[Mova](https://md.mova.elamur.ai/free/en/) · [this page in HTML](https://mova.elamur.ai/free/en/docs/languages/he/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/) / [Modern Hebrew](https://md.mova.elamur.ai/free/en/docs/languages/he/) / 3. Phonology

# 3. Phonology

**On this page**

1. [3.1 Overview](#31-overview)
2. [3.2 Consonants](#32-consonants)
3. [3.3 Mizrahi pronunciation: pharyngeals and trilled r](#33-mizrahi-pronunciation-pharyngeals-and-trilled-r)
4. [3.4 The rhotic](#34-the-rhotic)
5. [3.5 Spirantization of b, k, p](#35-spirantization-of-b-k-p)
6. [3.6 Vowels](#36-vowels)
7. [3.7 The syllable and consonant clusters](#37-the-syllable-and-consonant-clusters)
8. [3.8 Stress](#38-stress)
9. [3.9 Assimilation](#39-assimilation)
10. [3.10 Casual speech](#310-casual-speech)
11. [3.11 Prosody and intonation](#311-prosody-and-intonation)
12. [3.12 Summary for processing](#312-summary-for-processing)
13. [References](#references)

This chapter describes the sounds of Modern Hebrew **as it is spoken today** by the majority of Israeli speakers. Linguists call this pronunciation "General Israeli Hebrew" or "Standard Israeli Hebrew" (Schwarzwald 2011). The chapter covers the consonants and the historical mergers behind them, the pharyngeal and trilled-r pronunciation of Mizrahi speakers, the stop/spirant alternation of ב כ פ and how speakers level it, the five vowels, shva and epenthesis, stress and its grammatical role, voicing assimilation and other assimilations, the syllable and its clusters, the reductions of casual speech, and intonation. The spelling side of the same facts (points, shva signs, dagesh) is in [2. Script and spelling](https://md.mova.elamur.ai/free/en/docs/languages/he/02-script-and-spelling/). Transliteration conventions are defined there in §2.12.1.

Index: [Modern Hebrew (he): a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/he/) · previous: [2. Script and spelling](https://md.mova.elamur.ai/free/en/docs/languages/he/02-script-and-spelling/) · next: [Nouns, adjectives, pronouns and numerals](https://md.mova.elamur.ai/free/en/docs/languages/he/04-nouns-adjectives-pronouns/) · related: [05. The verb: system, binyanim, tense and use](https://md.mova.elamur.ai/free/en/docs/languages/he/05-verbs/) (weak roots), [10. Historical layers: Biblical, Mishnaic, Medieval and Modern Hebrew](https://md.mova.elamur.ai/free/en/docs/languages/he/10-historical-layers/) (Biblical and Mishnaic Hebrew)

---

## 3.1 Overview

Modern Hebrew has a small, regular sound system, quite unlike the reconstructed sound system of Biblical Hebrew or the Tiberian reading tradition:

- **five vowels** /a e i o u/, with no phonemic length;
- about **23–27 consonants**, depending on whether one counts the loan phonemes /dʒ ʒ tʃ w/ and the glottal stop, and whether the speaker keeps the pharyngeals /ħ ʕ/;
- **no emphatic** (pharyngealized) consonants and, in General Israeli speech, **no pharyngeals**;
- a **uvular** rhotic /ʁ/;
- **contrastive stress**, mostly on the last or second-to-last syllable;
- a stop/spirant alternation **b~v, k~χ, p~f** inherited from older Hebrew. It is still part of the morphology, but it is now irregular and being levelled.

The sound system of the revived language took shape in the first half of the 20th century. It grew out of the Sephardi reading tradition that the revivers chose as their norm, as realized by speakers whose first languages were mostly Yiddish, Russian, Polish and German (Zuckermann 2003, 2020; Schwarzwald 2011). Its phonetics are therefore largely European: a uvular r, no pharyngeals, and [ei] for many /e/. Its phonology, the distribution of sounds in words, stays close to the written tradition. Laufer (1999) gives a phonetic sketch in the IPA *Handbook*. Bolozky (1997) and Asherov & Cohen (2019) give phonological and phonetic descriptions.

## 3.2 Consonants

### 3.2.1 Inventory

The table gives the consonant phonemes of General Israeli Hebrew in IPA. The usual Hebrew letters are added in brackets, and sounds that occur only in loanwords are marked †. Sounds kept only by some speakers are in parentheses.

|  | Bilabial | Labiodental | Alveolar | Postalveolar | Palatal | Velar | Uvular | Pharyngeal | Glottal |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| **Plosive** | p (פּ) b (בּ) |  | t (ת ט) d (ד) |  |  | k (כּ ק) g (ג) |  |  | ʔ (א ע) |
| **Affricate** |  |  | ts (צ) dz† | tʃ† (צ׳) dʒ† (ג׳) |  |  |  |  |  |
| **Fricative** |  | f (פ) v (ב ו) | s (ס שׂ) z (ז) | ʃ (שׁ) ʒ† (ז׳) |  |  | χ (ח כ) ʁ (ר) | (ħ) (ח) (ʕ) (ע) | h (ה) |
| **Nasal** | m (מ) |  | n (נ) |  |  |  |  |  |  |
| **Lateral** |  |  | l (ל) |  |  |  |  |  |  |
| **Approximant** | w† (ו, וו) |  |  |  | j (י) |  |  |  |  |

Notes:

- **/χ/** is usually described as uvular [χ] or a velar-uvular [x~χ] (Laufer 1999; Asherov & Cohen 2019). This manual writes it **kh**.
- **/ʁ/** is the rhotic. It is a voiced uvular fricative or approximant [ʁ~ʁ̞], and more rarely a uvular trill [ʀ] (§3.4). Phonologically, it patterns with the sonorants in some ways (it is allowed as the second member of an onset, as in *pri* "fruit") and with the gutturals in others (it historically resisted gemination). It voices and devoices with its neighbours: *khatsrot* "yards" has a voiceless [χ] before [ts].
- **/ʔ/** is a marginal phoneme. It appears in careful speech between vowels and after a prefix (*ma'amar* "article", *le'ekhol* "to eat", *ha-'ir* ~ *ha-ir* "the city"), but it is very often left out (§3.2.3).
- **/h/** is weak. It is often dropped, especially at the start of unstressed syllables and in the article *ha-* in casual speech (§3.11).
- **Loan phonemes**: /dʒ ʒ tʃ/ are well established in loanwords and slang (*djuk* "cockroach", *zhaket* "jacket", *chips*) and are written with the geresh ([02-script-and-spelling.md §2.5](https://md.mova.elamur.ai/free/en/docs/languages/he/02-script-and-spelling/#25-the-geresh-sounds-that-hebrew-letters-lack)). /w/ occurs in loans and names (*Washington*, *wódka*~*vódka*) and is often replaced by [v]. /dz/ is marginal (*godzila*).
- There is no **/θ/ /ð/ /ɣ/** in speech. They existed as spirant allophones of ת ד ג in older Hebrew and were lost (§3.2.2).

### 3.2.2 Mergers: the letters that sound alike

Several distinctions of Biblical and Mishnaic Hebrew have merged in General Israeli Hebrew. This is why the spelling has more consonant letters than the speech has phonemes ([02-script-and-spelling.md §2.1.2](https://md.mova.elamur.ai/free/en/docs/languages/he/02-script-and-spelling/#212-letters-with-two-values-and-letters-with-one-value)):

| Older contrast | Letters | Modern (General Israeli) | Example | Note |
| --- | --- | --- | --- | --- |
| /t/ vs. emphatic /tˤ/ | ת : ט | both [t] | תָּעָה *ta'a* "strayed" = טָעָה *ta'a* "erred" | emphatics lost their secondary articulation |
| /k/ vs. /q/ | כּ : ק | both [k] | כָּרָה *kara* "dug" = קָרָה *kara* "happened" |  |
| /ħ/ vs. spirant [x] of /k/ | ח : כ | both [χ] | סָחַר *sakhar* "traded" = שָׂכַר *sakhar* "hired" | kept apart by many Mizrahi speakers (§3.3) |
| /ʔ/ vs. /ʕ/ | א : ע | both [ʔ] or nothing | אִם *im* "if" = עִם *im* "with" | kept apart by many Mizrahi speakers |
| /w/ vs. spirant [β] of /b/ | ו : ב | both [v] | (spelling contrast only) | /w/ returned in loans |
| /s/ vs. /ɬ/ (sin) | ס : שׂ | both [s] | סָם *sam* "drug" = שָׂם *sam* "put" | the merger is already ancient |
| emphatic /sˤ/ | צ | [ts] | צִפּוֹר *tsipor* "bird" | an affricate in the Ashkenazi and modern reading |
| spirants [θ ð ɣ] of ת ד ג | ת ד ג | [t d g] | בַּיִת *báyit* (not *bayiθ*) | Yemenite and some Ashkenazi traditions kept some of them |
| geminates (dagesh forte) | — | single consonants | דִּבֵּר *diber* (not *dibber*) | gemination lost in General Israeli speech |

Sources: Laufer (1999); Bolozky (1997); Schwarzwald (2011); Wikipedia "Modern Hebrew phonology" (accessed 2026-10-17); for the Tiberian values, Khan (2020).

As a result, a General Israeli speaker cannot spell many words from their sound alone. Children and adults make spelling errors along exactly these lines (ט/ת, כ/ק/ח, א/ע/ה), and the spelling of a root has to be learned. For NLP, this also means that a phonemic transcription loses information that the spelling keeps (and the other way round, §2.4.4).

### 3.2.3 The glottals and the "guttural" letters

The four letters traditionally called gutturals (*otiyot garoniyot*), א ה ח ע, plus ר, behave in a special way in the morphology. They do not take a dagesh, they prefer an *a* vowel next to them, and they take a *khataf* vowel where other letters take a shva. In modern speech:

- **א and ע** are a glottal stop [ʔ] or nothing. Between vowels, the stop is often weakened to a glide or a break in voice, or dropped completely, leaving two vowels in a row: *ma'amar* [maʔaˈmaʁ ~ maaˈmaʁ ~ maˈmaʁ] "article", *ta'im* [taˈim] "tasty", *sha'ón* [ʃaˈon] "watch". After a consonant, it usually disappears: *mar'é* [maʁˈe] "view". At the start of a word, a glottal onset is optional.
- **ה** is [h] at the start of a stressed syllable in careful speech, and is often dropped elsewhere: *ha-yéled* [(h)aˈjeled], *mahír* [maˈiʁ ~ maˈhiʁ] "fast". It is never pronounced at the end of a word, except for the mappiq suffix (*la(h)* "to her") in careful speech.
- **ח** is [χ] for most speakers and [ħ] for some (§3.3).

Even so, the **morphology remembers** the old gutturals. Verbs whose root has א, ה, ח, ע or ר still take the "guttural" vowel patterns: *yaakhol* "will eat", not **yikhol*; *nish'ar* "remained" with an *a*; *lishmóa* "to hear" with a "furtive" *a* after the final ע (the *patakh gnuva*: *shoméa* "hears", *tapúakh* "apple"). These patterns are learned as part of the inflection classes ([05. The verb: system, binyanim, tense and use](https://md.mova.elamur.ai/free/en/docs/languages/he/05-verbs/)), not derived from the modern pronunciation.

## 3.3 Mizrahi pronunciation: pharyngeals and trilled r

Speakers of Mizrahi (Oriental) origin, whose families came from Arabic-speaking countries (Iraq, Yemen, Morocco and others), traditionally keep some older distinctions in their Hebrew:

- **ח as a pharyngeal fricative [ħ]**, distinct from כ [χ]: *khaver* [ħaˈveʁ] "friend" vs. *mikhtav* [mixˈtav] "letter";
- **ע as a pharyngeal [ʕ]** (less often kept than [ħ]): *ani* [ʕaˈni] "poor" vs. אֲנִי *ani* [ʔaˈni] "I";
- an **alveolar trilled or tapped r** [r ~ ɾ];
- in some speakers, especially older Yemenites, **gemination** and an extra vowel for some shva na. A few elderly Yemenite speakers also keep emphatics such as [sˤ].

This pronunciation was promoted as the **prestige norm** in the early state: the Academy and the national broadcaster recommended it, and radio and TV announcers were trained to use it. In everyday speech, however, the European-based pronunciation became the norm. Since about the 1970s–80s, the Mizrahi features have declined among younger speakers and are now heard mainly from older speakers, in some traditional and religious contexts, and in some singers and public figures. They also work as an ethnic and social marker (Gafter 2014, 2016; Zuckermann 2005; Schwarzwald 2011). The pharyngeals are sometimes used to separate otherwise identical words. For example, some speakers who otherwise drop ע keep it in *ba'ali* "my husband" to distinguish it from *ba li* "I feel like (it)" (Wikipedia "Modern Hebrew phonology", citing Asherov & Cohen 2019).

The Arab citizens of Israel who speak Hebrew as a second language generally have the pharyngeals and emphatics of Arabic in their native phonology and may keep ח ע distinct in Hebrew. Their Hebrew is also recognizable by other features, such as /p/ said as [b] by some speakers, since Arabic has no /p/. We have not found a held source that describes this systematically; see «wanted» in [Sources](https://md.mova.elamur.ai/free/en/docs/languages/he/sources/).

| Feature | General Israeli | Traditional Mizrahi |
| --- | --- | --- |
| ח | [χ] | [ħ] |
| ע | [ʔ] / Ø | [ʕ] |
| ר | [ʁ] | [r] / [ɾ] |
| gemination | none | some, in careful speech |
| tsere | [e] ~ [ei] | [e] |
| social value today | unmarked | ethnic marker, "authentic", formal-traditional |

## 3.4 The rhotic

The Hebrew Language Committee and Eliezer Ben-Yehuda, who based the standard on the Sephardi pronunciation, recommended a trilled alveolar r. But most of the early speakers were Ashkenazim from Central and Eastern Europe, many of whom already had a uvular r in Yiddish or German, and the **uvular** r became the general one. Today, nearly all Jewish Israelis who do not have a Mizrahi background use a uvular fricative or approximant [ʁ ~ ʁ̞], which is often voiceless next to voiceless consonants [χ̝]. A uvular trill [ʀ] also occurs. The alveolar trill is an ethnic and stylistic variable (Gafter 2021; Laufer 1999).

The uvular r has phonological consequences:

- it is **very close to /χ/**: the two differ mainly in voicing. Before a voiceless consonant, /ʁ/ may become voiceless, and in clusters such as *khatsrot* "yards" or *ártsot* "lands" the r may be heard as a weak [χ];
- it appears in **onset clusters** (*pri* "fruit", *brera* "choice", *dror* "freedom", *tsrikha* "must (F)"), but a word-initial *rC-* cluster is avoided by epenthesis (*re-* in *rekhov* "street", not *rkhov*; §3.6).

## 3.5 Spirantization of b, k, p

### 3.5.1 The old rule

In Biblical and Mishnaic Hebrew, the six letters ב ג ד כ פ ת (*begadkefat*) had two pronunciations each: a stop at the start of a word and after a consonant, and a spirant (fricative) after a vowel. This was an allophonic rule (Khan 2020). Three of the six alternations survive in Modern Hebrew:

| Stop (dagesh) | Spirant (no dagesh) | Example (stop) | Example (spirant) |
| --- | --- | --- | --- |
| בּ [b] | ב [v] | *bayit* → *báyit* "house" | *be-váyit* "in a house" |
| כּ [k] | כ [χ] | *katav* "wrote" | *yikhtov* "will write" |
| פּ [p] | פ [f] | *paras* "spread" | *yifros* "will spread" |

The other three (ג ד ת) no longer alternate: they are always [g d t].

### 3.5.2 Why it is no longer a rule

In modern speech, the alternation is **no longer predictable** from the sound context, for three reasons:

1. **Mergers created spirants and stops that do not alternate.** [v] from ו never alternates with [b], [χ] from ח never alternates with [k], and [k] from ק never alternates with [χ]: *khibek* "hugged" (ח) – *yekhabek* (still [χ]); *kibel* "received" (ק) – *yekabel* (still [k]). Unless the speaker knows the spelling, they cannot know whether a given [k] or [χ] should alternate.
2. **Gemination was lost.** In the pi'el, a geminated middle consonant blocked spirantization (*dibber* → *diber*). Without gemination, a stop after a vowel is no longer explained: *siper* "told" has [p] after a vowel.
3. **Loanwords** bring stops and spirants in every position: initial [f] (*film*, *falafel*), [p] after vowels (*kapuchíno*), final [p] (*tsip* "chip").

So in Modern Hebrew, [b]/[v], [k]/[χ] and [p]/[f] are **separate phonemes** (Bolozky 1997; Schwarzwald 2011). The alternation survives as a **morphophonological** pattern tied to particular lexemes and inflection classes. Each stem has to be learned with the right consonant in each slot.

### 3.5.3 Where it still applies

The alternation is regular in **native verbs**, which change form by binyan and tense, and in prefixed forms:

```
כתב / יכתוב / לכתוב / מכתב
katav / yikhtov / likhtov / mikhtav
write.PST.3MS / write.FUT.3MS / INF-write / letter
'he wrote / he will write / to write / a letter'
```

```
פתח / יפתח / נפתח / מפתח
patakh / yiftakh / niftakh / mafte'akh
open.PST.3MS / open.FUT.3MS / open.PASS.PST.3MS / key
'he opened / he will open / it was opened / a key'
```

```
בנה / יבנה / נבנה / מבנה / בבניין
bana / yivne / nivna / mivne / ba-binyan
build.PST.3MS / build.FUT.3MS / build.PASS.PST.3MS / structure / in.DEF-building
'he built / he will build / it was built / a structure / in the building'
```

It also applies after the **proclitic prefixes** ב כ ל ו in careful speech: *báyit* "house" – *be-váyit* "in a house"; *pe* "mouth" – *be-fe* "orally"; *kise* "chair" – *le-khise* (careful) / *le-kise* (common) "to a chair". The prefix מ *mi-* is different: it historically doubled the next consonant, so it is followed by a stop (*mi-kfar* "from a village", *mi-báyit* "from a house"). With proper names and loanwords, and increasingly with ordinary nouns in casual speech, the other prefixes do **not** trigger spirantization either: *be-Paris*, *le-Bat Yam*, *ke-politikai* "as a politician".

### 3.5.4 Levelling in everyday speech

Since the rule is no longer phonological, speakers often **level** the paradigm. They keep one consonant, the one in the most frequent or basic form (usually the past 3MS or the noun), throughout the paradigm. The literature describes this as one of the most common deviations from the normative grammar (Bolozky 1997; Schwarzwald 2011; Ravid 1995). Typical cases:

| Normative | Colloquial (levelled) | Root | Gloss | Levelled to |
| --- | --- | --- | --- | --- |
| *lekhabes* | *lekabes* | כ־ב־ס | to launder (pi'el) | [k] of past *kibes* |
| *yekhabes* | *yekabes* | כ־ב־ס | he will launder | [k] |
| *lefater* | *lepater* | פ־ט־ר | to fire (from a job) | [p] of past *piter* |
| *mefater* | *mepater* | פ־ט־ר | fires (PRS) | [p] |
| *lefarsem* | *leparsem* | פ־ר־ס־ם | to publish (quadriliteral, pi'el pattern) | [p] of past *pirsem* |
| *lefaspes* | *lepaspes* | פ־ס־פ־ס | to miss (an opportunity) | [p] of *pisfes* |
| *yikhtov* | (stable) | כ־ת־ב | will write | not levelled: high frequency keeps it |
| *nikhnas* | (stable) | כ־נ־ס | entered |  |
| *be-váyit* | (stable) | בית | in a house | high-frequency word |

Some patterns are worth stating:

- **Pi'el verbs** (and the quadriliteral verbs inflected like them) level most often. The past and the noun of action have the stop (*kibes*, *kibus*; *piter*, *piturim*), while the future, infinitive and participle have the spirant after the prefix vowel (*yekhabes*, *lekhabes*, *mekhabes*). Speakers extend the stop.
- **Pa'al verbs** of high frequency (*katav–yikhtov*, *patakh–yiftakh*, *bakha–yivke*) keep the alternation. Less frequent ones level more often.
- **New verbs** formed from loanwords and acronyms, and many denominal verbs, **do not** spirantize at all: *lefakses* "to fax" (פקסס, from *faks*) keeps [f] after any prefix, and *lefarmet* "to format" (פרמט) keeps [f] where a native pi'el would have [p] in the past: past *firmet*, not **pirmet*. Such verbs follow their base, not the rule (Bolozky 1999).
- **Children** level more than adults. Normative forms are taught at school and corrected in formal registers (Ravid 1995).

For a written-text system, levelling is invisible in unpointed spelling: לכבס is written the same whether said *lekhabes* or *lekabes*. It matters for speech recognition and synthesis, and for pointed text.

## 3.6 Vowels

### 3.6.1 The five vowels

|  | Front | Central | Back |
| --- | --- | --- | --- |
| **Close** | i |  | u |
| **Mid** | e |  | o |
| **Open** |  | a |  |

- There is **no phonemic length** and no reduced vowels (the traditional *khataf* and *shva na* are realized as full [e], [a] or [o], or as nothing).
- /e/ and /o/ are true mid vowels [e̞ o̞]. /a/ is central [a~ä].
- Vowels are somewhat **longer under stress** and in open syllables, but this is not contrastive (Laufer 1999).
- **Diphthongs**: native [aj] (*báyit* "house", *khay* "alive"), [ej] (*beyt* "house of", *eyn* "there is not", *beytsa* "egg", *leyl* "night of"), [oj] (*goy* "nation", *oy* "oh"), and [uj] in passive participles of final-weak roots (*banuy* "built", *ra'uy* "worthy"). Loans add [aw] (*áwt* "out", in sports).
- **/e/ as [ej]**. Many speakers say tsere, especially before a yod in the spelling, as [ej]. In the speech of some, any stressed /e/ can be raised or diphthongized: *ken* [kein], *sheni* [ʃeiˈni] "second" (Laufer 1999; Wikipedia). It is a stylistic and regional variable, not a contrast.

### 3.6.2 Vowel sequences

Because ע, א and ה have weakened, Modern Hebrew has many **vowel sequences**, a vowel directly after another in two syllables: *ta'ut* [taˈut] "mistake", *me'od* [meˈod] "very", *ra'á* [ʁaˈa] "saw (F)", *sha'á* [ʃaˈa] "hour", *mo'adon* [moaˈdon] "club". Identical vowels may merge in fast speech: *ma'amar* [maˈmaʁ]. Some speakers insert a light [ʔ] or [j] between vowels.

### 3.6.3 Shva and the vowel zero

The written *shva* stands for either nothing or a short vowel ([02-script-and-spelling.md §2.2.2](https://md.mova.elamur.ai/free/en/docs/languages/he/02-script-and-spelling/#222-shva-na-and-nakh)). In Modern Hebrew speech, a written shva is pronounced as [e] or as nothing according to **phonotactics and morphology**, not according to the traditional na/nakh rules (Bolozky 1997; Laufer 1999):

- **[e] in prefixes**: *be-*, *le-*, *ke-*, *ve-*, *she-*, and the participle prefix *me-* (*medaber*), plus the future prefixes before a shva (*te-*, *ye-*: *yedaber* "will speak").
- **[e] to break up a forbidden cluster** at the start of a word (§3.7).
- **[e] in some morphological slots**, such as between two identical consonants (§3.6.4) and in some plural and suffixed forms: *dvarim* "things" / *divrey* "words of", *yeladim* "children" (from *yéled*).
- **Nothing** elsewhere: *zman* "time", *sfarim* "books", *kfar* "village", *mitkhilim* "beginning (PL)".

Speakers also vary. Some careful or Mizrahi speakers pronounce more shvas (*zeman*, *shemone*), and some fast speakers pronounce fewer (*bseder* for *be-séder* "OK").

### 3.6.4 Epenthesis

An epenthetic **[e]** is inserted in two main situations:

1. **Between two identical, or nearly identical, consonants** that would otherwise come together across a morpheme boundary, especially before a consonant-initial suffix. Typical cases are the past tense suffixes *-ti*, *-ta*, *-t*, *-nu*, *-tem/-ten* after a stem ending in **t, d** (or **n** before *-nu*): Root Without epenthesis With epenthesis Glossל־מ־ד **lamadti* *lamádeti* (colloquial) / *lamádti* (normative) I learned י־ל־ד **yaladti* *yaládeti* (coll.) / *yaládti* (normative) I gave birth ש־ד־ד **shadadti* *shadádeti* (coll.) / *shadádti* (normative) I robbed ש־כ־נ *shakhánu* (from *shakhan+nu*) — we dwelt (two n written נ once: שכנו) Normative grammar and careful speakers say *lamádti*, *yaládti* (written למדתי, ילדתי). The epenthetic forms *lamádeti*, *yaládeti* are frequent in casual speech, and some younger speakers drop the epenthesis again in rapid speech (Wikipedia "Modern Hebrew phonology"; Bolozky 1997). The vowel is not written in unpointed spelling, so the two forms look the same.
2. **To avoid an impermissible onset cluster** at the start of a word, especially one beginning with a sonorant (*rC-*, *lC-*, *mC-*, *nC-*) or ending in a glottal (*Cʔ-*, *Ch-*): *rekhov* "street", *levanim* "white (PL)", *menahel* "manager", *ne'elam* "disappeared". The shva of the spelling is then said [e] (§3.7).

There is also a third, paradigmatic kind: the **segolate** nouns have an [e] (or [a]) between the last two root consonants in the singular, which goes away in other forms: *yéled* "boy" / *yaldá* "girl" / *yeladím* "boys"; *séfer* "book" / *sifrí* "my book". This is a matter of morphology, treated in [Nouns, adjectives, pronouns and numerals](https://md.mova.elamur.ai/free/en/docs/languages/he/04-nouns-adjectives-pronouns/).

## 3.7 The syllable and consonant clusters

The syllable template is **(C)(C)V(C)(C)**:

- **Onsets** may have two consonants at the start of a word, and they are common: *kfar* "village", *shmone* "eight", *ptsatsa* "bomb", *dli* "bucket", *tmuna* "picture", *gvina* "cheese", *spontani* (loan). Three-consonant onsets occur only in loans (*strategya* "strategy", *sprint*).
- **Constraints on onsets** (Bolozky 1997; Laufer 1999): a sonorant /l m n ʁ j/ as the **first** member of an onset is generally avoided in native words, and an [e] is inserted (*le-*, *me-*, *re-*, *ne-*). In fast speech the [e] is often very short or absent (*mnahel* for *menahel*), but the careful forms keep it; two **identical** consonants, or two that differ only in voicing, cannot start a word: **ddvarim*, so *dvarim* is fine but *te-tagid* has no cluster; a glottal /ʔ h/ cannot be the second member: **CʔV*, so *ne'elam* [neʔeˈlam], not **nʔelam*.
- **Codas**: a single consonant is the norm (*khat*, *tov*, *ken*). Two-consonant codas occur in a handful of native forms: the past 2FS suffix *-t* after a consonant (*katavt* "you (F) wrote", *halákht* "you (F) went"), *kosht* "truth" (archaic), and loans (*sport*, *park*, *test*, *kemp*). Speakers often insert a vowel or drop a consonant in the 2FS forms: *katavt* → [kaˈtavt ~ kaˈtaft ~ kaˈtavet] (§3.10).

**Glide formation.** Unstressed /i/ next to another vowel becomes [j]: *ra'iti* [ʁaˈiti ~ ʁaˈjti].

## 3.8 Stress

### 3.8.1 Where the stress falls

Modern Hebrew stress is **lexical and morphological**: it is not predictable from the sounds of the word, but it follows rules tied to word classes and affixes. The traditional grammar names two positions:

- **milra** (מִלְרַע, "below", i.e. on the last syllable): *tapúakh* → in this manual's notation *tapuakh* (unmarked), *shulkhan* "table", *khanut* "shop", *medaber* "speaks". This is the most common position in native words.
- **mil'el** (מִלְעֵיל, "above", i.e. on the second-to-last syllable): *yéled* "boy", *báyit* "house", *séfer* "book", *katávti* "I wrote", *éfo* "where".

**Antepenultimate** (third-from-last) and earlier stress occurs in **loanwords** (*univérsita* "university", *politika* → *polítika* "politics", *télefon*, *ótobus*~*otobús* "bus", *kafetérya*), in some **names** (§3.8.4), and in some **suffixed forms** (*shalákhtem* is not standard, but colloquial *katávtem* "you (pl.) wrote", normative *ktavtém*).

Bolozky (1997) and Schwarzwald (2011) note a general **drift towards penultimate stress** in colloquial speech, especially in nouns and adjectives used as names and in hypocoristics. An example is colloquial *kóva* for normative *kova* "hat" (Wikipedia, citing Bolozky).

### 3.8.2 Stress by word class and affix

| Category | Stress | Examples |
| --- | --- | --- |
| Segolate nouns (CéCeC, CáCaC, CóCeC, CéCaC) | penult | *yéled* "boy", *séfer* "book", *bóker* "morning", *ná'ar* "lad", *ókhel* "food", *ózen* "ear", *khéder* "room" |
| Other native nouns and adjectives | final | *talmid* "pupil", *mikhtav* "letter", *gadol* "big", *khatul* "cat" |
| Feminine *-a* | final | *yalda* "girl", *gdola* "big (F)", *tmuna* "picture" |
| Feminine *-et / -at* (segolate feminine) | penult | *rakévet* "train", *makhbéret* "notebook", *tizmóret* "orchestra", *mazkéret* "memento" (but *mazkira* "secretary (F)", final) |
| Plural *-im / -ot* | final (stress moves to the suffix) | *yeladim*, *sfarim*, *yeladot*, *khanuyot* |
| Dual *-áyim* | penult | *yadáyim* "hands", *shnatáyim* "two years", *ofanáyim* "bicycle" |
| Past 1SG, 2MS, 2FS, 1PL (*-ti, -ta, -t, -nu*) | penult (stem stays stressed) | *katávti*, *katávta*, *katávnu* |
| Past 3FS *-a*, 3PL *-u* | final | *katva*, *katvu* |
| Past 2PL *-tem/-ten* | final (normative) / penult (colloquial) | *ktavtém* / *katávtem* |
| Present F *-et* | penult | *kotévet* "writes (F)", *medabéret* |
| Present F *-a* (some binyanim and weak roots) | final | *roa* → *ro'á* "sees (F)", *kama* "rises (F)" |
| Future, infinitive | final (with stem) | *yikhtov*, *likhtov*, *tikhtevu* |
| Future with *-i / -u* | final | *tikhtevi*, *yikhtevu* |
| Possessive suffixes | on the suffix or the syllable before it | *sifrí* "my book", *sifréykha* → *sfaréykha* "your (M) books", *sifrénu* "our book" |
| Adverbs and function words | mixed, often penult | *éfo* "where", *mátay* "when", *láma* "why", *étmol* "yesterday", *máshehu* "something" (but *akhshav* "now", *ulay* "maybe" final) |
| Loanwords | as in the source, often penult or antepenult | *télefon*, *kafé* (final), *ótobus*, *demokrátya*, *muzéon* "museum" |

### 3.8.3 Stress shifts

**Suffixes move the stress.** When a stress-taking suffix is added, the stress moves onto it, and the vowels of the stem may reduce or change (vowel reduction in the pretonic syllables, and segolate stems taking their underlying form):

| Base | Suffixed | Gloss |
| --- | --- | --- |
| *davar* | *dvarim*, *dvar-* (CS) | thing / things / thing of |
| *yéled* | *yeladim*, *yaldi* | boy / boys / my boy |
| *séfer* | *sfarim*, *sifri*, *sifréykhem* | book / books / my book / your (PL) book |
| *khatul* | *khatulim*, *khatula* | cat / cats / she-cat |
| *mishpakha* | *mishpakhti*, *mishpakhot* | family / my family / families |
| *gadol* | *gdolim*, *gdola* | big / big (PL) / big (F) |

**Verb paradigms** shift stress between suffixed and unsuffixed forms (*katav* → *katva* → *katávti*), and the stem vowels change with it ([05. The verb: system, binyanim, tense and use](https://md.mova.elamur.ai/free/en/docs/languages/he/05-verbs/)).

**Construct state.** A noun in the construct state (*smikhut*) is **unstressed** or weakly stressed and forms one phonological word with the following noun, which carries the main stress. Its vowels reduce accordingly: *báyit* "house" → *beyt séfer* "school" (lit. house-of book), *mishpakha* → *mishpákhat Kohen* "the Cohen family" ([Nouns, adjectives, pronouns and numerals](https://md.mova.elamur.ai/free/en/docs/languages/he/04-nouns-adjectives-pronouns/)).

**Prefixes** are unstressed and never shift the stress: *ha-yéled*, *ba-báyit*, *u-va-sfarim*.

### 3.8.4 Stress in names, nicknames and loans

**Personal names** often have a different stress from the word they come from, usually penultimate (mil'el), and this tells name and word apart:

| As a word | As a name |  |
| --- | --- | --- |
| *shoshana* "lily, rose" | *Shoshána* |  |
| *yarden* "the Jordan river" | *Yárden* (girl's or boy's name) |  |
| *David* (biblical, formal) | *Dávid* (everyday name) |  |
| *Yosef* (biblical) | *Yósef*, *Yósi* |  |
| *Mikhal* (biblical, formal) | *Míkhal* |  |
| *rina* "song, joy" | *Rína* |  |

Place names show the same pattern: *Tel Aviv* (final), *Kfar Sába* (penult, against *saba* "grandfather"), *Kinéret* (Lake Kinneret), *Khéyfa* "Haifa". Hypocoristic and slang forms favour penultimate stress: *Ávi*, *Rámi*, *Khézi*, *Yóni*.

**Loanwords** keep a non-final stress, often the source language's: *telefon* → *télefon*, *ótobus* "bus", *kafetérya*, *banána*, *kompyúter* "computer", *polítika* "politics". Some loans have two variants, a normative or older one with final stress and a colloquial one with earlier stress: *otobús*/*ótobus*, *muzeón*/*muzéon* "museum", *sigaryá*/*sigárya* "cigarette".

### 3.8.5 Stress as a contrast

Stress distinguishes words, and it often distinguishes **grammatical forms** of the same root:

| Penult (mil'el) | Final (milra) |
| --- | --- |
| בֹּקֶר *bóker* "morning" | בּוֹקֵר *boker* "cowboy" (from *bakar* "cattle") |
| יֶלֶד *yéled* "boy" | יֵלֵד *yeled* "will give birth (3MS, in the phrase 'the mountain will give birth')" |
| אֹכֶל *ókhel* "food" | אוֹכֵל *okhel* "eats (MS)" |
| בָּאָה *bá'a* "she came" (past) | בָּאָה *ba'a* "comes (F)" (present) |
| קָמָה *káma* "she got up" (past) | קָמָה *kama* "gets up (F)" (present) / "standing grain" |
| שָׁבָה *sháva* "she returned" | שָׁבָה *shava* "returns (F)" |
| בִּירָה *bíra* "beer" | בִּירָה *bira* "capital city" |

The most important pairs are the last group: in **hollow (ע״ו/ע״י) verbs**, the past 3FS and the present FS are spelled identically, and differ only by stress (*bá'a* / *ba'a*, *káma* / *kama*, *sháva* / *shava*, *sára* / *sara*). Unpointed text cannot show this. A parser must use context, especially the tense of neighbouring verbs and time adverbs.

## 3.9 Assimilation

### 3.9.1 Voicing assimilation

Obstruents agree in **voicing** with a following obstruent. The assimilation is **regressive** (the first consonant takes the voicing of the second), and is usual in natural speech, especially among younger speakers (Laufer 1999; Wikipedia, citing Bolozky 1997):

| Spelling | Careful | Natural speech | Gloss |
| --- | --- | --- | --- |
| לסגור | *lisgor* [lisˈɡoʁ] | [lizˈɡoʁ] | to close |
| זכות | *zkhut* [zχut] | [sχut] | a right |
| חשבון | *kheshbon* [χeʃˈbon] | [χeʒˈbon] | bill, account |
| אבטיח | *avati'akh* | [afˈtiaχ] (when said *avti'akh*) | watermelon |
| כדור | *kadur* | (no cluster) | ball |
| מתגעגע | *mitga'agea* [mitɡaʔaˈɡea] | [midɡaaˈɡea] | misses (someone) |
| לקבל | *lekabel* | (no cluster) | to receive |
| עכשיו | *akhshav* [aχˈʃav] | (both voiceless) | now |
| גדולה | *gdola* [ɡdoˈla] | (both voiced) | big (F) |
| קבוצה | *kvutsa* [kvuˈtsa] | [kfuˈtsa] or [ɡvuˈtsa] | group |

Note: /v/ behaves as an obstruent for assimilation: [kfutsa] or [ɡvutsa] for *kvutsa*. The two outcomes show that some speakers also assimilate progressively in this cluster.

The rhotic /ʁ/ also devoices next to voiceless consonants ([χ̝]), and in clusters with /χ/ it may be hard to hear at all (§3.10).

### 3.9.2 Other assimilations

- **Nasal place assimilation**: /n/ becomes [ŋ] before velars (*anglit* [aŋˈɡlit] "English", *bank* [baŋk]). Assimilation of /n/ to [m] before labials also occurs in fast speech, but we have no held source with counted data.
- **Dental-stop fusion**: across a word or morpheme boundary, two dental stops /t d/ fuse into one consonant, except at the end of a prosodic unit (Wikipedia "Modern Hebrew phonology"): *at tagidi* → [aˈtagidi] "you (F) will tell".
- **The *-t-* of the hitpa'el prefix** metathesizes and assimilates in spelling and sound with a following sibilant, which is historical and fixed: *hitlabesh* "got dressed", but *histakel* "looked" (ס + ת → סת), *hishtamesh* "used", *hitstalem* "had a photo taken" (צ + ת → צט: written הצטלם, [hitstaˈlem]), *hizdaken* "aged" (ז + ת → זד, with voicing: [hizdaˈken]). These are morphological facts ([05. The verb: system, binyanim, tense and use](https://md.mova.elamur.ai/free/en/docs/languages/he/05-verbs/)).

## 3.10 Casual speech

Fast colloquial speech reduces and deletes a great deal. The following are common processes (Laufer 1999; Bolozky 1997; Dekel 2014; Wikipedia "Modern Hebrew phonology"):

- **Deletion of unstressed vowels** two or more syllables away from the main stress: *zot oméret* "that is (lit. this says)" → [stoˈmeʁet]; *be-séder* "OK" → [bˈseder] ~ [pseder]; *ani lo yodéa* "I don't know" → [ˈanlo ˈjodea] ~ [ˈaːjoˈdea]; *ma pitom* "no way" → [mapˈtom].
- **Deletion of /ʁ/ and its vowel before /χ/**: *be-dérekh klal* "usually" → [beˈdeχklal]. Before a pause, the full form is kept.
- **Deletion of /h/** everywhere except, in careful speech, at the start of a stressed syllable: *ha-yom* → [aˈjom], *ahava* "love" → [aaˈva], *ehiye* "I will be" → [eˈje].
- **Deletion of /l/ after an unstressed vowel**, typically in *lo* "not" and *el* "to": *ani lo yodéa* → [aˈni o ˈjodea] (Wikipedia, citing Bolozky).
- **Contraction of *et ha-*** (accusative marker + article) to [ta] or [et a]: *ra'iti et ha-séret* "I saw the film" → [ʁaˈiti taˈseʁet].
- **Contraction of *she-* and *ha-* sequences** with pronouns: *she-ani* → [ʃaˈni], *ma ata* → [maˈta] "what (are) you".
- **Glide formation** and vowel merger: *ra'iti* [ʁaˈjti], *ta'im* [taˈim ~ tajm].
- **Final /t/ deletion in the 2FS past** and simplification of the cluster: *halákht* "you (F) went" → [haˈlaχ(t)]; *katavt* → [kaˈtaft], [kaˈtavet] (the epenthetic form).
- **Reduction of the copular pronouns and frequent phrases**: *ma nishma* "what's up" → [manʃma]; *ein li* "I don't have" → [en li]~[ˈeli]; *ma ze* "what is it" → [ˈmaze].

These reductions are invisible in writing and very frequent in the spoken corpora. The Knesset treebank (IAHLTknesset) is a transcription of parliamentary speech, but it is edited into standard spelling, so it shows the syntax of speech without its sounds ([11. Modern Hebrew in Universal Dependencies](https://md.mova.elamur.ai/free/en/docs/languages/he/11-ud-analysis/)).

Example of a casual utterance:

```
אני לא יודע מה הוא רוצה
ani lo yodéa ma hu rotse   →   [ˈanlo ˈjodea ˈmau ʁoˈtse]
I NEG know.PRS.MS what he want.PRS.MS
'I don't know what he wants.'
```

## 3.11 Prosody and intonation

Systematic descriptions of Hebrew intonation are fewer than of segments. The points below are common ground in the descriptions (Laufer 1999; Wikipedia "Modern Hebrew phonology"; a corpus-based description of Hebrew intonation is on the «wanted» list), with the caveat that detailed corpus studies are needed.

- **Prosodic phrasing.** An utterance is divided into intonation units, each with one main pitch accent, usually on its last content word. The construct state, the proclitic prefixes and many function words are unstressed and attach to a neighbouring stressed word.
- **Declaratives** have a falling final contour.
- **Yes/no questions** differ from statements **by intonation alone**, with a final **rise** (or a rise-fall, ending high). There is no change in word order, and the particle *ha'im* (האם) is formal and mostly written:

```
אתה בא?
ata ba?   (final rise)
you.MS come.PRS.MS
'Are you coming?'
```

```
האם אתה בא?
ha'im ata ba?   (formal, written)
Q you.MS come.PRS.MS
'Are you coming?'
```

- **Wh-questions** (*mi* "who", *ma* "what", *éfo* "where", *láma* "why", *mátay* "when", *eykh* "how", *káma* "how much/many") usually have a **falling** contour, with a high pitch on the wh-word, or a rise-fall. A rising wh-question sounds like an echo question or a request for repetition: *ma?* "what?" (rising) vs. *ma ata rotse.* "what do you want." (falling).
- **Lists and continuations** have a rising or level contour on the non-final items.
- **Focus and contrast** are marked by moving the pitch accent to the focused word, and often by fronting it ([7. Syntax](https://md.mova.elamur.ai/free/en/docs/languages/he/07-syntax/)).
- **Discourse particles** such as *nu*, *ma*, *ze*, *kaze*, *ken*, *beseder* have their own characteristic contours, studied in conversation analysis. *Kaze* "like" is often unstressed and low in pitch (Maschler 2009; see [9. Pragmatics and politeness](https://md.mova.elamur.ai/free/en/docs/languages/he/09-pragmatics-politeness/)).
- **Speech rate** in Hebrew is fast compared with many European languages, and the reductions of §3.10 are part of normal, not sloppy, speech.

## 3.12 Summary for processing

For a text-based system such as Mova:

1. **Unpointed spelling hides most of the phonology**: vowels, stress, b/v k/kh p/f in most positions, shin/sin, gemination, shva. Stress is grammatically meaningful (§3.8.5) and invisible in writing.
2. **Spelling keeps distinctions that speech has lost** (ט/ת, ק/כ, ח/כ, א/ע/ה, ס/שׂ, ו/ב). Matching speech transcripts or user text with typos requires a phonological key that merges these letters.
3. **The alternations of §3.5** link written forms of one lexeme: ב/ו is not an alternation in spelling (both [v] are ב in the paradigm of a ב root), but כ and פ keep one letter for both values. A root-and-pattern analysis can work on letters directly.
4. **Spoken data** (transcribed) shows the colloquial forms: epenthetic forms, levelling and reductions appear in spelling only in informal writing (e.g. *בסדר* stays, but chat may write *בסדר*/*בסדרר*/*בסדד*...). Spoken registers have to be modelled separately ([8. Lexicon and register](https://md.mova.elamur.ai/free/en/docs/languages/he/08-lexicon-register/)).

---

## References

- Asherov, Daniel & Evan-Gary Cohen (2019). A phonetic description of Modern Hebrew consonants and vowels. *Brill's Annual of Afroasiatic Languages and Linguistics* 11(1) (cited via Wikipedia, not held locally)
- Bolozky, Shmuel (1997). Israeli Hebrew phonology. In Alan S. Kaye (ed.), *Phonologies of Asia and Africa*, vol. 1, 287–311. Eisenbrauns. (not held locally)
- Bolozky, Shmuel (1999). *Measuring Productivity in Word Formation: The Case of Israeli Hebrew*. Brill. (not held locally)
- Coffin, Edna Amir & Shmuel Bolozky (2005). *A Reference Grammar of Modern Hebrew*. Cambridge University Press. (listed in $RESEARCH_RAW/hebgram/grammars.tsv; not held)
- Dekel, Nurit (2014). *Colloquial Israeli Hebrew: A Corpus-based Survey*. De Gruyter Mouton. (not held locally)
- Gafter, Roey J. (2014). "The most beautiful and correct Hebrew": Authenticity, ethnic identity and linguistic variation in the greater Tel Aviv area. PhD thesis, Stanford University. (not held locally)
- Gafter, Roey J. (2016). What's a stigmatized variant doing in the word list? Authenticity in reading styles and Hebrew pharyngeals. *Journal of Sociolinguistics* 20(1), 31–58. (not held locally)
- Gafter, Roey J. (2021). Trilling as a sociolinguistic variable. *Sociolinguistic Studies* 14(4). (cited via Wikipedia)
- Khan, Geoffrey (2020). *The Tiberian Pronunciation Tradition of Biblical Hebrew*, 2 vols. Open Book Publishers. [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/) ; local: $RESEARCH_RAW/hebgram/open/khan-tiberian-pronunciation-v1-2020/ and $RESEARCH_RAW/hbobook-obp0163-khan-tiberian-pronunciation-tradition-biblical-hebrew/
- Laufer, Asher (1999). Hebrew. In *Handbook of the International Phonetic Association*, 96–99. Cambridge University Press.
- Maschler, Yael (2009). *Metalanguage in Interaction: Hebrew Discourse Markers*. John Benjamins. (not held locally)
- Ravid, Dorit (1995). *Language Change in Child and Adult Hebrew: A Psycholinguistic Perspective*. Oxford University Press. (not held locally)
- Schwarzwald, Ora (Rodrigue) (2011). Modern Hebrew. In Stefan Weninger, Geoffrey Khan, Michael P. Streck & Janet C. E. Watson (eds.), *The Semitic Languages: An International Handbook*, 523–536. De Gruyter Mouton. (not held locally)
- Wikibooks (Hebrew). *Dikduk ha-Ivrit ha-Hadasha* [Grammar of Modern Hebrew], pages ההגה/עיצורים [Consonants] and ההגה/תנועות [Vowels] (CC BY-SA). [he.wikibooks.org/wiki/דקדוק העברית החדשה/ההגה](https://he.wikibooks.org/wiki/דקדוק_העברית_החדשה/ההגה) ; local: $RESEARCH_RAW/hebgram/sa/wikibooks-he-dikduk-haivrit-hahadasha/wikitext/0005.wiki, 0006.wiki
- Wikipedia. Modern Hebrew phonology. [en.wikipedia.org/wiki/Modern Hebrew phonology](https://en.wikipedia.org/wiki/Modern_Hebrew_phonology) (accessed 2026-10-17)
- Zuckermann, Ghil'ad (2003). *Language Contact and Lexical Enrichment in Israeli Hebrew*. Palgrave Macmillan.
- Zuckermann, Ghil'ad (2005). Abba, why was Professor Higgins trying to teach Eliza to speak like our cleaning lady?: Mizrahim, Ashkenazim, prescriptivism and the real sounds of the Israeli language. *Australian Journal of Jewish Studies* 19, 210–231.
- Zuckermann, Ghil'ad (2020). *Revivalistics: From the Genesis of Israeli to Language Reclamation in Australia and Beyond*. Oxford University Press.
- [← previous 2. Script and spelling](https://md.mova.elamur.ai/free/en/docs/languages/he/02-script-and-spelling/)
- [next → Nouns, adjectives, pronouns and numerals](https://md.mova.elamur.ai/free/en/docs/languages/he/04-nouns-adjectives-pronouns/)
- [↑ back to Modern Hebrew](https://md.mova.elamur.ai/free/en/docs/languages/he/)