:

1. An **open unstressed** syllable has a long vowel (or a vocal shewa): *dā-ḇār*. (Exception: a short vowel before a vocal shewa or a hatef, "half-open" syllable: יַעֲ־מֹד.)
2. A **closed unstressed** syllable has a short vowel: *miḏ-bār*, *kol-* (כָּל־ read *kol*).
3. A **stressed** syllable may be open or closed, long or short: *dā-ḇā́r*, *qā-ṭál*.

The Tiberian evidence agrees with rules 1 and 2 as rules of phonetic length, not of phonemic quality (Khan 2020, §I.2.2).

### 7.3 The rule of shewa

Two vocal shewas cannot follow each other at the start of a word. When a prefix with shewa (בְּ, כְּ, לְ, וְ) comes before a word that begins with shewa, the first becomes **i** and the second becomes silent (GKC §28a, §102d):

| expected | actual | gloss |
| --- | --- | --- |
| *bə-bərîṯ | בִּבְרִית *biḇrîṯ* | in a covenant |
| *lə-pənê | לִפְנֵי *lip̄nê* | before (lit. to the face of) |
| *wə-yəhî | וִיהִי *wîhî* | and let it be |
| *bə-yəmê | בִּימֵי *bîmê* | in the days of |

Before a hatef, the prefix takes the corresponding short vowel: לַאֲדֹנִי, בֶּאֱמֶת, כַּאֲשֶׁר (GKC §102d). Before yod with shewa, the hireq of the prefix and the yod merge into î (וִיהִי).

In our data, GEN.17.1 has וְהְיֵה … pointed וֶהְיֵה *wehyēh* "and be": before ה with hatef segol, the conjunction takes the matching segol.

> **GEN.17.1** · `wlc:GEN:398` … הִתְהַלֵּךְ לְפָנַי וֶהְיֵה תָמִים *hiṯhallēḵ ləp̄ānay wehyēh ṯāmîm* — walk before-me and-be blameless — "walk before Me, and be thou wholehearted." (JPS 1917)

---

## 8. Stress, pause and maqqef

### 8.1 Where the stress falls

Most words are stressed on the **last** syllable (*milraʿ*): דָּבָ֫ר, קָטַ֫ל, הַשָּׁמַ֫יִם is an exception because -*ayim* is a segolate-like ending. Words stressed on the **penultimate** (*milʿel*) are (GKC §15c, §29):

- the segolate nouns: מֶ֫לֶךְ *méleḵ*, סֵ֫פֶר *sḗp̄er*, קֹ֫דֶשׁ *qṓḏeš*, נַ֫עַר *náʿar*;
- the dual and some plural endings: עֵינַ֫יִם, מַ֫יִם, שָׁמַ֫יִם;
- forms with the directional ending ־ָה: אַ֫רְצָה, הָהָ֫רָה (the ending is unstressed, GKC §90c);
- verb forms with some suffixes: קָטַ֫לְתִּי, קָטַ֫לְתָּ (perfect 1cs, 2ms);
- wayyiqtol forms of hollow and III-he verbs, where the stress retracts: וַיָּ֫קָם, וַיִּ֫בֶן.

**Stress shift as grammar.** The stress distinguishes some forms. The perfect with *waw*-consecutive (weqatal) moves the stress of the 1cs and 2ms forms to the end: קָטַ֫לְתָּ "you killed" but וְקָטַלְתָּ֫ "and you will kill" (GKC §49h). The wayyiqtol often draws the stress back: יָקוּם → וַיָּ֫קָם (GKC §49d). The Hebrew text of our trees keeps the accents in FORM, so the stress position is recoverable from the accent sign (the accent stands on the stressed syllable, except prepositive and postpositive accents; chapter 02).

> **DEU.6.5** · `wlc:DEU:198` וְאָהַבְתָּ אֵת יְהוָה אֱלֹהֶיךָ בְּכָל־לְבָבְךָ וּבְכָל־נַפְשְׁךָ וּבְכָל־מְאֹדֶךָ *wəʾāhaḇtā ʾēṯ YHWH ʾĕlōheḵā bəḵol-ləḇāḇəḵā ûḇəḵol-nap̄šəḵā ûḇəḵol-məʾōḏeḵā* — and-you-shall-love OBJ YHWH your-God with-all-your-heart and-with-all-your-soul and-with-all-your-might — "And thou shalt love the LORD thy God with all thy heart, and with all thy soul, and with all thy might." (JPS 1917)

In this verse the weqatal וְאָהַבְתָּ is accented in the text on its last syllable (FORM `אָ֣הַבְתָּ֔`, with the disjunctive accent zaqef on תָּ), while the perfect אָהַבְתָּ alone is stressed on *-háḇ-*. MACULA marks it `Type=weqatal`.

**Secondary stress** is marked by *meteg* (gaʿya), a short vertical stroke to the left of a vowel; it also marks that a vowel before a vocal shewa is long (and so a qamets there is gadol, not hatuf): חָֽכְמָה *ḥāḵəmâ* (verb, "she was wise") vs חָכְמָה *ḥoḵmâ* (noun, "wisdom") (GKC §16c–i; Khan 2020, §I.2.9). In our trees the meteg (U+05BD) is kept in FORM, e.g. וַֽיְהִי in GEN.1.5 and GEN.1.31.

### 8.2 Pausal forms

At a major pause — the last word of a verse (sof pasuq, with silluq) or of its first half (atnah), and sometimes at other strong disjunctive accents — words take **pausal forms**: a short or reduced vowel in the stressed syllable is lengthened or restored, and the stress may move (GKC §29i–v; Khan 2020, §I.2.8).

| context form | pausal form | process | Torah example |
| --- | --- | --- | --- |
| אֶרֶץ *ʾereṣ* | אָרֶץ *ʾāreṣ* | segol > qamets in the stressed syllable | אָרֶץ at the major breaks; but note that with the article the word is always הָאָרֶץ, in and out of pause (GKC §35o), as in GEN.1.1 |
| מַיִם *mayim* | מָיִם *māyim* | patah > qamets | GEN.1.2 הַמָּֽיִם (end of verse) vs הַמַּיִם GEN.1.7 |
| שָׁמַיִם *šāmayim* | שָׁמָיִם *šāmāyim* | patah > qamets | GEN.2.4 וְשָׁמָֽיִם (end of verse) vs הַשָּׁמַיִם earlier in the same verse |
| קָטְלוּ *qāṭəlû* | קָטָ֫לוּ *qāṭā́lû* | reduced vowel restored, stress moves back | (frequent with plural verbs) |
| אַתָּה *ʾattâ* | אָתָּה *ʾāttâ* | patah > qamets |  |
| אָנֹכִי, אֲנִי | אָנִי *ʾānî* | hatef > qamets |  |

> **GEN.2.4** · `wlc:GEN:34` אֵלֶּה תוֹלְדוֹת הַשָּׁמַיִם וְהָאָרֶץ בְּהִבָּרְאָם בְּיוֹם עֲשׂוֹת יְהוָה אֱלֹהִים אֶרֶץ וְשָׁמָיִם *ʾēlleh ṯôləḏôṯ haššāmayim wəhāʾāreṣ bəhibbārəʾām bəyôm ʿăśôṯ YHWH ʾĕlōhîm ʾereṣ wəšāmāyim* — these generations-of the-heavens and-the-earth in-their-being-created in-day-of making YHWH God earth and-heavens — "These are the generations of the heaven and of the earth when they were created, in the day that the LORD God made earth and heaven." (JPS 1917)

The verse shows the same word twice: הַשָּׁמַיִם in context, וְשָׁמָיִם in pause at the end of the verse. MACULA gives both the same lemma שָׁמַיִם, and the article is a separate `det` token. The verse also has אֶרֶץ twice: הָאָרֶץ, whose qamets comes from the article and not from pause, and the bare context form אֶרֶץ. The tree lemma is always the context form, the FORM is what the reader reads.

How regular is it? Counted in our Torah trees (2026-10-17), the lemma מַיִם has the FORM with qamets (מָיִם, מָּיִם) 53 times, and every one of them carries silluq, atnah or meteg (the marks of a pausal word at a verse end or half-verse end); the FORM with patah (מַיִם, מַּיִם) occurs 117 times in context. For אֶרֶץ the count is blurred by the article: 444 forms אָרֶץ against 282 אֶרֶץ, most of the qamets forms after הָ.

### 8.3 Maqqef

A *maqqef* (־, a raised hyphen) joins two or more words into one stress unit; only the last carries the main stress and the accent, and the first loses its stress, so its vowel may shorten: כֹּל → כָּל־ *kol-*, אֵת → אֶת־ *ʾeṯ-*, עַל־, מִן־, כִּי־ (GKC §16a–b; Khan 2020, §I.2.11). In our data the maqqef is kept at the end of the FORM of the first word (`עַל־`, `כִּי־`, `אֶת־`) and MACULA's tokenisation splits the words at the maqqef as usual: the two words become two tokens and the maqqef is not a token of its own. The object marker counts in the trees show the effect: אֵת with full tsere where it carries an accent, אֶת־ with segol before maqqef — one lemma אֵת (Strong 0853, `Type=direct_object_marker`, 4 153 tokens in the Torah).

---

## 9. Vowel change: reduction and lengthening

Because the stress is mostly final, adding an ending to a word moves the stress to the right, and the vowels before it change in a regular way (GKC §27; Davidson 1916 §§ on "vowel changes in the noun"). The basic pattern for changeable vowels:

- the vowel **of the new pretonic syllable** (directly before the stress) is long in an open syllable;
- the vowel **two syllables before the stress** (propretonic) is reduced to shewa (or a hatef under a guttural);
- if two shewas would result at the start, the rule of shewa applies (§7.3).

| base | + plural -îm | + construct | + heavy suffix -ḵem | rule |
| --- | --- | --- | --- | --- |
| דָּבָר *dāḇā́r* | דְּבָרִים *dəḇārî́m* | דְּבַר *dəḇár* | דְּבַרְכֶם *dəḇarḵém* | propretonic ā > ə; pretonic ā stays; in the construct (no stress of its own) the last ā shortens to a |
| זָקֵן *zāqḗn* | זְקֵנִים *zəqēnî́m* | זְקַן *zəqan* | — | same; ē > a in a closed unstressed syllable |
| חָכָם *ḥāḵā́m* | חֲכָמִים *ḥăḵāmî́m* | חֲכַם | — | under ח the shewa is a hatef |
| כּוֹכָב *kôḵā́ḇ* | כּוֹכָבִים *kôḵāḇî́m* | כּוֹכְבֵי (pl.) *kôḵəḇê* | — | ô is unchangeable |

This is the most frequent alternation in the language and explains much of the noun (chapter 04) and verb (chapter 06) inflection. It also explains why one lemma has many FORM strings in our trees: the lemma דָּבָר appears as דָּבָר, דְּבַר, דְּבָרִים, דִּבְרֵי, דְּבָרָיו and so on.

**Lengthening in the open pretonic syllable** is the counterpart: a short vowel (originally *a*, *i*, *u*) in an open syllable just before the stress becomes ā, ē, ō: *qatal > qāṭal (GKC §27c). Suchard (2019, §§2.1–2.8) argues that several of these rules — the shortening of long vowels before the stress, segolization, begadkefat spirantization and vowel reduction — affect Biblical Aramaic in exactly the same way, and so belong to the shared reading tradition rather than to the spoken language of the biblical authors.

---

## 10. Assimilation and sandhi

**Nun assimilates** to a following consonant when no vowel separates them; it shows up as a dagesh forte in the next letter (GKC §19c, §66):

| underlying | surface | example |
| --- | --- | --- |
| *yinpōl | יִפֹּל *yippōl* | he falls (I-nun verb) |
| *min-kōl | מִכֹּל *mikkōl* | from all (GEN.2.16, GEN.3.1) |
| *min-mennû | מִמֶּנּוּ *mimmennû* | from it (GEN.2.17) |
| *yinttēn | יִתֵּן *yittēn* | he gives |
| *ʾant | אַתְּ *ʾatt*; *ʾantā > אַתָּה | you |

Before a guttural or resh, which cannot double, the vowel of מִן lengthens instead: מֵעֵץ *mēʿēṣ* "from a tree" (GEN.2.17), מֵאַרְצְךָ "from your land" (GEN.12.1). The independent form מִן־ stays before the article (מִן־הָאֲדָמָה GEN.2.9).

> **GEN.2.16–17** · `wlc:GEN:46`, `wlc:GEN:47` מִכֹּל עֵץ־הַגָּן אָכֹל תֹּאכֵל׃ וּמֵעֵץ הַדַּעַת טוֹב וָרָע לֹא תֹאכַל מִמֶּנּוּ *mikkōl ʿēṣ-haggān ʾāḵōl tōʾḵēl; ûmēʿēṣ haddaʿaṯ ṭôḇ wārāʿ lōʾ ṯōʾḵal mimmennû* — from-every tree-of-the-garden eating you-may-eat; and-from-tree-of the-knowledge good and-evil not you-shall-eat from-it — "Of every tree of the garden thou mayest freely eat; but of the tree of the knowledge of good and evil, thou shalt not eat of it" (JPS 1917)

Here are all three outcomes of מִן: assimilation before כ (מִכֹּל), compensatory lengthening before ע (מֵעֵץ), and the doubled suffix form מִמֶּנּוּ. Notice also the first word of verse 17: the conjunction is וּ, because the next consonant is the labial מ (below).

**The article** ha- is followed by a doubling of the next consonant (הַשָּׁמַיִם), which fails before gutturals and resh (§4). After the prepositions בְּ, כְּ, לְ the ה of the article drops and its vowel moves to the preposition: *bə-hā-ʾāreṣ > בָּאָרֶץ *bāʾāreṣ*, *lə-ha-ʾôr > לָאוֹר (GKC §35n). MACULA keeps the article as a token here but with an empty FORM (1 679 such tokens in the Torah).

> **GEN.1.5** · `wlc:GEN:4` וַיִּקְרָא אֱלֹהִים לָאוֹר יוֹם וְלַחֹשֶׁךְ קָרָא לָיְלָה *wayyiqrāʾ ʾĕlōhîm lāʾôr yôm wəlaḥōšeḵ qārāʾ lāyəlâ* — and-called God to-the-light day and-to-the-darkness he-called night — "And God called the light Day, and the darkness He called Night." (JPS 1917)

Tokens 4–6 of this tree are `לָ` (ADP, case) + `` (DET, empty FORM, lemma הַ) + `אוֹר` (NOUN).

**The conjunction ו** "and" is pointed (GKC §104d–g):

| form | where | example |
| --- | --- | --- |
| וְ *wə* | the default | וְהָאָרֶץ GEN.1.2 |
| וּ *û* | before a labial ב מ פ, and before a consonant with shewa | וּמֵעֵץ GEN.2.17, וּבְכָל־ DEU.6.5, וּשְׁלֹשִׁים GEN.5.5 |
| וִ *wi* | before yod with shewa (merges as וִי) | וִיפַת GEN.29.17 |
| וַ, וֶ, וֳ | before a hatef: the matching short vowel | וַאֲחִי GEN.14.13, וֶהְיֵה GEN.17.1 |
| וָ *wā* | often before the stressed syllable of a word pair, especially in pause | תֹהוּ וָבֹהוּ GEN.1.2, טוֹב וָרָע GEN.2.9 |
| וַ + doubling | the *wayyiqtol* prefix (a different morpheme) | וַיֹּאמֶר GEN.1.3 |

All of these are one token in our trees with lemma וְ and Strong `2050b` — 13 768 tokens in the Torah, the commonest word of the text (torah-en.tsv key `lemma:וְ`).

**Other assimilations.** The t of the Hithpael prefix swaps places with a following sibilant (metathesis: *hitšammēr > הִשְׁתַּמֵּר) and becomes ṭ after ṣ (הִצְטַדֵּק), and assimilates to a following dental (הִטַּהֵר, הִדַּבֵּר) (GKC §19n, §54b–c). The ל of the verb לקח behaves like nun: יִקַּח *yiqqaḥ* "he takes" (EXO.24.7 וַיִּקַּח) (GKC §66g).

---

## 11. The weak letters א ה ו י

**ʾalef** loses its consonantal value at the end of a syllable and lengthens the preceding vowel ("quiescence"): *raʾš > רֹאשׁ *rōš*; *yaʾmur > יֹאמַר *yōmar* (GKC §23a–d). The silent א stays in the spelling. So לֵאמֹר "saying" is *lēmōr* (EXO.20.1), from *lə-ʾĕmōr*. In MACULA, לֵאמֹר is split into לֵ and אמֹר, and the second token has an empty English gloss (EXO.20.1, word 9): the gloss "saying" is not repeated on the infinitive.

**he** at the end of a word is normally a mater for a vowel (â, ê, ô, ē): תּוֹרָה, שָׂדֶה. When it is a real consonant it carries a dot, **mappiq**: אַרְצָהּ "her land" vs אַרְצָה "to the land" (GKC §14a). The difference is grammatical — mappiq ה is the 3fs suffix, mappiq-less ־ָה is the directional ending or the feminine ending. MACULA splits both into separate tokens (the suffix as PRON with `Type=pronominal`, the directional ending with `Type=directional_he`), so the ambiguity is resolved in our trees.

**waw and yod** contract with a preceding *a* into diphthongs and long vowels: *ay > ê (בַּיִת "house", construct בֵּית), *aw > ô (מָוֶת "death", construct מוֹת); *iy > î, *uw > û (GKC §24). This explains the alternations of hollow and III-he roots (chapter 07) and of nouns like בַּיִת/בֵּית, עַיִן/עֵין, יוֹם (from *yawm).

---

## 12. Historical sound changes

### 12.1 Proto-Semitic consonants that merged

Hebrew writing has 22 letters for a language that inherited about 29 proto-Semitic consonants. Several pairs merged, some after the alphabet was borrowed (GKC §6; on the Greek evidence also Blau 1982, see "wanted" in sources.md):

| proto-Semitic | Hebrew letter | evidence that the distinction lived on |
| --- | --- | --- |
| *ḥ and *ḫ | ח | Greek transcriptions of the Septuagint keep *ḫ as χ but render *ḥ without a consonant: Χαρράν for חָרָן (place, *ḫ), Ἰσαάκ for יִצְחָק (*ḥ) |
| *ʿ and *ġ | ע | the Septuagint has γ for *ġ: Γάζα עַזָּה, Γόμορρα עֲמֹרָה; but no consonant for *ʿ: Ἠλί עֵלִי, Ἀμαλήκ עֲמָלֵק (GKC §6, citing these pairs) |
| *ś (lateral) and *š | שׂ vs שׁ | one letter ש for two sounds, distinguished only by the later dot; in the Tiberian reading שׂ = ס = [s] (Khan 2020, §I.1.21) |
| *ṯ and *š | שׁ | Aramaic has ת: Hebrew שׁוֹר "ox", Aramaic תּוֹר |
| *ḏ and *z | ז | Aramaic has ד: Hebrew זָהָב "gold", Aramaic דְּהַב |
| *ṱ, *ṣ, *ḏ̣ | צ | Aramaic has ט / ע: Hebrew אֶרֶץ, Aramaic אֲרַע |

The ʿ/ġ pair is in our text in one verse:

> **GEN.10.19** · `wlc:GEN:253` וַיְהִי גְּבוּל הַכְּנַעֲנִי מִצִּידֹן בֹּאֲכָה גְרָרָה עַד־עַזָּה בֹּאֲכָה סְדֹמָה וַעֲמֹרָה וְאַדְמָה וּצְבֹיִם עַד־לָשַׁע *wayhî gəḇûl hakkənaʿănî miṣṣîḏōn bōʾăḵâ ḡərārâ ʿaḏ-ʿazzâ bōʾăḵâ səḏōmâ waʿămōrâ wəʾaḏmâ ûṣəḇōyim ʿaḏ-lāšaʿ* — and-was border-of the-Canaanite from-Sidon as-you-go to-Gerar until-Gaza as-you-go to-Sodom and-Gomorrah and-Admah and-Zeboiim until-Lasha — "And the border of the Canaanite was from Zidon, as thou goest toward Gerar, unto Gaza; as thou goest toward Sodom and Gomorrah and Admah and Zeboiim, unto Lasha." (JPS 1917)

עַזָּה (Gaza) and עֲמֹרָה (Gomorrah) had *ġ; כְּנַעֲנִי (Canaanite) and לָשַׁע had *ʿ. In the Masoretic reading they are all [ʕ].

The ḥ/ḫ pair is visible in two names that English renders alike:

> **GEN.11.31** · `wlc:GEN:297` … וְאֶת־לוֹט בֶּן־הָרָן … וַיָּבֹאוּ עַד־חָרָן וַיֵּשְׁבוּ שָׁם *wəʾeṯ-lôṭ ben-hārān … wayyāḇōʾû ʿaḏ-ḥārān wayyēšəḇû šām* — and-OBJ-Lot son-of-Haran … and-they-came until-Haran and-they-dwelt there — "and Lot the son of Haran … and they came unto Haran, and dwelt there." (JPS 1917)

The person is הָרָן with *h*; the city is חָרָן with *ḫ* (Akkadian *Ḫarrānu*, Septuagint Χαρραν). In MACULA they are different lemmas with different Strong numbers, which the MISC field keeps apart.

### 12.2 Vowel history

- **Loss of case endings and final short vowels.** Proto-Hebrew nouns ended in case vowels (*-u*, *-i*, *-a*); their loss made many words end in a consonant cluster, which Hebrew then broke with a helping vowel — the **segolates**: *malk > מֶלֶךְ *meleḵ*, *sipr > סֵפֶר *sēp̄er*, *qudš > קֹדֶשׁ *qōḏeš* (GKC §84a, §93; chapter 04). The original vowel resurfaces before suffixes: מַלְכִּי, סִפְרִי, קָדְשִׁי. The directional ending ־ָה is a relic of the old accusative *-a* according to GKC §90a–c, although modern work derives it differently (see chapter 04).
- **The Canaanite shift** *ā > ō: proto-Semitic long *ā became ō in Canaanite (including Hebrew and the Amarna glosses), which is why Hebrew has שָׁלוֹם *šālôm* against Arabic *salām*, the participle קֹטֵל *qōṭēl* against Arabic *qāṭil*, and אָנֹכִי against Akkadian *anāku* (GKC §9q; the Amarna evidence in Baranowski 2016, $RESEARCH_RAW/hbo-2016-baranowski-biblical-hebrew-wayyiqtol-evidence-amarna).
- **Attenuation** *a > i in closed unstressed syllables: *yaqtul > יִקְטֹל *yiqṭōl*, *maktab > מִכְתָּב (GKC §27s; the Philippi law, that stressed *i became *a in some closed syllables, is discussed in GKC §29f under "the tone").
- **Pretonic changes, segolization, spirantization and reduction** (§§3, 9), which Suchard places in the reading tradition shared with Biblical Aramaic (Suchard 2019, §§2.1–2.7).
- **Monophthongization** *ay > ê, *aw > ô in unstressed syllables (§11).

### 12.3 The epigraphic and Qumran evidence

Inscriptions of the monarchic period (Siloam, Arad, Lachish) are written with few matres and show dialect features — e.g. the northern Samaria ostraca write יין "wine" as *yn* (contracted *ayn > ên) where Judean writes יין with the diphthong spelled out (GKC §2; Hutton 2015, $RESEARCH_RAW/hbobook-2015-hutton-epigraphy-philology-hebrew-bible; Stefaniak 2026 on Arad, $RESEARCH_RAW/hbo-2026-stefaniak-old-hebrew-inscriptions-tel-arad). The Dead Sea Scrolls use many more matres (כול for כֹּל, לוא for לֹא) and sometimes long forms of pronouns and suffixes (הואה, -כמה) that the Tiberian tradition lacks (Hornkohl 2020, $RESEARCH_RAW/hbo-2020-hornkohl-hebrew-dead-sea-scrolls-tiberian). These are discussed in chapter 12.

---

## 13. Other reading traditions

The Tiberian is one of several medieval vocalization systems (Khan 2020, §I.0.1–0.4; Kantor 2023, $RESEARCH_RAW/hebgram/nc/kantor-classification-reading-traditions-2023 — NC, cited only):

| tradition | vowel signs | main differences from Tiberian |
| --- | --- | --- |
| Tiberian | sublinear and supralinear points, 7 qualities + shewa + 3 hatefs | — |
| Babylonian | supralinear signs, 6 qualities (no segol/patah distinction) | no segol; different treatment of shewa |
| Palestinian | supralinear signs, used inconsistently | qamets and patah often merged, tsere and segol merged — close to the later Sephardi reading |
| Samaritan | oral reading tradition with no full pointing of the Pentateuch | gutturals largely lost; stress mostly penultimate; many morphological differences (Ben-Ḥayyim; see wanted) |

The modern Israeli and the Sephardi pronunciations are closest to the Palestinian type: five vowels (a e i o u), no length, qamets read [a] except as hatuf, shewa mostly [e] or zero, ח = [χ], ע = [ʔ] or zero, ת and ד without fricatives, ב כ פ with them. The Ashkenazi pronunciation reads qamets [ɔ] or [u] and tav rafe as [s] (Suchard 2019, §1 and n. 2, on the Ashkenazi holem). None of these is the Tiberian pronunciation of §2 and §5 (Khan 2020, Preface).

---

## 14. In Mova

**What the trees keep.** The Torah trees (*.conllu, from MACULA Hebrew, CC BY 4.0) keep the full Tiberian pointing:

- `# text` is the verse in the Westminster Leningrad Codex, with points, accents, maqqef, paseq (׀) and the end-of-verse sof pasuq (׃) and the section marks ס / פ;
- FORM is the surface form of each token with its points and accents; the maqqef stays on the left word (`עַל־`), the meteg (U+05BD) stays, a paseq can stay inside the FORM (`אֱלֹהִ֤ים ׀` in GEN.1.5);
- LEMMA is the pointed dictionary form, without accents, often with the stop dagesh the context form lacks (FORM `תֹ֙הוּ֙`, LEMMA `תֹּהוּ`).

So every rule of this chapter is visible in the data: begadkefat dots, pausal qamets, maqqef shortening, hatefs on gutturals, the empty article after a preposition. None of them is encoded as a feature; they must be read off the strings.

**What the trees do not keep.** There is no phonological transcription, no syllabification, no stress feature (the accent marks stand where the stress is, but nothing labels them), no pausal flag. MACULA's own data has a transliteration field; `tools/macula` does not carry it into the CoNLL-U. A transliteration in the scheme of this manual could be produced deterministically from the pointed FORM by the rules of §§3–8, with the known ambiguities (qamets gadol/hatuf, vocal/silent shewa after a short vowel with meteg) resolved by the rules of GKC §9u, §10, §16. That converter does not exist yet.

**Comparing words across sources.** `macula compare` and the dictionary builder (`macula dict`) compare lemmas by their **consonants only** (function `consonants()` in tools/macula/src/main.rs), "since the order of points differs between sources" (tools/macula/README.md). This makes the comparison blind to vowels — so it cannot tell, say, דָּבָר "word" from דֶּבֶר "plague" (both דבר); those are kept apart by the Strong number in MISC. It also does not unify plene and defective spellings (אבות vs אבת), which differ in consonants.

**Pausal and context forms** of one word share a lemma in the trees (אֶרֶץ, שָׁמַיִם, מַיִם), so counts by lemma (torah-en.tsv `count`) already merge them. Counts by FORM do not.

**Accents and pragmatics.** `macula prag` reads nothing of the phonology; its rules use the verb `Type`, person, and a few particles (see chapter 09). A future reading of the accents (a disjunctive accent before a word in a quotation, the atnah as the verse's main break) would give a prosodic phrase structure that could help both parsing and pragmatics (on the syntax–prosody interface: Naudé & Miller-Naudé 2018, $RESEARCH_RAW/hbo-2018-naude-interface-syntax-prosody-differentiating-left; Pitcher 2021, $RESEARCH_RAW/hbo-2021-pitcher-towards-prosodic-model-tiberian-hebrew).

**Sefer Yetzirah.** The crate `yetzirah` holds the letter classes of Sefer Yetzirah — three mothers, seven doubles (בג"ד כפר"ת), twelve simple, and the five places of the mouth (2:3) — from the two recensions in `docs/sefer-yetzirah/text/`, each quoting its source and checked by `cargo run -p yetzirah -- check`. It works on letters, not on pronunciation; it does not know the dagesh. The match between the "doubles" and the begadkefat letters plus resh is a fact of the book, stated in §2, not something the crate infers. `yetzirah -- letter ב` prints what the three readings (standard, Gra, ours) say about bet.

**Gaps (2026-10-17).**

- no transliterator and no syllabifier; no stress or pause feature;
- no normalization of plene/defective spellings;
- the empty-FORM article tokens (1 679) are a CoNLL-U oddity: a token with no surface. UD would rather make לָאוֹר a multiword token `לָאוֹר` = `לְ` + `הַ` + `אוֹר`; our files keep MACULA's split without the range line;
- qere/ketiv: the trees follow MACULA's choice and do not mark which reading was taken (chapter 01).

---

## 15. Sources

- **GKC** — W. Gesenius, E. Kautzsch, A. E. Cowley, *Gesenius' Hebrew Grammar*, 2nd English ed., Oxford 1910. §§5–29 (script and sounds), §35 (article), §49 (waw consecutive stress), §54 (Hithpael assimilation), §66 (I-nun), §84a, §90, §93, §102, §104. Public domain. Text: $RESEARCH_RAW/hebgram/open/gesenius-kautzsch-cowley-1910-wikisource/gesenius-kautzsch-cowley-1910-wikisource.txt (Wikisource); scan: $RESEARCH_RAW/hebgram/pd/gesenius-kautzsch-cowley-1910/; [en.wikisource.org/wiki/Gesenius' Hebrew Grammar](https://en.wikisource.org/wiki/Gesenius%27_Hebrew_Grammar)
- **Khan 2020** — Geoffrey Khan, *The Tiberian Pronunciation Tradition of Biblical Hebrew*, vol. 1–2, Open Book Publishers, Cambridge 2020, [doi.org/10.11647/OBP.0163](https://doi.org/10.11647/OBP.0163) [our card](https://md.mova.elamur.ai/free/en/docs/papers/hbobook-obp0163-khan-tiberian-pronunciation-tradition-biblical-hebrew/), CC BY 4.0. §I.0 (traditions), §I.1 (consonants), §I.2 (vowels, shewa, maqqef, gaʿya), §I.3 (dagesh), §I.5 (summary tables and the transcription of Gen 1:1–13, pp. 608–620). $RESEARCH_RAW/hebgram/open/khan-tiberian-pronunciation-v1-2020/, …-v2-2020/
- **Suchard 2019** — Benjamin D. Suchard, "Sound Changes in the (Pre-)Masoretic Reading Tradition and the Original Pronunciation of Biblical Aramaic", *Studia Orientalia Electronica* 7 (2019) 52–65, [doi.org/10.23993/store.74104](https://doi.org/10.23993/store.74104) [our card](https://md.mova.elamur.ai/free/en/docs/papers/hbo-2019-suchard-sound-changes-pre-masoretic-reading/) (the PDF states CC BY 3.0; our registry lists CC BY-NC-SA — cited only). $RESEARCH_RAW/hbo-2019-suchard-sound-changes-pre-masoretic-reading/paper.pdf
- **Davidson 1916** — A. B. Davidson, *An Introductory Hebrew Grammar*, 19th ed., Edinburgh 1916, public domain. $RESEARCH_RAW/hebgram/pd/davidson-introductory-hebrew-grammar-1916/
- **Kantor 2023** — Benjamin Kantor, *The Linguistic Classification of the Reading Traditions of Biblical Hebrew*, OBP 2023 (licence NC in our registry: cited, not copied). $RESEARCH_RAW/hebgram/nc/kantor-classification-reading-traditions-2023/
- **Hornkohl 2020** — Aaron Hornkohl, on the Hebrew of the Dead Sea Scrolls and the Tiberian tradition. $RESEARCH_RAW/hbo-2020-hornkohl-hebrew-dead-sea-scrolls-tiberian/
- **Hutton 2015** — Jeremy M. Hutton (ed. vol.), *Epigraphy, Philology, and the Hebrew Bible*, 2015. $RESEARCH_RAW/hbobook-2015-hutton-epigraphy-philology-hebrew-bible/
- **Stefaniak 2026** — on the Old Hebrew inscriptions of Tel Arad. $RESEARCH_RAW/hbo-2026-stefaniak-old-hebrew-inscriptions-tel-arad/
- **Baranowski 2016** — Krzysztof Baranowski, on the Biblical Hebrew wayyiqtol and the evidence of Amarna. $RESEARCH_RAW/hbo-2016-baranowski-biblical-hebrew-wayyiqtol-evidence-amarna/
- **Pitcher 2021** — towards a prosodic model of Tiberian Hebrew. $RESEARCH_RAW/hbo-2021-pitcher-towards-prosodic-model-tiberian-hebrew/
- **Naudé & Miller-Naudé 2018** — the interface of syntax and prosody in left dislocation. $RESEARCH_RAW/hbo-2018-naude-interface-syntax-prosody-differentiating-left/
- **MACULA Hebrew** — Clear Bible / Biblica, [github.com/Clear-Bible/macula-hebrew](https://github.com/Clear-Bible/macula-hebrew/), CC BY 4.0; text and morphology from the Open Scriptures Hebrew Bible (CC BY 4.0). Our trees: $MOVA_DATA/corpora/judaism/text/torah-wlc-macula/; converter: tools/macula.
- **JPS 1917** — The Holy Scriptures according to the Masoretic Text, Jewish Publication Society 1917, public domain, via eBible.org. $MOVA_DATA/corpora/judaism/text/torah-jps1917/
- **Sefer Yetzirah** — Warsaw 1884 edition, standard and Gra recensions, public domain. docs/sefer-yetzirah/text/; crate yetzirah.
- Wanted (not held): J. Blau, *On Polyphony in Biblical Hebrew* (1982); Z. Ben-Ḥayyim, *A Grammar of Samaritan Hebrew* (2000); E. J. Revell on Masoretic pausal forms; J. Fox, *Semitic Noun Patterns* (2003); R. Steiner on the Septuagint transcriptions of ḥ/ḫ and ʿ/ġ.
- [← previous 2. Cantillation: the accents (te'amim)](https://md.mova.elamur.ai/free/en/docs/languages/hbo/02-cantillation/)
- [next → The Noun, the Adjective and the Numeral](https://md.mova.elamur.ai/free/en/docs/languages/hbo/04-noun/)
- [↑ back to Ancient Hebrew](https://md.mova.elamur.ai/free/en/docs/languages/hbo/)