Archive for the ‘Accents and Vocalization’ category

Malone, Joseph L., “Wave theory, rule ordering and Hebrew-Aramaic segolation,” Journal of the American Oriental Society 91 (1971): 44-66

February 9, 2009

The so-called segholates developed from mono-syllabic nouns ending with a consonant cluster that resulted from the loss of final short vowels sometime during the late second millennium in common NWS. Interestingly, Hebrew and Aramaic form segholates differently. The Hebrew pattern is qVtVl (with stress on the first syllable) while the Aramaic pattern is qətVl (with stress on the final syllable). For instance:

*káspu > Heb kέsεf but BA kəsáf

*sípru > Heb séfer but BA səfár

*rúgzu > Heb róγεz but BA rəgáz

Malone suggests that the different patterns may be explained by differences in sequencing, namely the order of epenthesis and stress shift in the two languages.

It should be noted that there are also some deviations to these patterns in individual words. For instance, *šákmu ‘shoulder’ > Heb šəkέm, and *hílmu ‘dream’ > BA hélεm. Further, in Hebrew the pattern qətVl seems to dominate in nouns II-ˀ and III-y such as *diˀbu ‘wolf’ > zəˀév and *gadyu ‘kid’ > gəḏí

Based on comparative evidence and internal reconstruction, the following stages may be reconstructed for Hebrew segholate formation:

a) Final short vowels are apocopated.  

b) Stress shifts to the final syllable.

c) Final consonant clusters are broken by an epenthetic vowel  (usually e, but a if preceding consonant is a guttural).

d) Post-vocalic non-geminate non-pharyngealized consonants are spirantized (ie bgdkpt letters).

f) Unstressed e lowers to ε.

g) Stressed vowels undergo certain changes in quality: á umlauts to έ when separated by one consonant from following ε (á > έ /_xCε) while í and ú lower to é and ó respectively.

h) Closed-syllable ε lowers and backs to a when immediately followed by a guttural.

Following are some examples:

*sípru > sípr (a) > síper (c) > sífer (d) > sífεr (f) > séfεr (g) 

*báˤlu > báˤl (a) > báˤal (c) 

*pátḥu > pátḥ (a) > páteḥ (c) > páθeḥ (d) > páθεḥ (f) > pέθεḥ (g) > pέθaḥ (h)

Biblical Aramaic, however, follows a different sequence:

a) Final vowels apocopated.

b) Final consonant clusters broken by epenthesis (usually e, but o when following vowel is u).

c) Stress shifts to final syllable.

d) Spirantization of post-vocalic bgdkpt.

e) Open-syllable short vowels reduced to ə in pretonic position.

f) a lowers and backs to ă when preceded by a guttural.

h) Segholate é lowers and backs to á when a guttural or r immediately follows.

For example:

*gábru ‘man’ > gábr (a) > gáber (b) > gabér (c) > gavér (d) > gəvér (e) > gəvár (h)

*qúšṭu ‘truth’ > qúšṭ (a) > qúšoṭ (b) > qušóṭ (c) > qəšóṭ (e) 

Comparison with other Aramaic dialects suggests that changes a-e belong to the Common Aramaic phase of the language. For instance, in Jewish Palestinian Aramaic (JPA) the changes are similar to Biblical Aramaic except that stage h is modified:

j) Segholate é is replaced by á, unless a guttural immediately precedes. Segholate ó tends to be replaced by á.

For example:

*rúgzu > rúgz (a) > rúgoz (b) > rugóz (c) > ruγóz (d) > rəγóz (e) > rəγáz (j)

In Christian Palestinian Aramaic (CPA) there is no evidence of the vowel changes in steps f, g, or h. In Syriac, different vowel changes occur in stages f and h:

f) In word-initial ˀə the ə is replaced by u when the following vowel is u, otherwise it is replaced by e or a. Further, word-initial > ī.

h) Segholate é or ó lower to á when followed by a guttural or r.

For example:

*yárḥu > yárḥ (a) > yáreḥ (b) > yaréḥ (c) > yəréḥ (e) > īréḥ (f) > īráḥ (h)

Mandaic follows the same stages as Syriac a-f, shares Biblical Aramaic stage h, and then has several peculiarly Mandaic developments. 

When we compare Hebrew stages a-h with Aramaic, it is evident that they are very similar. Stages a and d are identical, while Hebrew stage b is identical to Aramaic stage c and vice-versa. Thus the difference between the two is the order of epenthesis and stress-shift. It seems likely that both languages are sharing in the same process of sound change, otherwise it is difficult to explain how the same changes could have occurred independently in the two languages. But how could stages b and c be reversed in the two languages?

It seems impossible if we conceive of each change as an instantaneous event, but not if it is conceived as a process through time. The common Northwest Semitic speech community was relatively unified for a time, but overlapping pre-Aramaic and pre-Hebrew groups may have begun to form. The apocopation of final vowels seems to be a common change to all dialects, but if stress shift began in the pre-Hebrew area (probably centering in the South) and epenthesis in the pre-Aramaic area (probably centering in the North), then by the time stress-shift reaches Aramaic, epenthesis will already have occurred, and by the time epenthesis reaches Hebrew, stress-shift will already have occurred. 

Thus we have an example of the wave theory of language change. This theory also helps explain the deviations noted at the beginning. The Hebrew qətVl forms can be explained as forms in which epenthesis preceded the stress shift. These forms then follow a process augmented by the Aramaic stages e-h, along with Hebrew stage g. For instance:

*gábru > gábr (a) > gáber (c) > gabér (b) > gavér (d) > gəvér (e) > gəvár (h)

Similarly, the Biblical Aramaic qVtVl forms can be explained as forms in which stress shift has preceded epenthesis. These forms then follow a process augmented by the Hebrew stages f and g . For instance:

*ṣálmu > ṣálm (a) > ṣálm (c) > ṣálem (b) > ṣálεm (f) > ṣέlεm (g)

With such a wave model, it is not unexpected that these features would reach peripheral dialects at different times, if at all. Thus it is interesting to note that many of Hebrew qətVl forms seem to have a rural provenance: dəváš ‘honey’, səváḥ ‘thicket’, šəlɔẃ ‘quail’, etc.

Finally, the II-ˀ and III-y forms can also be explained by a switch in the order of epenthesis and stress shift. Linguistic change often originates sporadically, becomes conditionally regular, and then unconditionally regular. These special forms may represent the first phases of the process, when only some word final clusters (namely ˀC and Cy) were subject to epenthesis. At this point, stress shift had not yet become regular. In addition, the nature of the epenthetic vowel seems to have differed from the later stage.


Gibson, John CL, “Stress and Vocalic Change in Hebrew,” Journal of Linguistics 2 (1966): 35-56

February 7, 2009

In this article, Gibson traces the relationship between stress placement and vowel quality/quantity through the history of Hebrew. He begins by dividing Hebrew into three main stages: proto-Hebrew (PH), Biblical Hebrew (BH), and Masoretic Hebrew dealing primarily with Tiberian Hebrew (TH). Jakobson and Halle have argued that vowel quantity and stress tend not to both have phonemic status at the same time in a language. Gibson’s thesis is that it was vowel quantity that  was phonemic in PH, but that the situation was reversed by BH with stress becoming phonemic instead. This results in quality replacing quantity as the more distinctive feature of vowels.

Gibson draws evidence for Biblical Hebrew from the first millennium until the first few centuries of the Christian era. This includes the consonantal text of the Hebrew Bible, the Dead Sea Scrolls, and the various Hebrew transcriptions in Greek and Latin. Masoretic Hebrew is represented by the vocalized manuscripts from the medieval period. These texts preserve the consonantal structure of the biblical text, but reflect a pronunciation maintained in the academy centuries after Hebrew has ceased to be spoken. Proto-Hebrew must be reconstructed on the basis of common Northwest Semitic from the second millennium as represented by Ugaritic, Amorite, and Amarna Canaanite. Gibson assumes that NWS was a relative unity during this period and that the division of Canaanite and Aramaic cannot be traced back beyond 1000 BCE.

The consonantal phonemes of BH and TH were identical, though gemination must have had phonemic status in BH as shown by the minimal pair kabed (Qal pf 3ms) : kabbed (Piel pf 3ms). PH can be reconstructed with six vowel qualities: /a/, /i/, /u/, /e/, /o/, and a second long a-vowel /a2/. This a2 seems to be somewhere in between an open back and open front a. The /e/ and /o/ vowels merge with the reduced diphthongs from /ay/ and /aw/ respectively, and /o/ also merges with the result of the Canaanite shift /ā/ > /ō/.  BH seems to have the same inventory of vowels. However, in TH there are nine vowel qualities. The seven full vowels are /a/, /ε/, /i/, /ɔ/, /o/, and /u/. Thus TH has added /ε/, and /a2/ has opened to /ɔ/. TH also adds the shewa /ə/ and the hateph-qamets. the other reduced vowels seem to be allophonic.

A general stress shift seems to have occurred in the transition period c1300-1000 BCE. This seems to have been a time of great phonological and grammatical change in NWS, most notably the loss of final short vowels. In PH stress seems to have followed Arabic pattern with vowel length being an important factor in determining stress placement. A long vowel attracted the stress when in the penultimate or some previous syllable. If there are only short vowels, then stress moves to the beginning of the word. For example, the change of PH *ˤa2lamu > TH ˤolam indicates that stress must have been on the first syllable during the Canaanite shift. In contrast *ˀata2 ‘you’ > ˀatɔ, suggesting that unstressed /a2/ did not participate in the Canaanite shift. 

With the drop of final short vowels and the resulting changes, the main stress was freed from vowel quantity. It now fell on the ultimate or penultimate syllable without regard for the vowel length or the type of syllable. The separation of stress from vowel quantity began the process of switching stress to a phonemic status instead of vowel quality. This also began a process of changes in vowel qualities as the prosody changed.

There seem to be two systems of vowel changes. In both /i/ > /e/ and /u/ > /o/. The difference is whether /a/ > /a2/ or not. In nouns like *dabaru, the loss of the final vowel closes the final syllable and the stress shifts to the ultimate so that /a/ > /a2/, *dábaru > dabá2r. However, in nouns like ˀiṣbaˤu where the first syllable is closed or contains a long vowel, generally /a/ does not shift even though the stress shifts to the ultimate, *ˀíṣbaˤu > ˀiṣbáˤ ‘finger’. 

In closed syllables, stressed short vowels seem to have been retained in BH, at least in the dialect underlying the Secunda: PH gabru > BH gabr (Sec γαβρ). However, Qumran orthography suggests a change in quality of /u/ and perhaps /i/ had already begun, as in שוחוד ‘bribe’ (1QIsaa 38:20) which suggests *šuḥdu > šuḥd > šəḥod or šoḥəd

Forms that ended with a geminate consonant lost the gemination when the final vowel elided. Here also /a/ does not shift to /a2/,  thus PH *libbu > BH leb, but PH *rabbu > BH rab. Based on Ugaritic, final /w/ or /y/ had already assimilated to the following short vowel at the PH stage.  Again, /a/ does not shift, thus PH pírī > BH péri > TH pərí, while PH gádī > BH gádi > TH gədí.  

Short vowels in open syllables preceding the newly shifted stress underwent no change from PH to BH. However, in TH /a/ and /i/ followed the normal path of development, /a/ > /a2/ > /ɔ/ and /i/ > /e/. Short vowels in propretonic positions reduced to zero in BH and thus /ə/ in TH. A /u/ in these positions is extremely rare, but seems to have gone unchanged. 

The verbal system is also quite complicated. Gibson reconstructs six verbal forms in PH: yiqáttal, yáqtulu, yáqtula, yáqtul, qtúl, and qátala. The first form, yiqáttal, seems to have merged with yáqtulu during the PH stage (note that scholars now generally reject a yiqáttal form for common NWS, see Terry Fenton, “The Absence of a Verbal Formation *yaqattal from Ugaritic and North-west Semitic,” JSS 15 (1970): 31-41). 

When the short vowels dropped, yáqtulu and yáqtula merged with yáqtul. The resulting verb had the semantic range of present-future (imperfect) indicative, past indicative, subjunctive, and jussive. The qátala form also has a bivalence between past indicative and future indicative in certain contexts. This seems like unbelievable multivalence for single verb forms, but fortunately, in BH and TH, stress often helps differentiate between the past and future meanings. Gibson suggests that this is some of the strongest evidence that stress has indeed become phonemic in BH and TH.

In the regular verb, the inflected forms had stress on the penultimate syllable when it was closed, and it remained there in BH and TH. One would expect the following short vowels to drop when in an open syllable, but in the majority of cases they did not. For instance PH *qátalta (2ms) > TH qɔ´taltɔ (orthographically קטלת). But in BH we have forms in the Secunda such as αφαχθ  suggesting qátalt while from Qumran we have אמרתה suggesting qátalta2. Gibson suggests that this reflects dialectal differences from PH, thus there was a PH qátalta and a qátalta2 form. Similarly he suggests a qátalti and a qátaltī (2fs), etc.  The forms with short vowels dropped, while the other vowels shifted /a/ > /a2/ and /ī/ > /i/, etc.

In the other forms, stress shifted to the penultimate syllable accompanied by the loss of final short vowels. Thus PH *qátala > BH qatál and PH *qátalat > BH qatála2. In TH, stress shifts to the ultimate for forms with final vowels, and the previously stressed vowels reduce to shewa, thus BH qatála2 > TH qɔtəlɔ´. 

Returning to the bi-valence of PH yaqtul(u) and qatala, the past value of yiqtol and future value of qɔtal is only maintained in the so-called waw consecutive constructions. Note that the wa-yiqtól form already disrupts the stress pattern by maintaining the full vowel on the copula, which the Masoretes also note by adding a dagesh to the following consonant. This should be considered a case of internal juncture. With the weqɔtál form, however, the waw is joined as usual. 

In the weqɔtál form, stress is distinguished in only two cases, the 2ms weqɔtaltɔ´, and the 1cs weqɔtaltí. The movement of stress has not left any perceptible changes in the vowel quality, and thus it seems that this movement of stress to the final syllable is a late innovation of the Masoretic period. In the wa-yiqtól form stress distinguishes I-y verbs as well as those of the form  sɔbab (> sab) and qɔm. Thus PH yéšib(u) > BH/TH yéšeb, but PH wa-yéšib(u) > BH wa-yéšib > TH wa-yéšεb. Also PH yáqum > BH yáqom > TH yɔ´qom, but PH wáyaqum > BH wa-yá2qum > TH wa-yɔ´qɔm.

Khan, Geoffrey, “Some parallels in linguistic development between Biblical Hebrew and neo-Aramaic,” Pages 84-105 in Semitic Studies in Honor of Edward Ullendorff. Edited by Geoffrey Khan. Leiden: Brill, 2005.

August 13, 2008

The neo-Semitic languages are interesting because they provide many historically documented examples of developments that parallel those in the classical Semitic languages. Even if these parallels are only typological, Khan argues that they have heuristic value for understanding the history of classical Semitic since we so often must appeal to hypothetical reconstructions. In this article he gives examples of such parallels between Biblical Hebrew and neo-Aramaic.

He begins with the issue of the BGDKPT consonants. There are many exceptions to the general rule that these letters are plosives after consonants, but fricatives after vowels. Some of these exceptions can be accounted for by general rules. For instance, in the case where the preceding word ends in a vowel, a BGDKPT letter beginning a word is plosive if the preceding word is marked with a disjunctive accent (Gen 12:11 וַיְהִ֕י כַּאֲשֶׁ֥ר), but fricative if it is marked with a conjunctive accent (Neh 4:1 וַיְהִ֣י כַאֲשֶׁ֣ר). However, what of cases such as מַלְכֵ֥י (Gen 17:16) and שִׁכְבַ֣ת (Ex 16:13), where a fricative occurs after a silent shewa? Conversely, there are cases where a plosive occurs after a vowel such as the 2fs perfect of final guttural vowels like לָקַ֣חַתְּ.

Khan suggests that this reflects that the ‘rule’ of BGDKPT is no longer operating in the phase of Tiberian pronunciation that is reflected in the Masoretic vocalisation. Where a vowel has elided before a consonant after the rule stopped working, the consonant generally remains a fricative. In a living language, further developments would occur to resolve the problem. For instance, it is possible that the fricative and plosive allophones would obtain independent phonemic status. Indeed there may be a few minimal pairs to suggest that this was beginning to happen, such as לָקַ֣חַתְּ, ‘you (2fs) took’, versus לָקַ֣חַת, ‘to take (inf)’.

In Aramaic there is a similar development. The distinction between fricative and plosive was originally conditioned by the preceding vowel, but eventually this rule ceased to operate. In literary dialects such as Syriac there is evidence that the two allophones began to attain phonemic status, such as garḇā ‘scabies’ but qarbā ‘scabious’. In neo-Aramaic this has come to its logical conclusion so that the variants have become independent phonemes which contrast in many minimal pairs: šāta ‘year’ – šāṯa ‘fever’; marta ‘saying’ – marṯa ‘mistress’ (don’t mix that one up! There was this marṯa I used to know…). Further, in verbal roots the plosive and fricative realizations of a root consonant no longer vary among the various inflections, but one is chosen which occurs consistently. For example, kṯw ‘to write’ (< *ktb): kaṯwa ‘she writes’, makṯōwə ‘to register’, kṯāwa ‘book’.

After next discussing the gutturals, Khan moves to issues related to vowel length. In Tiberian Hebrew, there was a tendency to lengthen vowels in stressed syllables. There seem to be two historical periods of lengthening, and between these two periods various changes in quality occurred such as the shift from ā to a rounded back vowel å. In the first period, a in an open syllable was also lengthened. Thus (disregarding the phenomenon of pre-tonic lengthening):

*dabáru > *dābāŕu > *dāḇāŕ

Next is the shift in quality:

*dāḇāŕ > *dåḇåŕ

On the other hand, the vowel of a segholate was not lengthened since it was in a closed syllable, and therefore did not undergo the shift of quality. However, after the epenthetic vowel was inserted, the first syllable was opened. Therefore the vowel was lengthened during the second period, but after the change in quality had already occurred. Thus:

*náˁru > *náˁr > *nāˁar

Parallels to these developments can also be found in neo-Aramaic dialects where an a vowel in an open stressed syllable was similarly lengthened. However, if an originally closed syllable becomes opened, the a does not immediately lengthen.

Khan also addresses the question of why there is a pataḥ in the final stressed syllable of the 3ms perfect קָטַל, but a qameṣ in the final stressed syllable of a noun דָּבָר. Should they not have followed the same pattern of development? That is:

*qaṭála > *qāṭāĺa > *qåṭåĺ

Somehow the usual rule of vowel lengthening was blocked in the verbal form *qaṭála. Some have suggested that this is because the final vowel of verbs was elided before the final vowel of nouns, thus the last syllable would be closed rather than open. However, from neo-Aramaic, Khan suggests that the short vowel comes from analogy to the rest of the paradigm. For instance, in neo-Aramaic the present verb is built on the participle qāṭil, which is inflected with a series of suffixes expressing the pronominal subject. Most of these suffixes begin with a vowel, which causes the vowel in the second syllable to be elided, closing the first syllable and shortening the long ā. For example:

qāṭǝl + a > qaṭla ‘she kills’
qāṭǝl + i > qaṭli ‘they kill’

What is interesting is that in some dialects such as Jewish Arbel, the long ā in the base 3ms form qāṭǝl is also short: qaṭǝl. There is no historical reason for this vowel to be short except for analogy. Thus, in Hebrew the the vowel in the second syllable of the 3ms perfect may also have remained short by analogy to the rest of the paradigm.

Knudsen, Ebbe Egede, “Stress in Akkadian,” Journal of Cuneiform Studies, Vol. 32, No. 1 (Jan., 1980), pp. 3-16.

May 6, 2008

The focus of contemporary linguistics on spoken language seems to have increased the interest in phonetics among scholars of dead languages. I myself am fascinated by the question of how much phonetic info we can squeeze out of these texts. In this article, Knudsen analyzes the letters of Hammurapi to determine whether the stress pattern in Akkadian is closer to Arabic or Hebrew and Aramaic. 

In prosody, stress is often divided between accent and pitch. In Semitic languages, either accent or pitch may be a feature of a word or phrase. When it is a feature of larger segments of speech, pitch is referred to as intonation or contour. In modern Semitic languages, stress is perceived similarly to Indo-European languages – stressed syllables seem more prominent than un-stressed syllables. This is realized in a complex composed of pitch, expiratory strength, sonority, and duration. Stress combined with intonation may mark sentence and clause boundaries such as the pausal forms of Tiberian Hebrew.

The general rules of stress in Akkadian are: 1) Final, open, circumflexed long vowels are stressed. 2) Otherwise stress falls on the first syllable with a long vowel or the first non-final syllable closed by a consonant (counting from the end of the word). If neither of these occur stress falls on the first syllable.

These stress rules are very close to classical Arabic, except for final stress on circumflexed long vowels such as ibnû or rabû which seems to be an innovation in the tradition of Assyriology. It seems very probable however that the Akkadian stress rules were derived from the European tradition of reading literary and classical Arabic, hence any claim that agreement between Akkadian and Arabic stress may provide evidence for proto-Semitic stress patterns seems to be circular in nature.

In 1879 Zimmern introduced a set of Akkadian stress rules more similar to Hebrew/Aramaic which included final stress in verbal forms such as ikšúd as well as final circumflexed vowels. J Aro also published a corpus of Old Babylonian texts with abnormal plene spellings which he argued represented a variable accentuation. The question is whether Akkadian stress should be seen as in the Arabic system, the Hebrew/Aramaic system, or fixed on the first syllable. There is some help in the Aramaic and Greek transcriptions of Akkadian texts, but the lack of a living tradition demands some caution.

Knudsen suggests that the letters of Hammurapi represent a well-defined corpus for the discussion of stress. The possible reflexes of stress will most naturally occur in the use of plene spellings from which the following cases are deduced:

a. Between i-class vowels (e, i, ī) and an a belonging to a closed syllable, plene spelling of a represents a phonetic glide as in ib-bi-a-am.

b. In final syllables plene writing may render the characteristic intonation of a yes/no question: in-na-ak-su-u as an interrogative, cf in-na-ak-su. Even short vowels can be treated this way.

c. In penultimate syllables plene writing may render “the intonation of emphasis” as in an imperative. 

d. Plene writing of long vowels in initial and final position is obligatory, in medial position it is variable. 

e. 3ms imperfect I-weak use plene writing in first syllable to distinguish from preterite, i-il-la-ak  versus il-li-ik. This does not represent stress, nor a *yiʼallak form since ia is preserved for I-y in OB. Variation of plene and defective writing occurs very often in common spellings.

Conclusive evidence for word stress in Akkadian comes from the existence of the opposition of plene and defective writing in the case of a vowel preceding an enclitic –ma and in monosyllabic words mu-u and šu-u against lu, la and ša. Here the grouping of defective and plene writing follows the categories of words that would be treated as independent carriers of stress and those that are grouped with following words as stress-units.

The stress in the Old Babylonian texts suggests differentiation between main and secondary stress, and the stress pattern seems to be zweisilbengesetz where the character of the final two syllables determine stress placement. Thus Knudsen concludes that Akkadian stress follows more closely to that of Hebrew and Aramaic then Arabic.


Dotan, Aron, “The Relative Chronology of Hebrew Vocalization and Accentuation,” Proceedings of the American Academy for Jewish Research, Vol. 48. (1981), pp. 87-99.

March 28, 2008

There are three known systems of Hebrew vocalization: the Tiberian, Babylonian, and Palestinian. Interestingly, while we have manuscripts which reflect stages of development in the Babylonian and Palestinian systems, there is no such history for the Tiberian system. At the same time, the Tiberian system achieves a level of uniformity and perfection that far surpasses the other two. In this article, Dotan addresses the question of whether the Tiberian system was invented as a complete system, or whether it was the result of a process of evolution.

Since Wickes’ study, the standard view has been that the vowels and accents were introduced in the 6th or 7th century CE. The points do not seem to have existed before the sealing of the Babylonian Talmud c500 CE, and Jerome mentions that the Jews of his day had no notation for vowels. At the other end, we know of two Masoretes who deal with reading the points from the end of the 8th and early 9th centuries, Pinḥas Rosh HaYeshiva and Asher ben Neḥemiah. Also, Moses ben Asher, from the second half of the 9th century, does not seem to be aware of the origin of the points which implies that they were already rather old.

Dotan suggests that the systems of accentuation and vocalization were originally different systems and that the accent signs preceded the vowel signs. In the Babylonian system, most of the signs are based on small Hebrew letters. For instance, a little ז corresponds to the accent זקף, while a small ו represents the vowel /u/. However, the sign corresponding to דגש, in the Babylonian system דיגשא, is not a small ד, but a small ז. The ד is instead used for the accent דחי, suggesting that the accents were established before the vowel points. There is similar evidence in the Palestinian system which uses dots and strokes, mostly above the line. The vowel signs all consist of two or three dots or strokes, while most of the accent signs are a single dot in various positions around the word again suggesting that the accents were developed first. Further, the earliest Palestinian Bible manuscripts contain mostly accents, with only an occasional vowel sign.

It seems clear then that the accents preceded the vowels in the Babylonian and Palestinian systems. There may also be hints of the same in the Tiberian system. Most of the Tiberian disjunctive accents are supra-linear signs which seem to be a continuation of the Palestinian and Babylonian systems. The Tiberian innovation is the infra-linear system which was introduced with the vowel signs and the conjunctive accent signs (with the exception of קדמא). It stands to reason that in the oral transmission of the text, accents and pauses were harder to preserve than vowels; and therefore, accent signs would have been introduced before vowel signs.

However, if accent signs were introduced before vowel signs, could it be that the it was the vowels which were introduced in the 6th or 7th century CE, but the accents were introduced earlier? Because it has been assumed that the vowels and accents were introduced at the same time, it seems that no one has investigated this question. In fact, both Jerome’s statements and the silence of the Talmuds are related to the נקודות, the vowel points. In fact, the Talmud refers several times to טעמים, the accents, but this was never taken to be written accents.

If indeed the accents are earlier, then Dotan suggests that the relationship between Hebrew and Syriac Masorah should be re-examined. It may not be the case that the Syriac signs are the earlier. Regardless, it is very likely that the Tiberian system was constructed gradually. Not only did the accents precede the vowels, but it seems that the accentuation system itself was developed in several phases.

Goetze, Albrecht, “Accent and Vocalism in Hebrew,” Journal of the American Oriental Society 59/4 (Dec, 1939), 431-459.

March 8, 2008

In this classic article, Goetze explores the relationship between Masoretic accentuation and vocalism. The standard approach had assumed that the accentuation of the MT was responsible for its vocalization (ie, whether a given vowel is lengthened, reduced, or even syncopated). However, Goetze suggests that over time the position of the accent has shifted so that some vowels may in fact be remnants of an older system.

The Masoretic system is interesting because it records phonetic information to aid in the pronunciation of the text when read aloud. That is, it preserves something of the prosody and phrasing of the spoken language. There are two important parameters that affect the quantity (and sometimes quality) of the vowel: 1) The position of the respective vowel in relation to the main stress of the word, and 2) The structure of the respective syllable – whether it is open or closed.

Goetze summarizes the basic system in the following table:


syllable-type initial open medial open closed
stressed ā ē ō ā ē ō a a1 o
pre-tonic ā ē ō ā ē ō a/i i o
2 b/f accent ə a/i i o
3 b/f accent ə a/i i o

However, there are a number of examples which cannot be explained by these rules:

1) A short vowel is reduced in a syllable occurring between the main and secondary stress, thus *dabarikā > BĀrəKĀ2. But why does *dabarakumū > barKΕΜ, with the vowel syncopated?

2) A short vowel is reduced in the pre-tonic syllable if it follows a “heavy” syllable, thus *siprakumū > sipKEM. But why does *malkatiya > malkāTĪ with the vowel lengthened? 

3) A short vowel in the pre-tonic syllable is syncopated internally after an open syllable, thus *dabarakumū > dəbarKEM and *damakumū > dimKEM. But why does *dabarikā > dəbārəKĀ and *damikā > daməKĀ with the vowel only reduced?

4) A short vowel is reduced in a word-initial pre-tonic syllable, thus zəRŌa(, but why is the vowel lengthened in šāLŌM?

5) How can the forms *dabaru and *qatalu yield BĀR with a long vowel in the stressed syllable, but qāTAL with a short vowel?

Allophonic changes in vowel length (and sometimes quality) often accompany stress. If the earlier patterns of accentuation were different than the Masoretic system, then it is possible that a given vowel may have been lengthened or reduced under the influence of an earlier accent. As the accent shifted positions, this new form would be the base for analysis under the Masoretic system proposed above rather than the original form. For example: *daBArīma > *dəBĀrim > bāRĪM.

This would also suggest that there is no such phenomenon as pre-tonic lengthening. That is, a pre-tonic short vowel is not lengthened under the influence of a following stressed syllable, but rather the length is a remnant of an earlier stress which has now shifted one syllable later.

The shift in stress was probably a gradual process of several shifts rather than a single sweeping change. Goetze identifies three main trends: 1) The main accent shifts one syllable toward the end of the word 2) In a word with a final secondary stress, the secondary stress and primary stress trade position such as *dəBĀrəkā > bārəKĀ 3) After the stress shift, the newly accented vowel lengthened in nouns such as *DĀbar > BĀR. However, in verbs the newly accented syllable remains short such as *QĀtal > qāTAL. This may be explained by the fact that the verb is typically followed by its subject with the two being treated together as a unit, preventing lengthening.

Using clues from the Amarna letters and early Hebrew orthography, Goetze proposes the following sequence of sound changes:

1) Before the Amarna period, syncope and early vowel reduction such as *baraKAtiyā > *barəKAtiyā

2) After the Amarna period, Philippi’s law (i > aíCCv) such as *littu > *lattu

3) Loss of short final vowels

4) Feminine ending –at > ah

5) –ah > ah(off-glide)

6a) Loss of certain long final vowels *LAhā > *LAH

6b) Stress shift to preserve other long final vowels *LAkā > *laKĀ

7) Simplification of diphthong –ah > ā

8) Switch of primary and secondary stress *BĀrəkā > bārəKĀ

9) General stress shift toward end of word



1. Shift of i > a here reflects Philippi’s Law

2. I have used Latin transliteration to better represent phonetics. Capital letters represent a stressed syllable. A macron (ā) indicates a long vowel. The reduced vowel, shewa, is represented as ə. Lastly, spirantized consonants are marked with an underline.

Aronoff, Mark, “Orthography and Linguistic Theory: The Syntactic Basis of Masoretic Hebrew Punctuation,” Language Vol 61 no 1 1985.

January 19, 2008

Aronoff begins by noting that modern linguistics tends to view spoken language as “true” language, while written language is a by-product. Still, the study of written language has often dominated the field. Aronoff argues that “written language is a product of linguistic awareness… any orthography must therefore involve a linguistic system.” In this article he examines the linguistic system of Masoretic accentuation – “A complete unlabeled binary phrase-structure analysis of every verse, based on a single parsing system.” The Masoretic system divided the text into verses, marked all the segmental phonological properties not marked by the consonantal text (vowels, dagesh, etc), and marked accents.

Aronoff describes Biblical Hebrew as a typical Greenberg V(erb)S(ubject)O(bject) language. VSO is the normal order, but the initial position is emphatic and SVO, OVS, and SOV also occur. Modifiers generally follow the VSO core. BH also has prepositions rather than postpositions; genitives, adjectives and demonstratives follow the words they modify; adverbs follow adjectives; WH-words are sentence initial; and relative clauses follow their head nouns.

Besides marking the position of stress, the accents also function as punctuation (Wickes 1881, 1887 is still the standard description of their use). Accents are either disjunctive or conjunctive. Each verse of the Bible is divided into two halves which can in turn be subdivided into halves and so on until no group of more than two words remains. Each accent clause is thus arranged into a hierarchical structure. Note that the Masoretes were not formally syntacticians, but they were interested in showing the sense of the text by marking the relationships among all the words in a verse.

Most interesting from his analysis is the separation of a topicalized element from the rest of the sentence by a major syntactical break. However, not all pre-verbal elements are treated as topicalized. For instance, when a subject pronoun precedes the verb it may or may not be separated from the sentence as topicalized. In nominal sentences the accentuation may also vary depending on if the first element is seen as topicalized. A further point of interest is that phrases introducing direct speech seem to be subordinated similar to adverbial phrases.

Lastly, Aronoff suggests that the Masoretic system does not seem to be based on recitation of the text, but rather the recitation must have become based on the accent system. This can be seen in the case where a word ending with a vowel is followed by a word beginning with a bgdkpt letter. Whether or not the letter is softened does not seem to follow the prosody of the phrase, but the hierarchical accentuation.If there is a conjunctive accent the letter is softened, if it is a disjunctive letter it remains hard. For instance, in both Judges 1:1 and 1:8 are sentences of the same syntactic pattern: V( S PP.

‏וַֽיִּשְׁאֲלוּ֙ בְּנֵ֣י יִשְׂרָאֵ֔ל בַּיהוָ֖ה‎ Jd 1:1b

‏וַיִּלָּחֲמ֤וּ בְנֵֽי־יְהוּדָה֙ בִּיר֣וּשָׁלִַ֔ם‎ Jd 1:8a 

In the second  case, בְנֵֽי־יְהוּדָה֙ is joined with a maqqef and thus there is no disjunctive accent between subject and verb, hence ב is softened. However, בְּנֵ֣י יִשְׂרָאֵ֔ל must be treated as two words. Therefore there is a disjunction between subject and verb, and ב remains hard. Aronoff gives other interesting examples as well.